<?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 Praveen Gupta Sanka on Medium]]></title>
        <description><![CDATA[Stories by Praveen Gupta Sanka on Medium]]></description>
        <link>https://medium.com/@praveenguptaz?source=rss-412e5987a786------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*7iBZ-mqNb11G2ElF.jpeg</url>
            <title>Stories by Praveen Gupta Sanka on Medium</title>
            <link>https://medium.com/@praveenguptaz?source=rss-412e5987a786------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 10:49:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@praveenguptaz/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[A/B Testing — Gold Standard of Measurement]]></title>
            <link>https://medium.com/@praveenguptaz/a-b-testing-gold-standard-of-measurement-3c4d2164d1c2?source=rss-412e5987a786------2</link>
            <guid isPermaLink="false">https://medium.com/p/3c4d2164d1c2</guid>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[a-b-testing]]></category>
            <category><![CDATA[data]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Praveen Gupta Sanka]]></dc:creator>
            <pubDate>Tue, 24 Dec 2024 21:13:30 GMT</pubDate>
            <atom:updated>2024-12-24T21:13:30.101Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9rINKHOG_Bc-QoNoIPQ8qQ.jpeg" /></figure><h3><strong>A/B Testing — Gold Standard of Measurement</strong></h3><p>A/B testing is a common industry practice to test and roll out changes for a product. This article summarizes the key characteristics of running A/B Tests and common pitfalls.</p><p>Have you ever noticed how the same app/website offers a completely different experience for you and someone else, like your friend or partner? This difference often stems from one of you being part of an “experiment”. Companies frequently run experience-changing tests — known as A/B tests — to improve their products.</p><p>Take Gmail, one of the most widely used apps in the world, as an example. In my account (see the image above), there’s a “Try Gemini” icon prominently displayed in the header bar. However, my friend’s account (image below) doesn’t show this option. Until recently, I hadn’t seen it either. This suggests that I might be in the “treatment group” of an A/B Test and my friend might be in the “control group”. The experiment could be testing the following hypothesis: Does exposing users to this button result in a higher awareness/adoption of the Gemini tool.</p><p><strong>Is anyone wondering “What is Control, Treatment, Hypothesis and Experiment?”, then this Post is for you as it will help you build the A/B testing concepts.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*N0A7QqLLcGpw03Ju" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*5lCEtM4W5E2atmN2" /></figure><h3><strong>What is A/B Test?</strong></h3><ul><li>A/B testing is a powerful way to introduce new experiences to a product — whether it’s a design update, a feature tweak, or a back end algorithm change — while minimizing risk. Instead of rolling out changes to all users at once, the new experience is shown to a select group of users, known as the <strong>“Treatment” group</strong>, while the rest of the users, referred to as the <strong>“Control” group</strong>, continue to see the existing experience.</li><li>This approach allows us to measure the impact of the change by comparing the behavior of the treatment and control groups. Key metrics, such as user engagement or conversion rates, determine whether the new experience delivers the desired results.</li><li>Imagine an e-commerce website testing a new color (blue to red) for its “Purchase” button. A portion of users (treatment group) sees the button in red, while the control group sees it in the existing blue. By analyzing the difference in click-through rates between the two groups, the company can decide which color performs better.</li><li>A/B testing not only helps improve products iteratively but also ensures data-driven decision-making with minimal disruption to the user experience.</li></ul><h3><strong>Why Do We Need A/B Testing?</strong></h3><p>Without A/B testing, it’s challenging to accurately measure the impact of a change. While we might observe correlations between a change and its outcome, we cannot confidently establish causation. This can lead to incorrect conclusions and misguided decisions.</p><p><strong>The Problem with Correlation</strong></p><p>Let’s consider an example: you decide to change the color of a “Purchase” button on an e-commerce website from blue to red. If sales increase by 5% after the change, it’s tempting to attribute this boost directly to the new button color. However, the increase might actually be due to other factors — such as seasonality, where sales are naturally expected to rise by 10% during that period.</p><p><strong>Why This is Inconclusive</strong></p><p>In this scenario, the change in button color is correlated with higher sales, but without a controlled test, there’s no way to determine if it caused the improvement. Did the red button truly enhance user engagement, or was the sales increase simply due to external factors?</p><h3>How A/B Testing Solves This</h3><p>By splitting users into a Control group (who see the original blue button) and a Treatment group (who see the new red button), A/B testing isolates the impact of the change. Comparing results between the two groups allows you to confidently determine whether the new button color drove the sales increase or if it was influenced by other factors.</p><p>A/B testing eliminates guesswork, providing clear, data-driven insights to guide product decisions and improve user experiences.</p><h3><strong>How to Run an A/B Test</strong></h3><p>Running an effective A/B test requires careful planning and execution to ensure unbiased and reliable results. Here’s a step-by-step breakdown:</p><p><strong>1. Randomization</strong></p><p>In a well-designed A/B test, the Control and Treatment groups must be representative of each other, with no inherent biases. For example, dimensions like gender, demographics, or user behavior should have consistent distributions across both cohorts. This ensures any observed differences are due to the tested change, not external factors.</p><p>The most commonly used randomization unit is the user, where each individual is randomly assigned to either the Control or Treatment group.</p><p><strong>2. Hypothesis Formulation</strong></p><p>Every A/B test begins with a clear hypothesis, such as:</p><ul><li>“A [specific change] will increase/decrease my [primary metric].”</li></ul><p>For instance:</p><ul><li>“Changing the button color from blue to red will increase the click-through rate by 10%.”</li></ul><p>The specific change (e.g., button color) is the only difference between the Control and Treatment groups, ensuring a controlled environment for testing.</p><p><strong>3. Define Primary Metrics</strong></p><p>The purpose of the test is to measure the impact of the change on key performance indicators (KPIs). Most organizations focus on primary metrics aligned with topline goals, such as:</p><ol><li>Revenue: Does the change drive higher sales?</li><li>User Experience: Does it improve engagement or retention?</li></ol><p>Clearly defining these metrics helps focus the experiment and assess its success.</p><p><strong>4. Conduct a Power Analysis</strong></p><p>Before starting the experiment, run a power analysis to determine how long the test should run. Power analysis ensures the sample size is large enough to detect statistically significant differences between the Control and Treatment groups.</p><p>Factors to consider:</p><ol><li>Expected effect size (e.g., how much improvement you anticipate)</li><li>Desired confidence level (commonly 95%)</li><li>Variability in the metric being measured</li></ol><p>This step prevents prematurely ending tests or running them longer than necessary.</p><h3><strong>Common Challenges in Running A/B Tests</strong></h3><p>While A/B testing is a powerful tool for data-driven decision-making, it’s not without its challenges. Here are some common issues that can arise during an experiment:</p><p><strong>1. Sample Ratio Mismatch (SRM)</strong></p><p>This occurs when the proportion of users allocated to the Control and Treatment groups is not as expected. For example, if the planned split was 50:50, but the actual allocation skews to 40:60, your results may be biased and unreliable.</p><p>Possible Causes:</p><ul><li>Errors in randomization logic</li><li>Filtering rules that unintentionally exclude certain users</li></ul><p>Impact:</p><ul><li>SRM can distort results, making it difficult to draw accurate conclusions.</li></ul><p><strong>2. Novelty Effect</strong></p><p>The novelty effect occurs when users’ initial engagement with a new feature or design is inflated because it’s fresh and exciting. Over time, this effect often diminishes as users become accustomed to the change.</p><p>Impact:</p><ul><li>Short-term spikes in metrics may not reflect long-term trends.</li><li>Decisions based solely on initial results can be misleading.</li></ul><p><strong>Upcoming articles</strong></p><ul><li>Power analysis is a Key concept to determine the length of the experiment. I will make a detailed Post exclusively for it.</li><li>There are a bunch of cases where we cannot run the A/B Tests due to various reasons. I will write a detailed post on the commonly used methods in such observational studies.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3c4d2164d1c2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Job Search — A Big Maze]]></title>
            <link>https://medium.com/@praveenguptaz/job-search-a-big-maze-94df1da45df6?source=rss-412e5987a786------2</link>
            <guid isPermaLink="false">https://medium.com/p/94df1da45df6</guid>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[job-search]]></category>
            <category><![CDATA[resume]]></category>
            <dc:creator><![CDATA[Praveen Gupta Sanka]]></dc:creator>
            <pubDate>Fri, 03 Feb 2017 04:34:02 GMT</pubDate>
            <atom:updated>2017-02-03T04:35:40.748Z</atom:updated>
            <content:encoded><![CDATA[<p>This post is targeted towards the audience who are about to finish their Graduation and in pursuit of their future career opportunities. This post is a piece of advice that I would give to myself if I travel back in time to September 2016 (when I started my Job Search).</p><p>When I term the ‘Job Search’ as Maze, I mean it. We often think that we are in pursuit of our destination (dream Job) until we find a blockade (Reject) and then we get back to Square One (Restart the Application process).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*bjIgc3D6HgK5I-GD.jpg" /></figure><p>Job Search could be tiresome, hectic, frustrating and whatever words that have similar meaning. Brace yourself to face hundreds of rejects and myriad of ups and downs during your job hunt. Gear up to be slammed by robots — Yes! Automated rejections by ATS (Applicant Tracking System). Believe me, they are going to be terrible. Here, I describe my haunting period and the tools I used to survive during this dreadful period. Hope you find this post useful.</p><p>This post would answer a lot of questions such as when should I start my job search, how many applications do I need to submit a day, the best way to optimize the Resume and many more.</p><p>First things first, I would suggest to begin the job search at least three months before the OPT period begins (considering Business Analytics Grads). The average time to secure a job in job hunt is 90 days as per my experience and of course, there are outliers ;).</p><p><strong>Resume Optimization</strong></p><p>Resume is the only marketing tool that we have in-hand before we jump into conversations with any interviewer. It has to be refined and optimized to ensure a good number of calls. Here, I talk about the ways the Resume could be optimized.</p><p><strong><em>Master Resume</em></strong> — Prepare a Master Resume which would have more number of Academic Projects (around eight) that will help in customizing the Resume in easier fashion. Every Resume that is used during the application process would be a subset of the Master Resume. A new version of Resume has to be prepared every time you sense things are not working out.</p><p><strong><em>Resume Revisions</em></strong> — I recommend a new version to be prepared every week for first 4 to 6 weeks (modifying the content of projects, summary or changing the order of sections etc.). The Resume should not be lost in abysm after the revisions. It so happened to me that after 5 or 6 revisions, I have a lot of catchy terms on my Resume but the meaning of the sentences were lost. So, <strong><em>be cautious</em></strong> during every revision.</p><p><strong><em>Resume Content</em></strong> — The ultimate version of your Resume should be in such a way that it is easily understood by a teenager. Common business terms should be used (otherwise, ATS would reject) and all the bullets should be in short and crisp sentences. I would suggest a sub-heading for every bullet. It would help the interviewer in asking questions succinctly during the interviews. A similar format, if followed for the titles of Academic projects would be beneficial.</p><p>For example, I defined my projects as shown below. Sub-headings help a lot in conveying more about the project in short time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*OK_qIs51r_XyD5wc.jpg" /></figure><p><strong><em>Sections of Resume</em></strong> — Coming to the order of sections on the Resume, I observed that it is better to put ‘Experience’ section above the ‘Academic Projects’ section. More weightage is given to the recent Experience by the interviewers, so extra care should be taken in articulating it. It is better to put <strong>Summary</strong> section which describes what is so special about you. I believe Summary is the only area on the Resume where we could talk about soft skills.</p><p><strong><em>Two-Page Resume</em></strong> — It is dogmatic to think that only single page Resume would work. It is better to try two-page Resume as well. You never know, which version / kind of Resume gets you the job. I had a better conversion rate for my two-page Resume when compared to the single page Resume. LaTeX version of the Resume definitely stands out when compared to the others. A LaTeX version of Resume could be prepared using ‘www.overleaf.com’.</p><p><strong><em>Target List</em></strong> — Prepare a target list of companies. Maintaining a spreadsheet containing all the companies that have been applied and what else needs to be applied would definitely help.</p><p><strong><em>Big Suggestion</em></strong> — It is better not to apply to the dream companies during the first month (I made this mistake and regretted for it). Let the Resume got through several versions and then apply to dream companies when you are confident that the Resume would be picked.</p><p><strong><em>JobScan </em></strong>— To optimize the Resume during initial phases, JobScan tool (www.jobscan.co) could be used. It gives you a % match between the Resume and the Job Description based on which corresponding actions could be taken.</p><p>Below is a snapshot showing the way the results from JobScan are displayed. It also would display what softskills the Resume does not contain that are required for the Job.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*Q5N1yZ110yBECBt2.jpg" /></figure><p><strong>Getting in touch with Recruiters</strong></p><p>Once you believe that your final version of Resume is out there and vexed of receiving automated rejects, something better needs to be tried. There is a better conversion rate when we get in touch with recruiters than just following the conventional application process. Here are different approaches to get in touch with recruiters</p><p><strong><em>LinkedIn Premium</em></strong> — It is better to start getting in touch with the Recruiters on LinkedIn. LinkedIn Premium allows for 5 InMails per month. The conversion rate through InMails for me was not high though. However, it is better to give it a try.</p><p><strong><em>Rapportive</em></strong> — This is an amazing add-on available on both Firefox and Chrome browsers. It helps in identifying if an email id is linked up with any LinkedIn accounts, thereby, telling us the validity of the email address. I personally felt that emails worked out better than InMails.</p><p>It is perfect if we know the contact person for the job. But what if we do not know the corresponding recruiter? Do the Google Search as shown and it gives you the top LinkedIn profiles of that company</p><blockquote>· ‘sites: linkedin &lt;company_name&gt; recruiters’</blockquote><p>So, the next question is, how do you identify the email id based on a person’s name?</p><p>Most of the times the email id would be –</p><p>· FirstName@domain.com</p><p>· FirstName.LastName@domain.com</p><p>· FirstLetter(FirstName)LastName@domain.com</p><p>So, we just enter either of these formats in the ‘To’ address of the email and it shows up if any LinkedIn account associated with it.</p><p>Let’s say that a person did not link up his LinkedIn account with his official email id (which recruiters rarely do), we would not find anything positive with Rapportive. Here we can make use of a website — hunter.io.</p><p>Below image gives you a snapshot of how the rapportive results are displayed</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*-Gzaft4ZjMJHvPWD.jpg" /></figure><p><strong><em>Hunter </em></strong>— Using Hunter, one can check the availability of a domain name and the validity of email addresses. So, all the expected email addresses could be tested using this to get in touch with the recruiters.</p><p>Never hesitate to contact the recruiters on LinkedIn / email / phone. What is the worst thing that could happen after emailing a recruiter? — They may ignore you. What if they don’t ignore — Don’t miss an opportunity.</p><p>Below image gives an idea of how the results are displayed from Hunter.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*5podXXKzypnrq040.jpg" /></figure><p><strong><em>Expand your wings (Referral)</em></strong> — Try contacting your friends, friends of friends and LinkedIn connections to make referrals for you. What is the worst thing that could happen? — They would not respond or they would respond negatively.</p><p><em>There is nothing that you would lose if you contact but a great career opportunity if you don’t</em>.</p><p><strong>Dream Companies</strong></p><p>There is an additional point I would like to mention when it comes to dream companies. See if their ATS provides a facility to create <em>search agents</em> that would notify you when a similar job is posted. Don’t step back to apply for positions that were created long time ago. There were couple of instances I got calls even when I have applied 4 months after the job is created.</p><p><strong>Last Resort — Easy Application</strong></p><p>Filling job application could be tiresome using several ATS systems such as Taleo, Successfactors, Brassring etc. These are some of the worst ATS systems, I have ever seen. You’ll be drained while applying through any of those portals. However, there are a few ATSs such as GreenHouse, JobVite, Lever etc. through which application could be finished quicker. Here is a list of companies using such easy ATSs.</p><p>www.github.com/j-delaney/easy-application</p><p>There is a reason why I suggested this as a last resort. If you try this way to apply for jobs during the initial phases itself, you may not be able to apply for greater set of companies that use the complicated ATS systems.</p><p><strong>Apply! Apply! Apply!</strong></p><p>Never ever stop applying, even when you have multiple on-sites lined up (This doesn’t mean that you should not prepare for the interviews) or even when you are awaiting on multiple final round results. The number of rejects are directly proportional to the number of applications. In short, if you are not getting good number of rejects, it means you are not applying enough. Job Application process should be stopped only after you receive an offer letter.</p><p>It is better to upload the Resume on all the platforms such as LinkedIn, Indeed, CareerBuilder, Monster etc. so that you could get more number of calls.</p><p>Never undermine the power of easy applications through LinkedIn and Indeed. However low the conversion rate through ‘LinkedIn Apply’ might be, keep applying. You never know that the company to which you did not apply might have shortlisted you.</p><p><strong>Staffing Associates / Job hunters</strong></p><p>It is better to apply to multiple staffing associates. During the initial phases, I did not understand the difference between consulting companies and staffing associates and I was reluctant to apply to them. But, it is always beneficial to get in touch with Headhunters or staffing agents as it is always good to have another person looking for your career opportunities.</p><p><em>Staffing Associates</em> — Apply to their client on behalf of you and ultimately, you would be working for the client and enrolled in the client’s payroll. Once you start working, you wouldn’t even look back at the Staffing agencies.</p><p><em>Consulting</em> — Apply to their client on behalf of you and ultimately, you would be working for the consulting company, working for the client and enrolled in the consulting company’s payroll. Applying to these should be at the bottom of the list and better be approached at a later point of time.</p><p><strong>Statistics</strong></p><p>From my experience, I found the conversion rate using direct applications is around 3% i.e. only 3 applications out of 100 would be shortlisted. Conversion rate by contacting recruiters using Rapportive is around 10%. Conversion rate using LinkedIn Apply is &lt;1%.</p><p>Overall, here is my Job hunt funnel — I have applied to more than 700 companies, attended more than 40 Recruiter calls (includes calls through referrals, InApply etc.), answered around 25 second rounds, been to 4 on-sites and ended up with 3 offers.</p><p><strong>Conclusion</strong></p><p>Through this post I wanted to share my experiences which could help someone in this process. I would like to end my post thanking everyone who helped me during my job search personally or virtually.</p><p>Stay tuned to know what kind of interview questions you can expect during the interviews!</p><p>All the best for your future prospects.</p><p><em>Originally published on my personal blog — http://analytics-edge.blogspot.com/</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=94df1da45df6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>