<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Mary D&#39;souza on Medium]]></title>
        <description><![CDATA[Stories by Mary D&#39;souza on Medium]]></description>
        <link>https://medium.com/@MaryDsa?source=rss-5d4ad1bf66aa------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*RVM-n9JeUlS6rze0YN-hFw.jpeg</url>
            <title>Stories by Mary D&amp;#39;souza on Medium</title>
            <link>https://medium.com/@MaryDsa?source=rss-5d4ad1bf66aa------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 15:07:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@MaryDsa/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[How to Read a Job Posting Headline and Act On It]]></title>
            <link>https://medium.com/@MaryDsa/how-to-read-a-job-posting-headline-and-act-on-it-17327e595798?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/17327e595798</guid>
            <category><![CDATA[job-data-scrape]]></category>
            <category><![CDATA[hiring-headlines]]></category>
            <category><![CDATA[job-posting]]></category>
            <category><![CDATA[job-signals]]></category>
            <category><![CDATA[hiring-insights]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 12 Nov 2026 11:30:38 GMT</pubDate>
            <atom:updated>2026-11-12T11:30:38.607Z</atom:updated>
            <content:encoded><![CDATA[<p>When companies announce job cuts in the same week, the headlines usually blend them into trendy angles. But each one has its own reason, and the reason decides what the news means for you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LLiILXs7strLOhuJcW6GWw.png" /></figure><p>Start with what was actually cut, look for the team, the city and the level of seniority. A cut in one function in one office says very little about the whole market.</p><p>Next, check whether the same company is hiring somewhere else. Many announcements come with new openings in another team or another country. That points to a company moving its effort, and the work is still being done.</p><p>Then look at the kind of roles that remain open. If the new postings ask for different skills than the old ones, the company has redesigned the work. If the postings look the same as before, the company is probably pausing.</p><p>Finally, remember what a posting is. It shows that a company plans to hire, and it does not prove the role was filled. Treat it as a direction, and expect a few listings to disappear without explanation.</p><h3>Three Causes that Produce the Same Headline</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*h8RcVTMPm_sPqMSqFByYmg.png" /></figure><p><strong>Structural reorganisation:</strong> A company redesigns how a function works, merging teams, removing management layers or replacing a manual process with a tool, and the affected roles are closed in one place while new ones are defined in another.</p><p>In posting data, total volume may barely move, but the requirement mix changes. Titles that did not exist a year earlier begin to appear, and familiar titles return with different tools, certifications and responsibilities named in the advert. Tracking<a href="https://www.jobspikr.com/use-case/role-architecture-evolution/"> role architecture evolution</a> is how this pattern becomes visible.</p><p>For the hiring market, demand has not shrunk so much as changed shape. Candidates holding the retired skill set meet a soft market, while candidates holding the new one meet a tight market, often inside the same company.</p><p><strong>Capital reallocation:</strong> Money, and the hiring that follows it, moves from one business line, product or geography to another. The company is not retreating; it is relocating its bet.</p><p>In the data, postings fall in one unit or location and rise elsewhere inside the same employer, with similar requirement sets on both sides. The decline and the rise are close in timing, usually within weeks, which is why<a href="https://www.jobspikr.com/use-case/location-strategy-intelligence/"> location strategy intelligence</a> reads this pattern more reliably than a headcount table can.</p><p>For the market, pressure moves with the money. The destination sees a new competitor for the same talent pool, and pay in that pool tends to harden, while the origin sees supply loosen.</p><p><strong>Cyclical cost control:</strong> Leadership expects weaker revenue and trims broadly to protect margin. It is the cause most people assume by default, and the one that least often explains a reduction on its own.</p><p>Postings fall across functions, locations and seniority levels together, the requirement mix of whatever remains is unchanged, and the pattern tends to reverse when conditions ease.</p><p>For the hiring market, that means a temporary increase in available candidates and softer pay pressure, with no lasting signal about skills.</p><h3>Four Checks Across a Ninety-Day Window</h3><p>Sorting a reduction into one of these causes takes four checks, applied to the same employer across a ninety-day window: thirty days before the announcement and sixty days after it.</p><p>The first check is volume before and after. Count open postings on the day of the announcement, then again at day thirty and day sixty, using the preceding thirty days as the baseline. A drop that fades is a different thing from a drop that holds, and<a href="https://www.jobspikr.com/historical-job-posting-data/"> historical job posting data</a> is what makes the baseline available after the fact.</p><p>The second is the requirement to mix and compare the skills, tools and responsibilities named in new postings with those in the postings that were withdrawn. Look for new skill names that recur across postings, since one instance may be an experiment while a cluster is a decision. If they differ, the cause is more likely structural. If they match, the company is either relocating the work or pausing it.</p><p>The third is location concentration, check whether postings fell in one place and rose in another. A matched movement points to reallocation, while a uniform fall points to cost control.</p><p>The fourth is seniority distribution, reduction that thins one layer, typically the middle, while the others continue suggests reorganisation. A reduction that falls evenly across levels suggests cost control. The same work reappearing at a lower level in a new location suggests reallocation.</p><p>Most commentary stops at the announcement, and the weeks that follow go unexamined. Yet the sixty days after are where the cause shows itself. It is also why a count taken on the day of the announcement is the weakest evidence available: it shows the reaction but none of the consequences.</p><p>A cyclical reduction shows hesitation and then a slow return of postings. A reorganisation shows new titles appearing. A reallocation shows the destination filling. The announcement carries the company’s explanation; the following sixty days carry its behaviour.</p><h3>What Job Posting Data Don’t Mention Directly</h3><p>A widely noted pattern of job posting data records intent, however, no company posts about the completed hires(unless its stakeholder/board roles). An advertised role has been approved and opened, but it has not necessarily been filled, and some are withdrawn, quietly reopened or never filled at all.</p><p>Several limits follow from that. The same vacancy can appear on many boards and must be deduplicated before counting. Some hiring never reaches a public posting, since referrals and internal moves leave no trace. Coverage differs by country and sector, so a small sample can mislead.</p><p>Reposting is the subtler problem. A role withdrawn and published again under a slightly different title looks like a closure followed by a new opening, and counting it twice inflates both the fall and the recovery. Matching on requirements, location and employer, rather than on title and date, removes most of this.</p><p>A withdrawn posting may also reflect nothing more than routine removal of a stale listing. Read within these limits,<a href="https://www.jobspikr.com/job-data/"> job data</a> is best used for direction and timing rather than for headcount.</p><p>It tells you what an employer intends and when, which is exactly the part that announcements leave vague, and the part planning depends on.</p><h3>What a Planning Team Considers Before Next Announcement Lands</h3><p>We noted a three steps order taken by corporates.</p><p>First, within a day, pull the thirty-day baseline for the role families your organisation competes for, defined by skills rather than titles. A baseline taken after the news is already contaminated by it.</p><p>Second, run the four checks at day thirty and again at day sixty, and name the cause in writing. A named cause can be corrected later, whereas an unnamed assumption cannot, and<a href="https://www.jobspikr.com/use-case/competitive-hiring-intelligence/"> competitive hiring intelligence</a> keeps the employer-level view in one place.</p><p>Third, act according to the cause rather than the headline. Structural reorganisation calls for a review of skills and<a href="https://www.jobspikr.com/emerging-roles-skills-radar/"> emerging roles</a>. Reallocation calls for a review of location and pay in the destination market. Cyclical control calls for patience, using the temporary supply while it lasts, with a review at day ninety.</p><p>Record the classification beside the date, so that the next announcement can be compared with this one and the team builds a reference set of its own.</p><p><a href="https://www.jobspikr.com/use-case/competitive-hiring-intelligence/">Read the Full Analysis</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=17327e595798" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Job Postings Say About Workforce Readiness]]></title>
            <link>https://medium.com/@MaryDsa/what-job-postings-say-about-workforce-readiness-f6028b502434?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/f6028b502434</guid>
            <category><![CDATA[hiringcompany]]></category>
            <category><![CDATA[workforce-readiness]]></category>
            <category><![CDATA[data-hiring]]></category>
            <category><![CDATA[job-insight]]></category>
            <category><![CDATA[jobs]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 05 Nov 2026 10:30:37 GMT</pubDate>
            <atom:updated>2026-11-05T10:30:37.934Z</atom:updated>
            <content:encoded><![CDATA[<h4>Why The Gap Between What Companies Report And What They Advertise Is The Useful Part</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cxRkKKQvScwQDXyozGl-gw.png" /></figure><h3>The Problem With Asking People If They Are Ready</h3><p>Ask an organisation whether its workforce is ready for a new capability, and the answer will usually be a rating. Somewhat ready. Mostly ready. On track.</p><p>Those categories are self-reported. The people choosing them are also the people responsible for the outcome, and that tends to push answers in one direction. They cluster towards the optimistic end.</p><p>Other people’s research shows the pattern clearly. A 2026 Skillsoft survey of 2,000 workers, managers and executives found that 77 percent of leaders believed their organisations had set employees up for success on AI, while only 24 percent of employees felt fully equipped.</p><p>A 2025 Workera survey of 800 learning leaders and 800 employees at large US companies found that 63 percent of leaders expected their organisations to be fully AI-ready within two years. Only 22 percent of employees shared that view.</p><p>Skillsoft’s German results, reported by heise, showed 73 percent of managers calling their workforce well prepared.</p><p>We cite these as other teams’ work, not ours. They are surveys, and they measure perception. That is exactly the point. Each figure records what people said on one day, not what the organisation did afterwards.</p><h3>Why Self Assessment Drifts Optimistic</h3><p>This drift has ordinary causes. First, the person answering is usually close to the plan. Saying, ‘we are behind,’ can feel like reporting your own failure. Saying. ‘we are on track,’ feels safe, and often feels true.</p><p>Second, readiness is rarely defined. One leader hears “ready” as “we have a strategy”. Another hears it as “people can do the work today”. A rating scale hides that difference and averages it away.</p><p>Third, surveys ask about the whole organisation, and the answer comes from the part of it the respondent can see. Senior people see the pilots, the steering meetings and the approved budgets. They see less of the teams still working the old way.</p><p>Fourth, timing matters. A survey is a snapshot of confidence on a given day, often just after a good announcement.</p><p>Put these together and optimism is the default, even among careful people. The fix is not a better question. It is a second source that does not depend on anyone’s mood.</p><p>That second source has to be observable by others. If a colleague can open the same postings and reach the same reading, you have something you can argue about productively. If it lives in one person’s judgement, you are back to a rating.</p><h3>What An Organisation Says And What It Pays For</h3><p>A survey records how an organisation describes itself. A job posting records something different. It is a request for money, time and management attention, written in public and aimed at a specific person.</p><p>Nobody posts a role to signal optimism, a posting costs something, and it has to describe real work. That makes it a better witness to what an organisation actually intends to do.</p><p>Neither source is complete. Surveys capture belief, and postings capture spending intent. The useful moment comes when they disagree.</p><p>If a planning team rates itself “mostly ready” for a capability but its postings show it is still hiring junior generalists to cover the basics, something does not add up. If it rates itself “early” while quietly posting for operations and governance roles, it may be further along than it admits.</p><p>The divergence is the informative part. Agreement tells you little you did not already assume. Disagreement tells you where to look.</p><p>Five signals make that divergence readable. Think of them as five questions you put to the postings, not as a score. Each answer is a direction of travel, and the pattern across all five matters more than any single one.</p><h3>What A Posting Actually Contains</h3><p>It helps to be precise about what you are reading. A single posting carries more than a title and a keyword list.</p><p>There is the <strong>title</strong>, which tells you how the organisation labels the work. There are the <strong>duties</strong>, which tell you what the person will actually do. There are the <strong>requirements</strong>, which tell you what the organisation believes it cannot teach. There is the <strong>level</strong>, which tells you how much judgement is expected. There is the <strong>function</strong>, which tells you who owns the work.</p><p>There are also traces of <strong>timing and repetition</strong>. How long a role stays open, and whether it returns, says something about how hard the organisation finds it to fill.</p><p>Each of these maps to one or more of the five signals below. Reading them together is more reliable than reading any one alone.</p><h3>Signal One, New Titles Versus Absorption</h3><p>When a capability matters, it either gets its own role titles or gets folded into existing ones. Both are visible in postings.</p><p><strong>What it indicates:</strong> New titles suggest a distinct, funded capability. Absorption suggests the organisation is asking existing roles to take on the work, often without new budget.</p><p><strong>The common misreading: </strong>New titles are read as readiness. They can be relabelled versions of old jobs. Absorption is read as neglect, yet it can be healthy when a capability becomes part of ordinary work.</p><p>Look at what the posting describes, not what it is called.</p><p><strong>How to check it:</strong> List every posting in a role family that mentions the capability. Sort them into postings with a new title and postings where an existing title gained new duties. Track the ratio quarter by quarter.</p><h3>Signal Two, Functional Spread</h3><p>A capability can sit in one team or run across many. Postings show which.</p><p><strong>What it indicates: </strong>If the same requirement appears in finance, operations, marketing and customer teams, the organisation is spreading it. If it stays in one function, it is still a specialism.</p><p><strong>The common misreading:</strong> Wide spread is read as maturity. Sometimes it is only copied language, a phrase that moved from one template to the next. Check whether the surrounding duties changed, not just the keyword.</p><p><strong>How to check it:</strong> Count the functions where the capability appears, then read a few postings from each. Real spread comes with different duties in each function. Copied spread comes with identical sentences.</p><h3>Signal Three, Seniority Creep</h3><p>Over time, the same role can start asking for more years of experience, or a more senior title for the same duties.</p><p><strong>What it indicates.</strong> Creep often means internal supply is thin. The organisation cannot grow the skill quickly enough, so it buys experience instead. It can also mean the capability needs judgement before it needs volume.</p><p><strong>The common misreading.</strong> Creep is read as ambition. It is just as often a sign of scarcity or title inflation. Compare the duties, not only the level.</p><p><strong>How to check it.</strong> Take one role and compare the years of experience and level wording across successive quarters. Then read the duties to see whether the work itself changed or only the label did.</p><h3>Signal Four, Operate Roles Next To Build Roles</h3><p>Early on, postings are mostly about building. Later, roles appear for running, monitoring, supporting and governing what was built.</p><p><strong>What it indicates.</strong> A mix of build and operate roles suggests work has moved from pilot to production. Someone now owns the outcome after launch.</p><p><strong>The common misreading.</strong> Operate roles are read as success. They can equally point to fragile systems that need constant care. Ask whether the operate roles are growing faster than the build roles, and why.</p><p><strong>How to check it.</strong> Tag each posting as build, run or govern. Chart the three groups over time. A healthy shift moves weight from build towards run without the govern group vanishing.</p><h3>Signal Five, Requirement Language Becoming Specific</h3><p>Vague requirements say “familiarity with new tools.” Mature ones name the workflow, the outcome and the standard expected.</p><p><strong>What it indicates: </strong>Specific language suggests the organisation has learned what it needs. It knows which tasks the role will do and how it will judge them.</p><p><strong>The common misreading:</strong> Specificity is read as quality. A long wish list can be a recruiter’s habit, and short, plain language is not always weak. Treat specificity as one signal among five, never the verdict.</p><p><strong>How to check it:</strong> Read a sample of postings from the earliest and latest quarters side by side. Note whether tasks, tools, outcomes and standards became more concrete, or whether the length simply grew.</p><h3>A Worked Example</h3><p>Here is an illustration, not a real case. Imagine a mid-sized insurer that tells its board it is “mostly ready” to use automation in claims handling. The survey of department heads agrees.</p><p>Its postings over six quarters tell a more complicated story. In the claims family, no new titles appear. The automation language shows up inside existing claims analyst roles, which suggests absorption. The language appears in claims and nowhere else, so spread is narrow.</p><p>Seniority drifts upwards: The same analyst role moves from asking for two years of experience to four. Operate roles are missing, because nobody has been hired to monitor the automated work. And the requirement language stays general, asking for “comfort with automation”.</p><p>Alone, any one of these might be harmless: Consider these insights together, they say the organisation is relying on a few experienced people to carry a capability nobody owns. “Mostly ready” is a generous description.</p><p>That is the value of the check. It does not prove the insurer is unready. It shows where its confidence deserves a second look.</p><h3>Running The Same Check On Your Own Postings</h3><p>You do not need outside data to begin. Your own history already contains the signals.</p><p>Start with your ten highest-volume role families, the groups of roles you hire for most often. They give the largest sample and the clearest trend.</p><p>For each family, pull the postings from the last four to six quarters. Record how the five signals moved. Did new titles appear? Did the capability spread to other functions? Did seniority rise? Did operate roles join the build roles? Did the requirements get more specific?</p><p>Then place that picture beside what the same family reported in your latest readiness survey. Mark each family as aligned, ahead of its rating or behind it.</p><p>Do not stop at your own data. Setting the pattern against the wider market, through<a href="https://www.jobspikr.com/historical-job-posting-data/"> historical job posting data</a> for similar organisations, shows whether your pace is typical or unusual.</p><p>Take the disagreements to the people who own those families. They usually have an explanation, and the explanation is where the planning insight lives.</p><p>A team that does this for ten families has a defensible view of readiness. It is built on what the organisation did, not only on what it said.</p><h3>Setting Up The Comparison Table</h3><p>A simple table keeps the work honest. Use one row per role family, ten rows in all.</p><p>In the first column, put the family’s self-reported readiness rating from your most recent survey or planning review. In the next five columns, record the direction of each signal, up, flat or down, with one line of evidence. In the final column, mark whether the postings and the rating agree.</p><p>Keep the evidence short and specific. “Analyst title now asks for four years, was two” is useful. “Seniority seems higher” is not.</p><p>Resist the urge to score everything numerically at the start. A first pass in plain words surfaces patterns faster, and numbers can come later once the team agrees what counts as a change.</p><p>Two cautions help. Use the same time window for every family, so comparisons are fair. And ask someone who did not write the survey to read the table, because they will question the things you stopped noticing.</p><h3>Who Should Be In The Room</h3><p>The check works best as a short working session, not a report that arrives by email.</p><p>Bring workforce planning, because they own the readiness question. Bring talent acquisition, because they know why a posting reads the way it does. Bring one business leader from a family that looks misaligned, because they can explain what the data cannot.</p><p>Talent acquisition deserves particular attention, they often know that a requirement was copied from an older template, or that a role stayed open for months because the market was thin. Those details change how a signal should be read.</p><p>Keep the session to one hour and one decision. The aim is to agree which two or three families need a closer look before the plan moves forward.</p><h3>What Postings Cannot Tell You</h3><p>Postings show intent, not completed hiring. A role can be posted and never filled, filled by someone different from the plan, or withdrawn.</p><p>Some postings stay open for months, are repeated, or exist mainly to collect applicants. Counting them as hires would overstate progress.</p><p>The reverse error is possible too, a quiet stretch of postings may mean a team filled roles quickly, or filled them from inside. Absence of advertising is not proof of absence of progress.</p><p>They also miss everything that happens inside the organisation. Training, redeployment and internal moves rarely appear as public advertisements.</p><p>So read postings as a leading indicator, they show where money is pointed. Completed hiring, retention and performance show whether it landed. Both matter, and the second set comes from your own records.</p><p>Postings reflect roles advertised in public, so sectors that hire through networks or internal moves will look quieter than they are. For work tied to automation,<a href="https://www.jobspikr.com/use-case/ai-automation-impact/"> AI and automation impact analysis</a> can show where roles are being reshaped. Read each role family on its own terms.</p><p>Use postings to ask better questions, and use hiring and performance data to answer them. A<a href="https://www.jobspikr.com/use-case/strategic-workforce-planning/"> strategic workforce planning</a> process works best when it holds both.</p><p>It also helps to watch how roles themselves are changing. Tracking<a href="https://www.jobspikr.com/use-case/role-architecture-evolution/"> role architecture evolution</a> shows when a title no longer describes the work behind it, and an<a href="https://www.jobspikr.com/emerging-roles-skills-radar/"> emerging roles and skills radar</a> shows which new titles are spreading across the market.</p><h3>Four Data Pitfalls To Avoid</h3><p>Even a careful team can mislead itself with posting data. Four traps come up often.</p><ul><li><strong>Duplicates and reposts:</strong> The same role can appear several times across sources or be reposted weekly. Without cleaning, one vacancy looks like five and demand looks inflated. Deduplicate before you count anything.</li><li><strong>Inconsistent titles:</strong> “Analyst”, “Associate” and “Specialist” may describe the same job in different companies. Group roles by duties and normalised titles, not raw text, or your signals will be noise. A shared<a href="https://www.jobspikr.com/use-case/skills-taxonomy-intelligence/"> skills taxonomy</a> helps here.</li><li><strong>Seasonality:</strong> Many organisations hire in waves. Comparing a busy quarter with a quiet one can produce a trend that is only the calendar. Compare like with like, such as the same quarter in successive years.</li><li><strong>Thin samples:</strong> A role family with a handful of postings cannot support strong conclusions. Treat small counts as prompts for conversation, not as findings.</li></ul><p>Source quality matters as well. Reliable<a href="https://www.jobspikr.com/job-data/"> job data</a> with clear coverage and stable fields makes these problems easier to manage, because you can trust that a change in the numbers reflects the world and not the collection.</p><h3>Questions For The Planning Meeting</h3><p>If you take only one thing into your next planning meeting, take questions rather than conclusions. Postings are better at prompting good questions than at settling them.</p><p>Which of our capabilities are being built through new roles, and which through stretching existing ones? Where are we paying for experience because we could not grow it? Who owns the work after launch? Where do our postings describe something more advanced than our survey admits, or less?</p><p>Add one more question about the outside world. Are other organisations in our sector moving faster or slower on the same signals? Insight from<a href="https://www.jobspikr.com/jobspikr-insights/"> job market data</a> can show whether a pattern is yours alone or shared across the market.</p><p>Good questions lower the temperature of a planning discussion. They replace the debate about whether people feel ready with a shared look at what the organisation is actually doing.</p><h3>The One Comparison To Run Before A 2027 Plan Is Signed</h3><p>Plans are written in the language of confidence. Capabilities will be built, teams will be ready, hiring will follow.</p><p>Before yours is signed, run a single comparison. Take the three capabilities your 2027 plan assumes are ready or nearly ready. Beside each one, place the last two quarters of your own postings for the roles that would deliver it.</p><p>If the postings show new titles, spread across functions, with operate roles and specific requirements, the plan stands on evidence. If they show absorption, one function, rising seniority and vague language, the plan rests on a rating.</p><p>Neither result is bad news, and both are useful. The first confirms the plan. The second tells you where the budget, the timeline or the hiring assumptions need to change while changing them is still cheap.</p><p>Surveys tell you how an organisation feels about its readiness. Postings tell you what it is paying to find out. Read them side by side, and trust the gap. That is the whole method; begin here.</p><p>The reward for doing so is modest but real. You walk into the 2027 planning round knowing which of your readiness claims are backed by what you are buying, and which are still hopes. That is a stronger position than any rating can give you.</p><p><a href="https://www.jobspikr.com/for-enterprise/workforce-strategy-and-planning/">Read the full analysis</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6028b502434" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Two-Quarter Lag: Why Compensation Reviews Are Always Late]]></title>
            <link>https://medium.com/@MaryDsa/the-two-quarter-lag-why-compensation-reviews-are-always-late-4f3261886702?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/4f3261886702</guid>
            <category><![CDATA[compensation]]></category>
            <category><![CDATA[job-insight]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[hiring-insights]]></category>
            <category><![CDATA[job-listings]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 14:20:51 GMT</pubDate>
            <atom:updated>2026-10-29T14:23:33.302Z</atom:updated>
            <content:encoded><![CDATA[<p>Most compensation reviews begin when offers start getting rejected, but by the time the new salary bands are approved, the market has often moved on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Elir_Tj6BiM_GKhlMvRMJg.png" /></figure><p>Recruiters report declined offers, HR gathers benchmarks, and a proposal moves through finance and leadership. Months later, new bands arrive but the market may have changed again.</p><p>The problem is a process designed to respond to outcomes instead of noticing market movement early.</p><h3>How A Market Shift Becomes A New Salary Band</h3><p>A compensation change moves through several stages, each with its own questions and queue. A typical review might take:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NmYj5oxYo6A5fgex9GIImw.png" /></figure><ul><li><strong>Signal gathering: One to four weeks,</strong> recruiters flag declines and hiring managers report difficult searches.</li><li><strong>Diagnosis: Two to four weeks,</strong> HR checks internal salaries, benchmarks, locations, and role requirements.</li><li><strong>Benchmarking: Two to six weeks,</strong> analysts compare market rates, estimate budget impact, and draft ranges.</li><li><strong>Approval: Two to eight weeks,</strong> finance and business leaders weigh affordability, consistency, and scope.</li><li><strong>Implementation: Two to six weeks,</strong> revised bands are documented and added to recruiter guidance.</li></ul><p>These are illustrative ranges, not fixed service levels. The process can stretch across a quarter or two, leaving the original evidence stale.</p><h3>Why Offer Declines Make A Weak Early Warning System</h3><p>An offer decline seems like a clear signal until you try to combine it with other declines.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7V4V0hWoywiB7hYoyWcOCQ.png" /></figure><p>One candidate says the base salary is low; another accepts a counteroffer. Someone else prefers remote work, stronger equity, a different title, or a shorter commute. Recruiters record these reasons in notes or applicant tracking systems, often in inconsistent ways. A compensation issue can disappear among candidate-specific explanations.</p><p>Declines explain individual decisions, but not when the market started moving or whether pressure is building across a role family.</p><h3>Scarcity Changes The Shape Of Pay</h3><p>Pay pressure rarely rises evenly across an organisation. It gathers around skills and locations where employers compete for a limited pool of talent. AI-Core roles are one example: early-career compensation can move quickly when demand for practical AI engineering skills outpaces the supply of candidates who can demonstrate them.</p><p>Traditional salary structures group jobs by title, level, and function. But two early-career roles in the same grade may require very different skills. If one begins demanding model deployment, evaluation, or specialised AI infrastructure experience, its market can move ahead of the assigned band.</p><p>JobsPikr’s<a href="https://www.jobspikr.com/blog/compensation-intelligence-report-pay-strategy-2026/"> 2026 Compensation Intelligence Report</a> examines how salary movement varies by role, skill, and region. Our<a href="https://www.jobspikr.com/report/ai-salary-benchmark-2026/"> AI salary benchmarks</a> offer another starting point for understanding advertised pay for AI talent.</p><p>The answer is not to raise every technical band whenever AI appears in a job description. Identify which skill combinations are becoming scarce, then test whether those specific bands still match the market. Broad increases can raise costs in roles that have not moved while leaving the hardest-to-fill jobs behind.</p><h3>Watch Advertised Bands Between Reviews</h3><p>Annual benchmarking supports governance and long-term planning, but should not be the only external check.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wxN1Qi8QTH8LIN9xbass3A.png" /></figure><p>Instead, continuously track advertised salary ranges for selected role families , especially jobs that are repeatedly difficult to fill. Compare postings by geography, seniority, and required skills, and watch how ranges change over time.</p><p>A practical routine should:</p><ul><li>Define the role families and locations to monitor.</li><li>Compare similar jobs, including seniority and required skills.</li><li>Review advertised minimums, maximums, and midpoints monthly or quarterly.</li><li>Separate genuine movement from changes in the mix of jobs advertised.</li><li>Set thresholds that trigger a formal review.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cegkPMGYAwmEiSG1Wo5Ezw.png" /></figure><p>Live job posting data can provide this signal between scheduled reviews. JobsPikr’s<a href="https://www.jobspikr.com/blog/compensation-analytics-for-enterprises/"> compensation analytics overview</a> explains how posting-level information can be organised by role, geography, and time. The aim is to provide current evidence before approval discussions.</p><h3>What Posting-Level Salary Data Cannot Tell You</h3><p>Advertised salary data shows what employers say they are prepared to pay, not what employees ultimately receive. Many postings omit salary; others give broad ranges based on experience, location, or flexibility. Listings can remain online after a role changes, and duplicates can distort counts.</p><p>Base salary is not total compensation. Posting data may not reveal equity, bonuses, sign-on payments, benefits, retirement contributions, or the conditions attached to them. A posted range also does not confirm the final offer or accepted amount. JobsPikr’s<a href="https://www.jobspikr.com/historical-job-posting-data/"> historical job posting data guide</a> notes that postings represent advertised labour demand, not completed hires or exact headcount.</p><p>Use this evidence alongside internal offer outcomes, employee pay data, recruiter feedback, and established surveys. Each source answers a different question.</p><h3>Start The Review Before Offers Fail</h3><p>A review should not begin only when declines become impossible to ignore. By then, the organisation has already spent time losing candidates and diagnosing the problem after the fact.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cegkPMGYAwmEiSG1Wo5Ezw.png" /></figure><p>Define the roles where market movement would affect hiring most. Monitor advertised ranges, record each observation’s source and date, and set thresholds for investigation. Bring that evidence together with internal salary and offer data before proposing a new band.</p><p>It gives teams time to investigate before a hiring problem becomes a budget emergency for the business.</p><p>The goal is not to chase every market increase. It is to recognise when the market relevant to your hiring plan has shifted and act while there is still time. Compensation reviews will always take work. They do not have to start late.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4f3261886702" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Where Did It Reappear? Reading Role Migration in Posting Data]]></title>
            <link>https://medium.com/@MaryDsa/where-did-it-reappear-reading-role-migration-in-posting-data-09811d1bedb2?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/09811d1bedb2</guid>
            <category><![CDATA[role-migration]]></category>
            <category><![CDATA[job-posting]]></category>
            <category><![CDATA[overseas-jobs]]></category>
            <category><![CDATA[hiring]]></category>
            <category><![CDATA[hiring-process]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 13:23:02 GMT</pubDate>
            <atom:updated>2026-10-22T13:23:02.483Z</atom:updated>
            <content:encoded><![CDATA[<p>The most common error in reading a labour market shock is assuming that demand which disappeared from your view has disappeared from the market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*En0XVyqiKlFxoz6W9b0XjQ.png" /></figure><p>A policy change lands, whether a visa rule tightens, an incentive lapses or a compliance requirement shifts. Openings for a particular role fall sharply in the country where it happened.</p><p>The chart slopes downward, the commentary follows, and the conclusion that hiring has contracted is written before anyone asks the question that matters: did the work stop, or was the requisition posted somewhere else?</p><p>Posting data, read across borders and over weeks rather than inside one country and over quarters, answers that question more often than most planners expect.</p><h3>Unseen Pattern from Inside One Country</h3><p>Role migration is what happens when an employer’s need for a function persists while its chosen location for filling that function changes. The need does not vanish with the policy; it relocates.</p><p>From inside the originating country, though, a relocated opening and a cancelled opening look identical. Both leave the same mark on a local dashboard: one fewer posting.</p><p>Only an observer positioned above both markets can see the second posting that appeared elsewhere. And only if that observer is counting the same role in both places, which is harder than it sounds, since titles, spellings and reporting lines differ from one market to the next.</p><p>This is why migration stays invisible to anyone whose data stops at a national boundary, and why it is so routinely misread as decline.</p><h3>The Mechanics: Same Requirements, Different Market, Different Band</h3><p>A role that reappears rarely reappears unchanged. The requirement set, meaning the skills, tools, certifications and responsibilities named in the advert, tends to survive almost intact, because the work itself has not changed.</p><p>What changes is the market around it and the terms offered. Seniority often shifts: a position advertised as senior in an expensive market may return as mid-level in a market where the surrounding team is younger and the employer is prepared to invest in development.</p><p>The salary band moves with local pay structures, so comparing nominal figures across countries misleads. The fair comparison is where the band sits within each market’s own distribution for that role.</p><p>Titles drift too, which is why matching on title alone loses the trail while matching on skills and responsibilities keeps it.</p><p>This is the reasoning behind<a href="https://www.jobspikr.com/use-case/global-role-harmonization/"> global role harmonization</a>: standardising roles, skills and geographies so that a “data platform engineer” in one market and a “backend developer, data” in another are recognised as the same demand.</p><p>It is also why the quality of the underlying<a href="https://www.jobspikr.com/job-data/"> job data</a> matters more than its volume. A migration can only be traced if enrichment has already done the mapping.</p><h3>Why Official Statistics Lag and Postings Do Not</h3><p>Official employment figures count people already hired, in categories revised on fixed schedules, and they are published after collection, reconciliation and often correction.</p><p>A role that moves in March may not register as employment change in the destination country until several quarters later. A posting must become an offer, the offer must become a start date, and the start must then be surveyed and reported.</p><p>Postings sit at the very beginning of that chain. An employer expresses intent when it publishes a requisition, and that intent is visible the day the advert goes live.</p><p>With sources refreshed daily, a movement between markets shows within weeks, not quarters.</p><p>There is an honest limit here. Postings measure intent, not hires, and the same vacancy is frequently syndicated across several boards, so deduplication has to come before any counting.</p><p>Read as direction and relative scale rather than as headcount, posting data still gives the earliest reliable view of where demand is heading.</p><p><a href="https://www.jobspikr.com/historical-job-posting-data/">Historical job posting data</a> adds the baseline against which a change can be judged.<a href="https://www.jobspikr.com/use-case/always-on-talent-signal-monitoring/"> Always-on talent signal monitoring</a> keeps that judgement current rather than retrospective.</p><h3>The Data Centre Example: Demand Appearing Where Nobody was Looking</h3><p>Data centres offer a close parallel. When demand for computing capacity rises, the roles it creates do not appear where the technology companies are headquartered.</p><p>Electrical engineers, facilities technicians, cooling and network specialists and construction managers appear where land, power, water and grid connection are available. That is often in secondary regions that barely register in any national hiring narrative.</p><p>An analyst watching the large technology hubs would have seen flat or weakening demand for these trades. A handful of smaller regions, meanwhile, showed steady growth in the same skill clusters.</p><p>The demand was real, substantial and located somewhere the chosen frame did not reach.</p><p>Cross-border role migration is the same phenomenon at a different scale. In both cases the question is not whether demand exists, but where the viewer’s frame happens to be pointed.</p><h3>The Practical Method: Follow the Role Family, Not the Border</h3><p>The method follows directly from the pattern. Begin by defining the role family through its skills cluster rather than its title, so that variants are grouped by what the work requires.</p><p>Then track that family across every market where the industry or the employer hires, not only the one affected by the change.</p><p>Index each market against its own pre-change baseline. This lets a small market and a large one be compared on movement instead of size.</p><p>Next, examine the seniority mix and the pay band in the destination, because a rise at a lower level means something different from a rise at the same level.</p><p>At employer level,<a href="https://www.jobspikr.com/use-case/competitive-hiring-intelligence/"> competitive hiring intelligence</a> shows whether the same companies that reduced postings in one market began posting in another.</p><p>The decisive test is timing and proportion. If a decline in the origin market is followed within weeks by a comparable rise in the same skill cluster elsewhere, the demand has migrated. If both fall together, the market has genuinely contracted.</p><p>The two readings call for opposite responses, and only a cross-market view can tell them apart.</p><h3>What this Means for Location Strategy and Capacity Planning</h3><p>For location strategy, migration shows where the market is already moving before the arrival of that movement becomes expensive.</p><p>A market receiving relocated roles will soon see tighter supply and upward pressure on pay. Entering after the postings have climbed means competing against employers who got there first.</p><p><a href="https://www.jobspikr.com/use-case/location-strategy-intelligence/">Location strategy intelligence</a> and<a href="https://www.jobspikr.com/use-case/talent-scarcity-saturation/"> talent scarcity and saturation analysis</a> are the practical tools here. They show which markets still have available supply for a role family and which are filling up.</p><p>For capacity planning in a multi-region organisation, the risk is local interpretation. A regional leader who sees postings fall in their own country may reduce hiring plans, while the group’s total demand for that capability is unchanged and has simply been routed through another hub.</p><p>Receiving hubs, meanwhile, need managers, onboarding capacity and equipment before the roles arrive, not after. The origin market may also face attrition, because competitors posting the same requirements nearby are now recruiting from the same pool.</p><p>Holding a single, harmonised view of each role family across regions prevents each region from reading its own dip as the group’s decline. That is the purpose of<a href="https://www.jobspikr.com/use-case/strategic-workforce-planning/"> strategic workforce planning</a> and<a href="https://www.jobspikr.com/for-enterprise/workforce-strategy-and-planning/"> workforce strategy</a>.</p><h3>Three Questions to Ask When a Policy Change Lands</h3><p>The first reports will describe what fell. The more useful work is establishing what moved, and three questions do most of it.</p><ol><li><strong>Which role families does the change touch, and what was their posting volume in the eight weeks before it?</strong> Define them by skills, not titles, so the baseline is complete.</li><li><strong>Where did the same requirement sets begin to appear afterwards, in what volume, and at what seniority and pay band?</strong> This locates the destination and shows what changed on arrival.</li><li><strong>Does the rise elsewhere match the fall at home in timing and proportion, and what remains unexplained?</strong> The unexplained portion is the genuine contraction, and it is usually smaller than the first reading suggested.</li></ol><p>Answering them takes weeks of posting data rather than quarters of official statistics, which is the argument for treating postings as the first instrument and not the last.</p><p>With coverage across more than 100 countries and 70K+ sources, refreshed daily,<a href="https://www.jobspikr.com/"> JobsPikr</a> is built to follow a role family across markets.<a href="https://www.jobspikr.com/jobspikr-insights/"> JobsPikr Insights</a> turns that view into the skill, role and location signals this kind of reading depends on.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=09811d1bedb2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stock & Flow: Reading a Labour Market That Is Doing Two Things at Once]]></title>
            <link>https://medium.com/@MaryDsa/stock-flow-reading-a-labour-market-that-is-doing-two-things-at-once-a26a4dff7f5e?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/a26a4dff7f5e</guid>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[job-market-trends]]></category>
            <category><![CDATA[hiring-strategy]]></category>
            <category><![CDATA[hiring-trends-2026]]></category>
            <category><![CDATA[labour-market]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 15 Oct 2026 12:42:37 GMT</pubDate>
            <atom:updated>2026-10-15T12:42:37.048Z</atom:updated>
            <content:encoded><![CDATA[<p>A record layoff tally and a rising hiring outlook landed within weeks of each other this autumn, and almost every piece of commentary picked one and called it the market. Both readings are defensible, and both are incomplete. The two figures measure different things, in different populations, over different periods. A labour market can release a large volume of workers from a small number of employers while a much wider group of employers expands. Reading it correctly means asking what each number counts before asking what it means.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f_t-sEJqjvgFHfYftiIFyw.png" /></figure><h3>The Two Figures and How Each Was Built</h3><p>The layoff figure comes from Layoffs.fyi, a tracker of publicly reported job cuts at technology companies, compiled from media reports and crowdsourced submissions. As reported in September, it recorded 128,536 workers across 299 companies as of 10 September 2026. The full-year 2025 total was 122,606 workers across 278 companies, so the 2026 tally passed the previous year’s total with roughly a quarter of the year remaining. The figure counts announced reductions, worldwide, in the technology sector only.</p><p>Trackers do not agree with one another. Challenger, Gray &amp; Christmas data reported by Fast Company put announced cuts at U.S.-based technology companies at 165,925 through September 2026. Differences in scope, geography and counting rules explain the gap. A single tracker’s total is therefore a measurement, not a fact about the whole market.</p><p>The hiring figure comes from the Experis Tech Talent Outlook, published by Experis, a division of ManpowerGroup, and reported by<a href="https://www.staffingindustry.com/news/global-daily-news/tech-companies-seek-soft-skills-as-q4-hiring-holds-at-37"> Staffing Industry Analysts</a>. The survey polled 4,258 technology and IT services employers across 42 countries between 1 and 31 July 2026. Some 51 percent planned to hire in the fourth quarter and 14 percent planned to reduce staff, giving a net employment outlook of 37 percent. That compares with 35 percent in the third-quarter report and is unchanged from a year earlier. Net employment outlook is the share of employers expecting to hire minus the share expecting to cut, and ManpowerGroup notes that its survey results<a href="https://investor.manpowergroup.com/news-releases/news-release-details/global-hiring-outlook-strengthens-q4-workforce-transformation"> reflect employer sentiment at the time of data collection</a>.</p><h3>Stock and Flow: Why Both Can Be True</h3><p>In strict terms, both figures describe activity over a period. The useful distinction is how each is weighted and what it accumulates.</p><p>The layoff tally accumulates. It adds every announced separation through the year into a running total of displaced workers. It is weighted by headcount, so one reduction of several thousand roles moves it as much as dozens of small ones. Reporting on the year’s largest cuts has named a small group of very large employers, including Oracle, Amazon and Dell. Uber’s September announcement of roughly 3,300 roles, about 10 percent of its workforce, shows how single decisions can shift a monthly total.</p><p>The hiring outlook is a rate. It records the share of employers intending to add or reduce staff in the coming quarter. It is weighted by employer count, so a firm with 40 staff carries the same weight as a firm with 40,000. It says how widespread hiring intent is, not how many jobs it represents.</p><p>A concentrated release of workers from large employers and a broad, moderate appetite for hiring among many others do not contradict each other. A few employers can dominate one measure while many employers dominate the other. The labour market is not turning. It is bifurcating.</p><h3>Three Causes Inside the Layoff Number</h3><p>The headline tally combines at least three distinct drivers, and each carries a different implication.</p><p><strong>Structural reorganisation.</strong> Employers are redesigning work, not only reducing it. In ManpowerGroup’s global survey, 62 percent of employers planning to add staff cited changing roles and skills as the primary driver, and 32 percent were backfilling positions. Some reductions are one half of a reshaping in which roles are closed in one place and opened in another. The implication is that displaced workers and new vacancies may exist in the same skill families, and the task is redeployment and reskilling, not demand recovery.</p><p><strong>Capital reallocation toward AI infrastructure.</strong> Some employers are cutting operating cost to fund heavy<a href="https://www.jobspikr.com/report/ai-report-2026/"> investment in AI capacity</a>. Oracle, for example, has expanded its restructuring budget to roughly US$2.8 billion, driven mainly by severance, according to its regulatory filing as reported by Yahoo Finance. The cut and the investment belong to one decision. The implication is that demand rises for infrastructure, data and AI engineering skills at the same time that other roles are removed. The signal is about where budget is moving, not about falling demand overall.</p><p><strong>Ordinary cyclical cost control.</strong> Some reductions reflect routine discipline in a soft environment. Zillow and Etsy each stated that their 2026 reductions were not driven by AI adoption, and both reported revenue growth in the same period. ManpowerGroup’s survey likewise found that expected decreases were driven more by a soft economic environment than by AI or automation. The implication is that these cuts are the most reversible, and they tend to ease when conditions stabilise.</p><p>Treating the three as one number leads to one policy response where three are needed.</p><h3>A Four-Question Filter for Any Workforce Headline</h3><p>Before a workforce figure informs a decision, four questions are worth asking.</p><ol><li><strong>What is being counted?</strong> Establish whether the figure is an accumulation of events, a rate, an intention or a level, and whether it is weighted by people or by employers.</li><li><strong>Who is in the sample, and how was it collected?</strong> Identify the population, geography, sector, collection method and any known divergence from comparable sources.</li><li><strong>Over what period, and is it announced or realised?</strong> Check the collection window, the date of publication and whether the figure describes completed actions or stated plans.</li><li><strong>What mechanism produces the number?</strong> Ask which causes sit inside it and whether they point in the same direction.</li></ol><h3>A Worked Example</h3><p><strong>The layoff tally.</strong> (1) It is an accumulation of announced separations, weighted by headcount. (2) It covers 299 technology companies, drawn from media reports and crowdsourced entries, and it differs materially from Challenger’s U.S.-based count. (3) It runs year to date to 10 September and records announcements, not confirmed exits. (4) It combines the three causes above in unknown proportions. Conclusion: it is reliable evidence of concentrated restructuring at large technology employers. It is not evidence about the typical employer, and it does not measure the demand for labour.</p><p><strong>The hiring outlook.</strong> (1) It is a net balance of intentions, weighted by employer. (2) It covers 4,258 technology and IT services employers across 42 countries, and the 37 percent is a difference between 51 percent and 14 percent. (3) It was collected in July, before most of the September announcements, and it records intent, not realised hires. (4) Employers report a mix of expansion and changing skill needs, and technical demand is led by AI modelling and application development at 34 percent and AI literacy at 30 percent, with communication and collaboration the most sought soft skill at 41 percent. Conclusion: it is reliable evidence of broad and steady hiring intent. It is not evidence of the number of roles, the quality of the match, or the speed of hiring. ManpowerGroup’s own global data shows that 28 percent of employers report faster time-to-hire than in 2025, 41 percent report no change and 30 percent report slower hiring.</p><p>Applied together, the filter shows two accurate readings of two different populations.</p><h3>What a Planning Team Should Do Differently</h3><p>In a bifurcating market, the planning error is to forecast a single direction. Four adjustments follow.</p><p><strong>Segment before aggregating.</strong> Plan by employer size, function and skill family. A sector-level headline blends large-employer restructuring with broad mid-market hiring, and the two call for different decisions.</p><p><strong>Pair the stock with the flow.</strong> Track announced reductions alongside live demand. Posting-level data, such as that compiled by<a href="https://www.jobspikr.com/"> JobsPikr</a>, observes hiring as it happens, which narrows the gap between a July survey and a September layoff headline.</p><p><strong>Treat the released talent pool as a supply event.</strong> Large-employer reductions increase the availability of experienced workers. Teams hiring for scarce skills may find a wider window, while teams exposed to the same skill families should expect competition for the roles that are being created.</p><p><strong>Replace single forecasts with triggers.</strong> Define in advance which indicators would move a plan, such as posting volumes in priority roles, changes in time-to-hire, or the scale of a further quarter of announcements, and decide in advance what each would change.</p><p>The central point is simple. A cumulative count and a forward-looking rate answer different questions, and a market in transition will keep producing figures that appear to disagree. Analysts and planning teams who ask what each number counts, who it covers, when it was taken and why it moved will read the market more accurately than those who choose a side. The workforce is reorganising around new skills and new capital priorities, and the evidence for that sits in both figures at once.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a26a4dff7f5e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Three Tiers, One Phrase: Why AI Hiring Keeps Missing]]></title>
            <link>https://medium.com/@MaryDsa/three-tiers-one-phrase-why-ai-hiring-keeps-missing-326bf161baff?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/326bf161baff</guid>
            <category><![CDATA[tech-jobs]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[indian-job-market]]></category>
            <category><![CDATA[job-insight]]></category>
            <category><![CDATA[ai-hiring]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 08 Oct 2026 08:46:35 GMT</pubDate>
            <atom:updated>2026-10-08T08:46:35.938Z</atom:updated>
            <content:encoded><![CDATA[<p>The reported AI talent gap in India sits somewhere between 53 and 60% depending on the skill family, and almost none of that number describes a shortage of people.</p><p>It describes a shortage of a particular kind of person, hidden behind a phrase that has quietly stretched to cover three very different jobs. “AI-skilled” now appears on job descriptions for people who build models, people who run data and cloud platforms around them, and people who monitor what those models do in production. Hiring teams write a single requisition, post it once, and then find the same role still open two quarters later.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_TAwtw2NS6s7kQXpOXm8FA.png" /></figure><p>This article examines what the headline numbers measure, why the shape of the gap matters more than its size, and how to write a job description that says which kind of AI hire you are actually after.</p><h3>What the headline figures measure</h3><p>The analysis begins with the gap itself. Cloud shows the widest supply gap among the skill families tracked, at 55 to 60 percent, with GenAI close behind at about 53 percent. Cybersecurity sits much lower, at 25 to 30 percent, and is narrowing. Demand for GenAI talent is expected to cross 10 lakh roles this year.</p><p>A gap percentage compares the capability employers are asking for against the capability available at the depth they need. It is not a count of unemployed engineers or empty seats. A 53 percent gap in GenAI does not mean half of the country’s GenAI jobs have no applicants. It means that, for the work these roles require, qualified supply covers only about half of what the market is asking for.</p><p>Two further figures sharpen the point. Only about 16 percent of IT professionals hold AI skills. The upskilling pipeline is also wide at the entry point and narrow at the advanced end: around 20 lakh professionals have been upskilled, but only about 3 lakh have advanced AI skills. That is roughly one in seven.</p><p>The shortage, therefore, does not lie with professionals who have simply used an AI tool. It lies with professionals who can build, run and govern AI in production. These are distinct populations, and the gap figure can be interpreted only once they are separated.</p><h3>Why the shape of the gap matters more than its size</h3><p>A single headline percentage invites a single response: hire more, train more, pay more. However, a gap with a defined shape calls for a targeted response.</p><p>If the shortage were evenly distributed, a broad hiring push would be effective. It is not. It is concentrated in roles close to the AI stack, while roles further from it operate in a very different market. When a single job description is written for “an AI professional” and released into all of these markets at once, three problems follow:</p><ul><li>Genuinely scarce roles are written as routine roles and remain open.</li><li>Roles that are not scarce are priced as though they were, and budget is lost.</li><li>Candidates cannot determine what is being asked of them, so strong candidates pass over the posting while poorly matched candidates apply in volume.</li></ul><p>The remedy begins with a shared vocabulary. The question is not whether a role needs AI talent, but which tier of AI talent it needs.</p><h3>The three-tier framework: Core, Adjacent and Support</h3><p>One practical way to segment the market is into three tiers, defined by how a role relates to the AI system itself.</p><p><strong>Core</strong> roles build and own intelligent systems. Examples include GenAI engineers and machine learning engineers. These people decide how the model is designed, trained, evaluated and shipped.</p><p><em>Hiring implication:</em> This is the scarcest tier and the one where the gap is most real. Expect long search cycles, competitive offers and a candidate pool that assesses the employer as rigorously as the employer assesses them. The job description needs to describe the system they will own, the stack, and the problem, not just a list of frameworks.</p><p><strong>Adjacent</strong> roles apply AI within established technology functions. Data engineering, cloud engineering and cybersecurity engineering are the clearest examples. These engineers are not building foundation models, but their work decides whether AI can run reliably, cheaply and safely inside the business.</p><p><em>Hiring implication:</em> This is where most organisations actually need to hire, and where role definitions are blurriest. The candidate you want is a strong engineer in an existing discipline who has real, demonstrable exposure to AI workloads. Screen for depth in the home discipline first, and AI fluency second.</p><p><strong>Support</strong> roles focus on monitoring, validation and first-line resolution of AI-enabled operations. They keep AI-enabled processes running and catch what goes wrong.</p><p><em>Hiring implication:</em> This tier is the most accessible, and it is also where pursuing a premium profile does the most damage. Using AI tools well is the baseline here, not the differentiator. Hire for judgement, process discipline and communication, and train for the tooling.</p><p>No tier is unimportant. The difficulty is that three distinct hiring problems are being managed under a single label.</p><h3>Where the market pays: four Core, twelve Adjacent, zero Support</h3><p>The clearest evidence that the tiers behave differently shows up in pay. Across AI, Data, Cloud and Platform, and Cybersecurity, one analysis identified 16 premium roles, the roles where scarcity is translating into a salary premium. Four are Core, twelve are Adjacent, none are Support.</p><p>Two observations follow-</p><p>First, the premium does not belong only to the people building models. Three quarters of the premium roles sit in the Adjacent tier. The market is paying for people who sit close to the AI stack and combine that proximity with domain knowledge in data, cloud or security. Technical depth alone is not what sets the premium; where that depth sits relative to the AI stack does.</p><p>Second, the Support tier is absent from the premium list entirely. That is not a verdict on the value of the work. It is a signal that supply there is healthier, so employers are not competing on price.</p><p>For anyone planning headcount, the practical reading is this: if you are budgeting for a “Core-level” premium across every AI-labelled role, you are overpaying in one tier and probably still underpaying in another. The money should follow the tier.</p><h3>The salary curve, and why early-career premiums break band structures</h3><p>The pay data adds a second, less convenient finding.</p><p>A GenAI developer with 0 to 2 years of experience earns around Rs 11.2 lakh per annum, rising to Rs 34.5 lakh by 6 to 8 years. A data engineer moves from Rs 9.1 lakh to Rs 29.6 lakh over the same span. A technical support engineer, by comparison, goes from Rs 5.2 lakh to Rs 14.6 lakh. By the 6 to 8 year mark, GenAI, ML and data science roles sit around Rs 31 to 35 lakh, nearly double what AI-Support roles earn.</p><p>The entry point is more instructive than the end point. An entry-level GenAI hire earns a little over twice what a support engineer earns at the same experience, and about three quarters of what that support engineer will earn after six to eight years. The ratio between tiers changes little as professionals progress. The premium is not earned gradually over a career; it is present from the first day.</p><p>Traditional band structures are not designed for this. Most compensation frameworks assume pay grows with tenure and that entry grades sit close together across functions. When a 0 to 2 year hire in one tier prices at roughly double a peer in another, one of three things happens:</p><ul><li>The offer is capped to fit the band, and the strong candidate declines.</li><li>The offer is stretched, and an internal equity problem opens up with colleagues who have been there longer.</li><li>The role is quietly retitled or re-levelled to justify the number, and the job architecture slowly stops meaning anything.</li></ul><p>A further consequence concerns the hiring pipeline. Entry-level absorption in the technology industry is already under pressure as revenue growth outpaces hiring. If early-career premiums are concentrated in a small number of Core and Adjacent roles, the entry-level market becomes a two-speed market, and a single “entry-level AI hire” budget stops making sense.</p><p>A more robust approach is to build bands by tier rather than by title, and to decide in advance which tiers carry an early-career premium.</p><h3>A practical section: write a job description that states its tier</h3><p>If the core problem is one phrase doing the work of three, the remedy is to have the job description perform the sorting. A tier-stated job description is not longer; it is more specific in four places.</p><p><strong>1. Open with the relationship to the AI system.</strong> Say in the first two lines whether this person will build the system, run it inside an existing function, or monitor and validate it. “You will own the evaluation and deployment pipeline for our retrieval models” signals a Core role unambiguously. “You will make our cloud infrastructure ready to run AI workloads cost-effectively” tells an Adjacent candidate they are in the right place. “You will monitor AI-assisted workflows and resolve first-line issues” does the same for Support.</p><p><strong>2. Separate must-have from nice-to-have.</strong> For Adjacent roles in particular, the must-have is usually depth in the home discipline, and AI exposure is the differentiator. Presenting both as equal requirements produces a posting that no realistic candidate can satisfy.</p><p><strong>3. Name the work, not just the tools.</strong> A list of frameworks ages quickly and attracts candidates who match on keywords alone. A description of what the person will build, run or catch in their first six months attracts the right people and filters out the rest.</p><p><strong>4. Match the level and the pay conversation to the tier.</strong> If the role is Core, say so and be ready to discuss compensation that sits outside your standard bands. If it is Support, do not borrow Core language to look attractive. Candidates notice the mismatch, as do existing employees.</p><p>A simple test before publication: remove the terms “AI-skilled” and “AI-driven” from the posting. If the role still reads clearly, with its tier obvious from what the person will actually do, the description is ready. If it does not, the label was carrying meaning that the description should have carried.</p><h3>Validate the tier you actually need by watching how peers write the same role</h3><p>The final step is one that most teams omit: checking your tier assumption against the market before you open the requisition.</p><p>Internal intuition about whether a role is scarce is a weak signal. Recruiters tend to recall the roles that were difficult to fill, and hiring managers tend to describe the role they would like to hire. The market offers a better signal, because every competitor is publishing the same decision in public: how they scope the role, which skills they list as required, what seniority they ask for, and whether the posting keeps coming back.</p><p>A few things are worth reading across peer postings for the same role:</p><ul><li><strong>How the role is framed</strong></li></ul><p>Are peers asking for someone to build the system, or to integrate and operate it? That is your tier, as the market defines it.</p><ul><li><strong>Which skills show up as required versus preferred</strong></li></ul><p>If most peers treat AI exposure as a preferred skill for a data or cloud role, the role is Adjacent, and a Core-style search is overkill.</p><ul><li><strong>Whether the same role keeps being reposted</strong></li></ul><p>Repeated reposts and long posting durations are often the clearest sign that a role is stuck, either genuinely scarce or mispriced or miswritten.</p><ul><li><strong>The seniority mix</strong></li></ul><p>If most peer demand sits at senior levels, a junior-heavy search will struggle no matter how well you write the posting.</p><p>These are the same kinds of signals behind JobsPikr’s<a href="https://www.jobspikr.com/use-case/talent-scarcity-saturation/"> talent scarcity and saturation intelligence</a>, which uses repost frequency, posting duration, demand density by geography and seniority demand ratios to separate roles that are truly hard to fill from roles that only feel that way. For a wider context on how demand is shifting across roles and skills, the<a href="https://www.jobspikr.com/report/"> JobsPikr labour market reports</a> are a good place to start.</p><p>The real question is — which tier is this role, and what does the market currently show for that tier? the talent gap is real. However, a gap of this size, spread across three different kinds of work, will continue to undermine hiring plans until each requisition states which kind it is seeking</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=326bf161baff" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Aggregate Job Numbers Rarely Tell Planning Teams What to Do]]></title>
            <link>https://medium.com/@MaryDsa/why-aggregate-job-numbers-rarely-tell-planning-teams-what-to-do-1d54ae076764?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/1d54ae076764</guid>
            <category><![CDATA[market-report]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[labour-market]]></category>
            <category><![CDATA[planning]]></category>
            <category><![CDATA[market-data]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 01 Oct 2026 11:56:59 GMT</pubDate>
            <atom:updated>2026-10-01T11:56:59.626Z</atom:updated>
            <content:encoded><![CDATA[<p>A record tech layoff tally, a WHO warning about retiring doctors, and a media jobs roundup featuring a tree care trade association all landed within ten days of each other, and read together they describe one thing rather than three. Each is a story about a workforce, each arrived with a single prominent number, and each number invites a conclusion that the underlying data does not support. The layoff figure suggests a technology labor market in collapse. The WHO figure suggests a uniform shortage of doctors. The media roundup, skimmed quickly, suggests that editors are simply finding work again. None of those readings is entirely wrong, and none is safe to act on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a8RlWBTMlR51VYy-N88bRw.png" /></figure><p>For anyone who plans hiring, compensation, location strategy or training budgets, the practical problem is the distance between the number that makes the news and the number that changes a decision. The three stories below show how wide that distance can be, and what usually closes it.</p><h3>What Does the 2026 Tech Layoff Record Actually Measure?</h3><p>By 10 September, Layoffs.fyi had recorded 128,536 technology layoffs across 299 companies in 2026. That is already higher than the 122,606 recorded across 278 companies for all of 2025, according to Business Standard’s coverage published on 11 September. Oracle leads the tally with 21,000 affected employees, followed by Amazon at 17,267, Dell at 11,000, Meta at 10,400, Microsoft at 4,800 and PayPal at 4,760. The headline treats these as one phenomenon. They are at least three, and each carries a different message for hiring markets.</p><h4>Structural reorganisation</h4><p>Uber plans to eliminate about 3,300 corporate roles, roughly 10 percent of its workforce, in order to remove management layers and consolidate teams. Its management ranks are set to shrink by 20 percent, and the company has not presented the move as AI driven. Cuts of this kind remove a category of work, namely the layer of coordination between senior leaders and the people doing the task. The implication for hiring markets is lasting. Those roles are unlikely to return in the same form, and displaced middle managers will compete for a smaller pool of equivalent positions.</p><h4>Capital reallocation toward AI infrastructure</h4><p>Alphabet, Amazon, Microsoft and Meta are expected to spend about $725 billion on capital expenditure in 2026, up 77 percent from roughly $410 billion in 2025. Oracle, which announced cuts of about 13 percent of its workforce in June, is raising its spending on cloud infrastructure and AI over the same period. In these cases the money has not disappeared. It has moved from payroll to data centres and specialised computing. The implication is that demand for infrastructure, security and specialised engineering talent can hold steady or even grow while headcount falls elsewhere inside the very same company.</p><h4>Ordinary cyclical cost control</h4><p>PayPal is running a multi-year turnaround that targets $400 million in savings by the end of this year and at least $1.5 billion over the next two to three years. Its 220 job cuts in India, about 4 percent of its local workforce, belong to that programme. Other firms are still correcting the hiring surge of 2021 and 2022. This is the most reversible category. One industry executive quoted in the coverage expects some rehiring where aggressive cuts damage quality or client delivery, although he does not expect the workforce structure of 2021 to return.</p><p>The same executive estimated that only a quarter to a third of the wave reflects genuine task replacement by AI, with the rest being cost reset and restructuring. The entry-level evidence is sharper. Stanford’s 2026 AI Index, cited in the same article, shows that employment among software developers aged 22 to 25 has fallen nearly 20 percent since 2024. A single global total cannot tell a planner which of these forces applies to the roles they actually need to fill.</p><h3>What Does the WHO Health Workforce Report Say About Retiring Doctors?</h3><p>On 18 September the World Health Organization published its first National Health Workforce Accounts report on workforce levels and trends, the opening edition of an annual series built on official country data. Its central finding on ageing is that nearly one in four doctors, on average, is older than 55 and will reach retirement age within the next decade, based on the latest data from 122 countries. In high-income countries that share rises to nearly one in three. The report also finds that 67 countries will not have enough health workers to meet population health needs by 2030, and it projects a global shortfall of 11.1 million workers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*17QYh3vq8rp7Z7xgJDHpbA.png" /></figure><p>The same report contains a genuine success story. Health workforce density, which covers doctors, nurses and midwives, dentists and pharmacists, rose 52 percent over two decades, from 44.8 per 10,000 people in 2006 to 67.9 in 2025. There are now more than 70 million health and care workers worldwide. A planner who read only that figure would conclude the problem is being solved, and a planner who read only the shortage figure would conclude it is universal.</p><h4>Why does an average conceal the distribution that matters?</h4><p>Both conclusions fail because the average hides where the workers are. According to the WHO, the density of medical doctors in its European Region is 13 times higher than in its African Region, and the density of nurses and midwives is six times higher. Worker density remains strongly correlated with national income. Ageing adds a second layer, because retirement risk and rising service demand press hardest on high-income countries, while lower-income countries face a shortage of a different kind, which is too few trained workers to begin with.</p><p>The word “shortage” therefore describes two separate problems that call for separate responses. One requires succession planning, retention and flexible retirement design. The other requires training capacity and recruitment. An average of the two describes neither, and a workforce plan built on it would misallocate effort in both.</p><h3>What Do Media Job Listings Reveal About Where Editorial Skills Are Moving?</h3><p>Mediabistro’s hot jobs roundup, published on 11 September, featured three openings that share almost nothing in subject matter. One is a staff writer role at a geopolitics interview show hosted by a former head of MI6. Another is a managing editor post at Northeastern Global News, a university operation that behaves like a daily newsroom and does not require higher education experience. The third is a director of editorial and content strategy at the Tree Care Industry Association, a nonprofit membership body. Yet all three ask for the same underlying capability: the ability to turn specialised information into accurate, compelling journalism, backed by deadline editing, rigorous fact-checking and a working knowledge of how audiences consume content across channels.</p><h4>How does skill migration distort sector-level readings?</h4><p>Most labor market dashboards count jobs by the industry of the employer. Under that lens, a fall in postings from traditional publishers looks like a contraction in editorial work. Yet if the same editor now sits inside a university, a trade association or a creator-led media company, the work has not vanished. It has been reclassified. Read by employer industry, the data shows a loss. Read by role and skill, it shows relocation.</p><p>The reverse error is just as costly. A team that tracks only media employers will miss the organisations outside the sector that are quietly building serious content operations and competing for the same talent. The skill stays constant, the employer industry changes completely, and any reading taken at the sector level will describe the wrong market.</p><h3>Why Do the Aggregate Number and the Actionable Number Differ?</h3><p>The three stories share a structure. In each case the aggregate answers the question of how big the issue is, while the actionable number answers the questions of where it sits, which roles it touches and what to do next. The layoff total cannot say which skills are being redesigned and which are being replaced. The WHO average cannot say which country faces a retirement cliff and which faces a training gap. The media employer count cannot say which organisations are now hiring editors.</p><p>In all three cases the gap is filled by the same kind of evidence: role-level, region-level, live posting data. Job postings record what employers are asking for right now, by title, location and skill requirement, and they update continuously rather than annually. That makes them well suited to testing a headline against reality. If a layoff report claims that engineering demand is falling, current postings show whether infrastructure and security roles are still being advertised. If a workforce report flags an ageing profession, postings by region show where replacement hiring is already under pressure.</p><p>Teams that want to see this approach in practice can explore JobsPikr’s<a href="https://www.jobspikr.com/labor-market-trends-analysis/"> labor market trends analysis</a>, which shows how posting data can be read by role and region, and its<a href="https://www.jobspikr.com/for-enterprise/workforce-strategy-and-planning/"> workforce strategy and planning</a> resources for enterprise teams. Those who want to understand how such data is delivered can review the<a href="https://www.jobspikr.com/jobspikr-data-pricing/"> JobsPikr data and pricing options</a>.</p><h3>Which Three Questions Should Planning Teams Ask of Any Workforce Headline?</h3><p>Before acting on a workforce headline, a planning team can run it through three questions. They apply equally to layoffs, shortages and hiring roundups.</p><p><strong>What is inside the number?</strong> Ask what causes and categories the total combines. A layoff figure may contain structural redesign, capital reallocation and cyclical cuts, and each implies a different hiring outlook.</p><p><strong>Where does the number apply?</strong> Ask how the total is distributed across regions, employer types and seniority levels. A global average can sit comfortably beside a local crisis, and the most affected group is often the one the average describes least.</p><p><strong>What does current role-level evidence say?</strong> Ask whether live postings, filtered by title, location and skill, confirm or contradict the headline. If they do not match, the headline describes the past or the average, and the decision should follow the postings.</p><p>A number can be accurate and still be the wrong number for the decision at hand. The teams that plan well are the ones that treat the headline as the start of the question and go looking for the object that actually fits it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1d54ae076764" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[JobsPikr API: What to Look for in Modern Job Data in 2026]]></title>
            <link>https://medium.com/@MaryDsa/jobspikr-api-what-to-look-for-in-modern-job-data-in-2026-4a3bb2bf4760?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/4a3bb2bf4760</guid>
            <category><![CDATA[api]]></category>
            <category><![CDATA[job-api]]></category>
            <category><![CDATA[job-data-scrape]]></category>
            <category><![CDATA[jobs-insight]]></category>
            <category><![CDATA[jobs]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 24 Sep 2026 12:02:01 GMT</pubDate>
            <atom:updated>2026-09-24T12:02:01.749Z</atom:updated>
            <content:encoded><![CDATA[<p>Every job data evaluation starts the same way: with a spreadsheet comparing record counts. But the record count is close to the least useful number on that sheet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FCrbWymvF9OFtZT-LVi9Tg.png" /></figure><p>A vendor that reports 40 million listings and another that reports 25 million aren’t necessarily answering the question a buyer actually needs to ask:</p><blockquote><strong>How many of those listings are current, correctly labelled, and usable without a cleanup project?</strong></blockquote><p>Coverage numbers are easy to print and harder to verify. In 2026, with job data feeding pricing models, labour-market dashboards, and recruitment products, that gap between what a datasheet claims and what a pipeline actually receives has become expensive to ignore.</p><p>This article looks at what actually separates job data providers once the marketing language is set aside, and where the<a href="https://www.jobspikr.com/jobspikr-api/"> JobsPikr API</a> fits into that comparison.</p><h3>Why Coverage Comparisons Can Mislead</h3><p>A raw count of job listings tells a buyer very little without knowing how those records are counted.</p><p>For example, two providers could handle the same posting differently:</p><ul><li><strong>Provider A:</strong> Counts the same posting found on three job boards as three records.</li><li><strong>Provider B:</strong> Recognizes them as one job and counts them once.</li></ul><p>Neither approach is necessarily wrong but comparing their headline numbers without knowing the methodology creates an uneven comparison.</p><p>The same issue applies to the word <strong>refreshed.</strong></p><p>A vendor might consider a listing refreshed when:</p><ul><li>The source page has been re-crawled and the listing still exists.</li><li>A field, such as salary or location, has changed.</li><li>The listing has been checked according to a fixed schedule.</li></ul><p>These are different processes, even though they may all be described as a refresh.</p><p><strong>Before comparing vendors, ask two questions:</strong></p><ol><li>What counts as a unique job record?</li><li>What counts as a refresh event?</li></ol><p>The answers provide more useful context than the headline coverage number alone.</p><h3>Specialist vs. Generalist Job Data Providers</h3><p>Another useful distinction is whether a provider was built specifically around job data or whether job postings are one category within a broader data business.</p><p>This can affect three areas:</p><h3>1. Schema stability</h3><p>A job-data specialist has a strong reason to maintain consistent classifications for:</p><ul><li>Job titles</li><li>Seniority levels</li><li>Employment types</li><li>Skills</li><li>Locations</li></ul><p>A generalist provider may treat job data as one of many verticals, which can make schema consistency something buyers need to investigate more closely.</p><h3>2. Enrichment depth</h3><p>Extracting the text of a job posting is only the starting point.</p><p>Useful job data may also require:</p><ul><li>Identifying seniority from a job title</li><li>Separating required and preferred skills</li><li>Standardising company names</li><li>Mapping different job titles to common job families</li></ul><p>These require job-specific logic rather than simple page extraction.</p><h3>3. Source coverage</h3><p>Job data comes from job boards, applicant tracking systems, and company career pages. A specialist can tailor its collection strategy to these sources and their update patterns rather than applying the same approach across unrelated data types.</p><p>This doesn’t automatically make a specialist better than a generalist. It simply gives buyers another question to ask:</p><blockquote><strong>Who owns the job data schema, and how consistently has it been maintained?</strong></blockquote><h3>Freshness and History Both Matter</h3><p>Two properties of a job data feed matter more than a single coverage number:</p><ul><li><strong>Freshness:</strong> How quickly new or changed postings appear in the feed.</li><li><strong>History:</strong> How far back the archive extends and how consistently it was collected.</li></ul><p>JobsPikr refreshes its feed daily across <strong>70,000+ sources in 100+ countries</strong> and maintains an archive of <strong>1 billion+ historical records</strong>.</p><p>The combination supports different use cases.</p><p><strong>For recruitment products:<br></strong>Freshness matters because showing candidates jobs that have already closed creates a poor experience.</p><p><strong>For labour-market research:<br></strong>Historical depth matters because identifying changes in hiring demand requires a consistent baseline over time.</p><p>Think of it this way:</p><p>Freshness tells you what is happening now. History helps explain how you got there.</p><p>A feed with frequent updates but no historical depth has limited value for trend analysis. A <a href="https://www.jobspikr.com/historical-job-posting-data/">feed with extensive history</a> but infrequent updates can quickly become stale for operational use.</p><h3>Normalization Is Where the Real Work Happens</h3><p>Raw job data isn’t automatically ready for analysis.</p><p>Job titles alone show why.</p><p>Consider:</p><ul><li>“Sr. Software Engineer”</li><li>“Senior SWE”</li><li>“Software Engineer III”</li></ul><p>These titles may represent similar roles, but a system treating each as a completely different string can’t easily answer:</p><blockquote><strong>How many senior engineering jobs were posted last month?</strong></blockquote><p>JobsPikr maps titles into consistent job families and seniority bands, reducing the need for a separate mapping project. The same principle applies to skills.</p><p>A job description might list required qualifications in a paragraph of free text. For that information to become useful for filtering or matching, those skills need to be extracted into structured fields.</p><p>JobsPikr separates explicitly required skills from those listed as preferred.</p><h3>Location and company normalization matter too</h3><p>The same location might appear as:</p><ul><li>Bangalore</li><li>Bengaluru</li><li>BLR</li></ul><p>Similarly, a company can appear under multiple name variations.</p><p>Standardizing these fields makes it easier to:</p><ul><li>Compare hiring activity across locations</li><li>Track a company’s job-posting history</li><li>Build labour-market dashboards</li><li>Analyse demand for specific skills or roles</li></ul><p>This is the difference between receiving raw listings and receiving data that can be used immediately.</p><h3>Delivery and Integration</h3><p>Data quality only matters if an engineering team can actually use the data.</p><p>JobsPikr provides several ways to work with the dataset, including:</p><ul><li><strong>JSON and CSV formats</strong></li><li><strong>Direct API access</strong></li><li><strong>Scheduled file delivery</strong></li></ul><p>The API also provides endpoints for specific use cases.</p><h3>Search endpoint</h3><p>Teams can query listings using criteria such as:</p><ul><li>Job title</li><li>Location</li><li>Company</li><li>Posting date</li></ul><p>This avoids pulling the entire dataset and filtering it locally.</p><h3>Expired listing endpoint</h3><p>This identifies postings that have been taken down.</p><p>For recruitment products, that can help prevent candidates from seeing jobs that are no longer available.</p><h3>Aggregate count endpoint</h3><p>Sometimes a team doesn’t need every matching record. It simply needs to know <strong>how many</strong> listings meet a particular filter.</p><p>The aggregate count endpoint provides those summary figures without requiring the full dataset to be downloaded.</p><p>Each of these capabilities removes a piece of infrastructure an engineering team would otherwise need to build around raw job records.</p><h3>When a Pre-Built Dataset Makes More Sense</h3><p>An API isn’t automatically the right choice for every use case.</p><p>An API can make sense when a product needs:</p><ul><li>Live queries</li><li>Filtered data pulls</li><li>Recurring updates</li><li>Data feeding an active application</li></ul><p>But a pre-built dataset may be more practical for:</p><ul><li>A one-time market-sizing exercise</li><li>A historical research project</li><li>A report that won’t be updated regularly</li><li>A single large-scale analysis</li></ul><p>The key question is simple:</p><p><strong>Will the data be queried repeatedly?</strong></p><p>If the answer is no, paying for API infrastructure, rate limits, and integration work may add unnecessary overhead. For a one-time analysis, a file containing the same underlying data can provide a shorter path to the result.</p><h3>What to Ask Before Choosing a Job Data Provider</h3><p>Instead of comparing vendors only by the number of records they claim to have, build your evaluation around the questions that affect your actual use case.</p><h3>Ask about coverage</h3><ul><li>What counts as a unique record?</li><li>How are duplicate listings handled?</li><li>Which sources are included?</li></ul><h3>Ask about freshness</h3><ul><li>What exactly counts as a refresh?</li><li>How often are different sources checked?</li><li>How quickly do new listings appear?</li></ul><h3>Ask about history</h3><ul><li>How far back does the archive go?</li><li>Is historical collection consistent across the period?</li></ul><h3>Ask about normalization</h3><ul><li>Are job titles standardized?</li><li>Are skills extracted into structured fields?</li><li>Are locations and company names normalized?</li></ul><h3>Ask about delivery</h3><ul><li>Which formats are available?</li><li>Can the data be accessed through an API?</li><li>Are scheduled file deliveries supported?</li><li>Can specific records be queried without downloading the full dataset?</li></ul><p>These questions give buyers a more practical way to compare job data providers in 2026. The headline number still matters but <strong>freshness, history, normalization, schema stability, and delivery</strong> determine how much of that data can actually be used.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4a3bb2bf4760" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Does the August JobSpeak Report Actually Say?]]></title>
            <link>https://medium.com/@MaryDsa/what-does-the-august-jobspeak-report-actually-say-d9ff0caa6f03?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/d9ff0caa6f03</guid>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[job-report]]></category>
            <category><![CDATA[india-job-report]]></category>
            <category><![CDATA[tech-jobs]]></category>
            <category><![CDATA[hiring-2026]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Thu, 17 Sep 2026 12:08:05 GMT</pubDate>
            <atom:updated>2026-09-17T12:08:05.140Z</atom:updated>
            <content:encoded><![CDATA[<p>The Naukri JobSpeak index rose 14 percent year on year in August, to 3,028 from 2,664. It also fell 6 percent from July. Both facts are in the same report, and only one of them made the headlines.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ogZzE8QEXD95sayGWRroqw.png" /></figure><p>That is not unusual for a labor market index. A single number gets pulled out, rounded, and turned into a headline, while the report around it carries three or four other numbers that complicate the story. For anyone using hiring data to plan content, staffing, or sourcing decisions, the complication is usually where the real information lives. This piece walks through what the August index found, what the report itself flags as a distortion, why the methodology matters, and what a broader dataset can add when the goal is separating a seasonal blip from an actual shift in demand.</p><h3>What the index actually found</h3><p>The JobSpeak index climbed 14 percent year on year in August, driven largely by sectors outside information technology. IT and software services grew too, but at a slower 11 percent, and global capability centers added 10 percent. The interesting movement was elsewhere.</p><p>Auto and auto ancillary hiring rose 19 percent, one of the stronger showings in the non IT basket. Healthcare followed at 16 percent, and retail grew 15 percent, both consistent with sectors that scale hiring ahead of a busy consumer season. Accounting and finance rose 14 percent, and insurance added 12 percent, a steady but less dramatic gain compared to some of its peers.</p><p>Two categories stood out well above the rest. AI and machine learning roles grew 31 percent year on year, continuing a trend that has shown up in hiring data for several quarters now as companies across sectors build out data and automation capability. Recruitment sector hiring itself surged 109 percent, a figure large enough to suggest hiring firms were staffing up in anticipation of exactly the kind of demand the rest of the report describes.</p><p>Growth, however, was not universal: Media, dotcom, and entertainment hiring fell 14 percent, the steepest decline in the report. Education, teaching, and training dropped 8 percent. Legal services fell 7 percent, and travel and hospitality, including hotels, restaurants, and airlines, declined 4 percent. These four sectors pulled in the opposite direction from the broader non IT growth story, and any summary that only quotes the 14 percent headline misses that the picture underneath is mixed, not uniformly positive.</p><h3>The asterisk the report itself flags</h3><p>Here is where the report gets more interesting than its headline. Naukri’s own release cautions that the year on year comparison needs to be read alongside a shift in the festive calendar. A larger portion of the festive period fell in August last year, while more of it falls in September this year. That single calendar shift can distort a year on year number in both directions, depending on the sector being measured.</p><p>For a sector where festive season hiring is seasonal and front loaded, such as retail or consumer durables, having less of the festive period land inside August this year compared to last year could understate the real momentum, since some of the seasonal hiring push that showed up in the earlier month last year has simply moved a few weeks later. For sectors less tied to the festive calendar, the effect is closer to noise. This is not a minor footnote. It means the same 14 percent headline could be quietly flattering some sectors and quietly penalizing others, purely as an artifact of when a holiday period happens to fall on the calendar, not because underlying demand actually changed.</p><p>This is also why the month on month decline matters as much as the year on year gain. A 6 percent drop from July to August, sitting right next to a 14 percent year on year rise, is a reminder that the index is not describing steady, linear growth. It is describing a hiring cycle that moves with its own seasonal rhythm, and August landed on a downswing within that rhythm even as it beat last year’s August.</p><h3>The wider methodology point</h3><p>None of the numbers above should be read as a count of actual jobs filled. The JobSpeak index is built from new job postings and recruiter searches on one platform, Naukri’s own resume database. Naukri discloses this clearly, and it is worth repeating because it changes what the number can responsibly be used for.</p><p>The index excludes gig work, hyperlocal hiring, and campus placements entirely. It measures posting and search activity, not confirmed hires, so a spike in postings could reflect increased hiring intent, increased competition for a shrinking pool of candidates, or simply a change in how recruiters are using the platform’s search tools. All three would move the index in the same direction while meaning very different things for the labor market.</p><p>This is not a criticism of Naukri’s reporting. Every hiring index built on a single platform’s activity will have the same structural limits, and Naukri’s disclosure of them is exactly what allows a reader to use the number appropriately, as a directional signal rather than a precise measurement. The problem only arises when a headline strips the number of its context and presents it as a settled fact about how many people got hired in August.</p><h3>What a multi source posting dataset adds</h3><p>This is where a broader, cross platform job posting dataset earns its place alongside a single platform index rather than replacing it. A dataset that pulls postings from many sources at once, rather than one recruiter platform, changes what questions can be answered.</p><p>Coverage widens beyond one platform’s user base, which matters because different platforms skew toward different sectors, seniority levels, and company sizes. Deduplication across sources becomes necessary and, done well, gives a cleaner count of unique open roles rather than raw posting volume, which can otherwise be inflated by the same role appearing on multiple boards or being reposted by the same employer.</p><p>The most useful capability, though, is the one directly relevant to the festive calendar problem above. With postings tracked across sources on a weekly basis, it becomes possible to compare the exact same calendar weeks year over year rather than the same named month. That lets a seasonal calendar shift, like festive period timing, be separated from a genuine change in hiring demand, because the comparison is anchored to the actual weeks in question rather than to a month label that happens to contain a different mix of festive days each year. A single platform index built for a broad readership will rarely publish at that granularity. A dataset built for analysis can.</p><h3>The number most likely to survive the seasonal noise</h3><p>Given all of the above, which figure from this report is worth trusting the most as a signal of real demand rather than calendar timing?</p><p>Fresher hiring stands out. Hiring of candidates with up to three years of experience grew 15 percent overall in August, and within that group, fresher hiring in insurance specifically grew 37 percent. Fresher hiring is less tied to the kind of senior, planned hiring that shifts around a festive calendar, and a jump this large in a single experience band, especially concentrated in one sector, points toward an actual change in demand rather than a timing artifact.</p><p>The headline number will always be the one that travels furthest. The number worth building a decision on is usually a few paragraphs further down, next to the caveat that explains what it does and does not measure.</p><p>For more on how job posting data is collected and structured at scale, see JobsPikr’s<a href="https://jobspikr.com/job-data/"> job data</a> overview, and for a look at how that data moves from raw posting to usable dataset, see<a href="https://jobspikr.com/blog/how-jobspikr-data-pipeline-processes-job-data/"> how JobsPikr’s data pipeline processes job data</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d9ff0caa6f03" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Washing Is Real and Job Posting Data Is How You See Through It]]></title>
            <link>https://medium.com/@MaryDsa/ai-washing-is-real-and-job-posting-data-is-how-you-see-through-it-77afe890fb1e?source=rss-5d4ad1bf66aa------2</link>
            <guid isPermaLink="false">https://medium.com/p/77afe890fb1e</guid>
            <category><![CDATA[ai-washing]]></category>
            <category><![CDATA[2026-trends]]></category>
            <category><![CDATA[labour-market-report]]></category>
            <category><![CDATA[layoff-insight]]></category>
            <category><![CDATA[job-posting]]></category>
            <dc:creator><![CDATA[Mary D'souza]]></dc:creator>
            <pubDate>Wed, 09 Sep 2026 10:24:25 GMT</pubDate>
            <atom:updated>2026-09-09T10:24:25.134Z</atom:updated>
            <content:encoded><![CDATA[<p>Ask an investor covering this year’s AI attributed layoffs what they actually believe, and most will tell you the same thing. They cannot fully tell whether the AI is real or just the explanation being used this quarter.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eCewAimyND_fQ3DI-7usTA.png" /></figure><p>There arepeople who read transcripts for a living, who build models on headcount and margin, who talk to management teams every quarter. And even they cannot separate a genuine productivity shift from a convenient narrative. If the professionals tracking this full time are uncertain, the rest of us reading press releases about AI driven restructuring should be more skeptical.</p><p>The term for what is happening is AI washing. A company announces layoffs, attributes them to AI adoption, and the market accepts the framing because it fits the story everyone already believes. Nobody asks: Was this AI replacing work, or was this a company that was overhired in 2021 and needed a clean explanation for a cut it was always going to make.</p><h3>Why AI Washing Is So Easy to Get Away With</h3><p>AI washing works because it borrows credibility from a trend that is genuinely real. Nobody doubts that AI is changing how work gets done. PwC’s 2026 Global AI Jobs Barometer, which analyzed more than one billion job advertisements across 27 countries, found that jobs requiring specific AI skills are growing almost eight times faster than the overall jobs market. The average wage premium for workers with AI skills has climbed to 62 percent. Companies most exposed to AI are seeing faster headcount growth than their peers, not slower, at 52 percent versus 36 percent.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RHtakzCj_99UhFGSo1xbAQ.png" /></figure><p>That last figure is the one that should give AI washing skeptics pause. If AI were simply replacing headcount at scale the way some layoff announcements imply, the most AI-exposed companies would be shrinking, not growing faster than everyone else. Instead, PwC describes a two track labor market. Roles that AI professionalizes, where it acts as a force multiplier for experts, are expanding. Roles that AI democratizes, where it makes a job easier for a non-expert to do, are growing more slowly.</p><p>The micro explanation attached to any single company’s layoffs is often unverifiable from the outside, and convenient enough that boards, comms teams, and even some investors are content to let it stand.</p><h3>The Randstad and Hays Problem</h3><p>For years, investors who wanted a read on the health of the broader labor market without waiting for lagging government data leaned on staffing stocks as a proxy. Randstad, the world’s largest staffing company, and Hays, one of the largest recruitment firms in the UK, became informal barometers. If placements were up, the thinking went, hiring demand was strong. If staffing revenue was low, something was slowing down in the real economy.</p><p>That proxy is breaking down, and it is breaking down for reasons that have nothing to do with the underlying labor market. Citi Research has already cut estimates across the sector, flagging technology risk from platforms built specifically to route around traditional staffing firms, alongside broader margin pressure. Staffing companies are not just reading the labor market anymore. They are also absorbing disruption to their own business model from AI powered hiring tools and direct sourcing platforms. When Randstad’s share price moves, an investor cannot cleanly separate how much of that move reflects employer demand for workers and how much reflects employers routing around Randstad entirely.</p><p>There is a second, quieter distortion — staffing firm revenue is a lagging and filtered signal even in normal conditions. It reflects only the placements that happen to run through an agency relationship, in the geographies and sectors that agency covers, weeks or months after the underlying hiring decision was made. It was always an approximation, now, it is an approximation of a business that is itself being reshaped by the exact technology investors are trying to measure. None of this is a criticism of Randstad or Hays as businesses. It is a statement about what their stock prices can and cannot tell an outside observer about whether AI attributed layoffs are real. The signal has always been indirect and is getting more indirect by the quarter.</p><h3>What a Direct Signal Actually Looks Like</h3><p>If the goal is to know whether a company’s AI story matches its hiring behavior, the more direct source is the hiring behavior itself. Job postings are a company’s own words about what it needs, published in real time, before any agency, index provider, or earnings call gets involved. A company that is genuinely automating a function will show it in the postings. Roles in that function shrink or disappear. New postings shift toward the skills needed to build, manage, or oversee the AI systems doing the work. A company that is using AI as cover for an unrelated cost cut usually shows a different pattern. Headcount drops, but the roles it keeps posting for look unchanged, or the AI-skill postings it does publish are thin relative to the scale of the claimed transformation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-oOOZ_WYsAgBa3w7uRGCyg.png" /></figure><p>This is the layer PwC’s Barometer data starts to expose at the macro level, and it is the layer JobsPikr’s job posting data is built to expose at the company and sector level. Instead of asking what a company said in a press release, the question becomes what a company is actually advertising for, role by role, month by month, across the countries and sectors where it operates. That is a much harder story to fake, because it requires a company to keep publishing job ads that are internally consistent with the explanation it gave investors. Most AI washing does not survive that level of scrutiny.</p><p>JobsPikr’s new AI Labour Market Report 2026 was built around exactly this gap. It tracks job posting volume, skill requirements, and wage signals across markets to show where AI adoption is genuinely changing hiring, and where the AI explanation is doing more narrative work than economic work. And sits closer to PwC’s finding that AI-exposed entry-level roles are seven times more likely to now require senior-level skills like judgement and leadership, a shift that only shows up if someone is actually reading the postings rather than the press release. The full methodology and findings are available at<a href="https://www.jobspikr.com/job-data/"> jobspikr.com/job-data/</a>.</p><h3>Reading the Signal</h3><p>A staffing stock tells an investor something about aggregate agency-mediated placement activity, filtered through a business that is itself under pressure. Job posting data tells an investor what a specific company is actually asking the market to hire for, unfiltered, in the company’s own language.</p><p>The investor who opened this piece uncertain about how much of this year’s AI narrative is real is asking the right question. The answer is not going to come from a quarterly staffing report or a well-produced press release. It is going to come from looking directly at what companies are posting, comparing it to what they are claiming, and treating the gap between the two as the actual signal.</p><p>That is the case for job posting data over proxy signals, and it is the case for looking past the AI washing headline to the hiring record underneath it. For a closer look at how this data is collected and structured across industries, visit<a href="https://www.jobspikr.com/"> jobspikr.com</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77afe890fb1e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>