<?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 Yoshihiro Sato on Medium]]></title>
        <description><![CDATA[Stories by Yoshihiro Sato on Medium]]></description>
        <link>https://medium.com/@yosatonet?source=rss-df612c2ee268------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*udXZjg8MnS1W_R_n</url>
            <title>Stories by Yoshihiro Sato on Medium</title>
            <link>https://medium.com/@yosatonet?source=rss-df612c2ee268------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 17:46:52 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@yosatonet/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[The Importance of Imagination as Information Literacy : Structural lessons from the failure of…]]></title>
            <link>https://medium.com/@yosatonet/the-importance-of-imagination-as-information-literacy-structural-lessons-from-the-failure-of-d875c51b087e?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/d875c51b087e</guid>
            <category><![CDATA[audit]]></category>
            <category><![CDATA[japan]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[modeling]]></category>
            <category><![CDATA[event-management]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Sat, 14 Nov 2026 02:12:23 GMT</pubDate>
            <atom:updated>2026-11-14T02:12:23.204Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Importance of Imagination as Information Literacy : <em>Structural lessons from the failure of Local Food Fight Tottori</em></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/531/1*ppoTpnZEqxiY-r2glcpXYg.png" /></figure><p><strong>Note:</strong> Everything I know about this incident comes from secondary sources — news reports and publicly available social media posts. Even posts attributed to the organizer cannot be verified as reflecting their true intent. I will work from confirmed facts, treating unverified claims as “reported to be” rather than established.</p><p>I was aware of this incident when it happened. At first, I chose not to write about it. The online reaction felt hysteric — less like concern for the people harmed and more like an audience enjoying the spectacle of outrage. I could understand victims speaking out. I could accept criticism grounded in structural argument, open to counter-evidence. But what I saw looked mostly like people having fun with someone else’s fire.</p><p>Then, browsing note (a Japanese publishing platform), I came across an article that stopped me. It broke down the structural problems clearly and carefully. That was the nudge. I had things I wanted to say about this case — things I had been holding back precisely because I didn’t want to be mistaken for one of the spectators. What I wanted to do was frame this as a risk problem: what structure allowed this to happen, and what structure might have prevented it.</p><h3>The facts</h3><p>On November 7–8, 2026, an event called “LOCAL FOOD FIGHT in Tottori” was held at a public lawn space in Kurayoshi City, Tottori Prefecture, Japan. The concept: restaurants from across Japan competing by sales volume, with celebrities in attendance. The program included a dedicated slot titled “VIP Vendors, Corporate Sponsors &amp; Government PR Announcement Time.”</p><p>Sources include FNN (Fuji News Network, a major Japanese broadcaster), local broadcaster BSS, and a Toyo Keizai Online article (one of Japan’s largest business news outlets) by media researcher Jo Kido, referenced by Yahoo News Japan.</p><ul><li>Projected attendance: 50,000 visitors</li><li>Projected economic impact: 1 billion yen (approx. $6.7M USD)</li><li>Booth fee: 50,000 yen per vendor (approx. $330)</li><li>Number of vendors: approximately 50</li><li>Actual attendance: a few hundred people</li><li>Vendors reported financial losses</li><li>Equipment shortfalls on-site: tents, power supply, washing facilities</li><li>Post-event communication from organizers: absent, insufficient, or unclear</li></ul><p>This is the extent of what I can confirm. Everything beyond this is analysis.</p><h3>Structural observations</h3><p><strong>1. The projections show no signs of hypothesis testing.</strong> According to AI-assisted research, an economic impact of 1 billion yen is comparable to a dome-scale professional sports event. A typical local festival — where most attendees are local and spending is limited to food and transport — generates roughly 150 to 500 million yen. The gap is not small. Whether any stress-testing of the 50,000 figure or the 1 billion yen projection was done, and if so what it found, is not publicly known. If the numbers were produced without systematic verification, that is itself a risk factor.</p><p><strong>2. Responsibility was not clearly assigned.</strong> There appear to have been at least two people in leadership roles, with a range of involvement from nominal figurehead to operational lead. When accountability exists on a spectrum between two people without explicit assignment, post-failure explanations tend toward deflection rather than acknowledgment. The structure creates that outcome regardless of intent.</p><p><strong>3. A listed company was involved beyond simple sponsorship.</strong> Reports indicate that a major publicly listed company participated in a capacity that would have triggered internal approval processes — budget review, sign-off. In principle, such processes could have included scrutiny of the event’s projections. Whether that scrutiny occurred is not publicly known. What is notable is that the presence of an institution with internal governance did not appear to function as a check on the event’s planning assumptions.</p><p><strong>4. Local government had access but did not check.</strong> The event program included a dedicated government PR slot. The mayor of Kurayoshi received a formal courtesy visit from the organizer before the event. No direct financial involvement by the city has been confirmed, but staff time and the use of the mayor’s name represent public resources. This means local residents have a legitimate channel — a formal citizen inquiry mechanism to the city government — to ask what, if anything, the city verified before its name appeared alongside this event.</p><p>Taken together: the projections appear untested; accountability was structurally diffuse; institutions with the capacity to check were present and did not check; and vendors, as the information recipients, bore the cost of that failure.</p><h3>What might have prevented this</h3><p>The desire to run an event can come from anywhere on a spectrum — from wanting the experience of organizing something, to building a personal track record, to genuine community contribution. Where this event sits on that spectrum is something only the organizers know. What matters structurally is that the initial impulse produced a plan. And plans, however detailed they appear, can be reverse-engineered to look more rigorous than they are. That can happen without malicious intent.</p><p>An optimistic projection — 50,000 visitors, 1 billion yen — might reflect a failure of imagination rather than deliberate deception. At the planning stage, the risk imagination was simply not working. If the plan proceeds from that stage to execution without an external check, it cannot be stopped from the inside.</p><p>The practical answer is third-party review: an independent check of the planning assumptions, with some form of certification that the check occurred. This does not require a large audit firm. It requires that someone outside the organizing team ask: <em>what is the basis for this number?</em> And that the answer be documented. Had that documentation existed — even a simple note stating that an external reviewer had examined the projections — vendors would have had different information when deciding whether to participate.</p><p>There is a psychological dimension worth naming as well. When organizers operate from an implicit position of “I am giving you an opportunity” — and when vendors internalize “I am fortunate to participate” or “I paid, so I have standing” — an invisible hierarchy forms. That hierarchy suppresses the questions that should be asked. In any business relationship, the baseline should be partnership: equal standing, with each party responsible for their own due diligence. When that principle erodes, trouble follows.</p><p>All of this comes back to imagination — the capacity to place yourself in another person’s situation, to anticipate consequences, to ask what happens if you are wrong. That capacity cannot be measured from outside. Which is why external checks are not a supplement to imagination. In its absence, they are the only substitute.</p><h3>Closing</h3><p>The victims in this case are the vendors. Measured against the organizer’s own projected economic impact of 1 billion yen, the shortfall in expected value is, at maximum, that entire figure. The actual losses are smaller — but no less real to the businesses that bore them.</p><p>What I find difficult is the knowledge that this was preventable. If hypothesis testing, modeling, and audit were normal parts of how events like this are planned — if skipping them were seen as a failure of professional practice rather than a reasonable shortcut — the structure that produced this outcome would not have held.</p><p>Those concepts exist. The tools exist. What is missing is not capability but expectation: the shared understanding that projections used to solicit participation from others carry a responsibility to be verified. That expectation does not yet exist in this context. Its absence is not a personal failing of any individual involved. It is a gap in the ecosystem — and gaps like this one persist not because no one cares, but because the infrastructure of expectation that would make verification normal has not yet been built. Building it is the work.</p><p>#EventManagement #RiskManagement #Audit #Modeling #InformationLiteracy #Japan</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d875c51b087e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Japan Wants to Ban Social Media for Minors. The Logic Doesn’t Hold.]]></title>
            <link>https://medium.com/@yosatonet/japan-wants-to-ban-social-media-for-minors-the-logic-doesnt-hold-c99526450029?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/c99526450029</guid>
            <category><![CDATA[sns]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[literacy]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Wed, 04 Nov 2026 23:34:08 GMT</pubDate>
            <atom:updated>2026-11-04T23:34:08.642Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/671/1*CS6rd44fZlr510WVT8oUKw.png" /><figcaption>Window Paint AI Generated</figcaption></figure><p>In late October 2026, a Japanese civic group called “The Association to Protect Children from SNS Harm” (子どもをSNS被害から守る会) submitted a petition to Japan’s government demanding that social media platforms be legally required to verify the age of users and ban those under 16. The petition cites Australia’s social media law as a pioneer model and calls for criminal penalties against non-compliant platforms.</p><p><a href="https://www.sankei.com/article/20261030-PAFRORT5TZMNZMNYHGO32PWA2Y/">https://www.sankei.com/article/20261030-PAFRORT5TZMNZMNYHGO32PWA2Y/</a></p><p>The petition gathered significant media coverage. The press conference photo tells a story of its own: a room full of older adults, no young people present, speaking on behalf of the demographic they seek to protect.</p><p>I’m an IT literacy educator based in Sendai, Japan. I’ve spent three decades working with spreadsheets, data modeling, and information risk. I think the petition’s heart is in the right place. Its logic is not.</p><h3>The Premise I Accept</h3><p>Social media is dangerous. I want to be clear about this, because the argument I’m making is not a dismissal of harm.</p><p>SNS platforms are engineered to induce dependency. Their recommendation algorithms are optimized for engagement, not wellbeing. The timeline of any active user — mine included — is a mixture of genuine information and content designed to bypass rational evaluation: pseudo-science, emotional manipulation, manufactured consensus. This is not a bug. It is the product.</p><p>The danger is not hypothetical and it is not exclusive to children. Adults are equally susceptible. The variable that separates safer users from more vulnerable ones is not age. It is whether they can <em>see the contamination</em> — whether they carry a framework for evaluating information before accepting it.</p><h3>Four Problems With the Petition</h3><p><strong>1. The people being protected have no voice.</strong></p><p>The petition’s three demands — age verification, platform penalties, Australia-style regulation — were designed entirely by adults, for children, without children. The press conference featured no young people. No testimony from adolescents about what they gain from social media, what they lose, or what kind of support they would actually want. Protecting someone while excluding them from the conversation about their own protection is not advocacy. It is control under a different name.</p><p><strong>2. There are no success criteria.</strong></p><p>I read the original petition text. It contains three demands and zero evaluation design. No definition of what “protection” looks like in measurable terms. No monitoring mechanism. No threshold at which the regulation would be judged to have succeeded or failed.</p><p>This is a structural problem, not a minor omission. A policy without falsifiability cannot be held accountable. If harm continues after prohibition, the response becomes “enforcement wasn’t strong enough.” If harm decreases, the prohibition gets the credit regardless of whether it caused the decrease. The petition has designed itself to be immune to evidence.</p><p><strong>3. Australia is not a success story — yet.</strong></p><p>The petition cites Australia as a model. Australia enacted its under-16 social media ban on December 10, 2025. The eSafety Commissioner’s own Phase 1 assessment found that account ownership among under-16s dropped from 52% to 42% — a 10-point reduction, with 42% still active. Parental awareness of the ban actually <em>decreased</em> post-implementation. eSafety itself characterized Phase 1 as insufficient and moved to Phase 2 with doubled fines and strengthened enforcement.</p><p>Phase 2 results are not yet published as of November 2026.</p><p>Australia has not concluded. Citing it as a proven model is not an error of emphasis — it is a misrepresentation of the evidence.</p><p><strong>4. The petition closes off falsification.</strong></p><p>A claim that cannot be disproven is not a policy proposal. It is a declaration of moral position. That may have its place in public discourse. It does not belong in legislation design.</p><h3>The Structural Confusion</h3><p>The petition conflates two distinct problems.</p><p>The first is <strong>platform design</strong>: social media companies engineer dependency and bear responsibility for the architecture of harm. This is a legitimate regulatory target.</p><p>The second is <strong>user cognition</strong>: what frameworks do users — children and adults — bring to their media consumption? This is an educational target.</p><p>Prohibition addresses neither. It removes access without changing the platform architecture (users migrate to unregulated alternatives), and it removes the opportunity to develop evaluative skills before adulthood.</p><h3>The Argument That Goes Unasked</h3><p>There is a longer version of this problem.</p><p>Every time humanity externalizes cognition into a tool, risk increases. The printing press enabled mass propaganda. Google enabled filter bubbles at scale. AI has made it possible to generate unlimited plausible-sounding falsehoods at zero marginal cost. This is not a new danger — it is the permanent condition of a species that builds cognitive prosthetics.</p><p>The internet did not create cognitive risk. The calculator did not create cognitive risk. They <em>amplified</em> pre-existing risk and changed its distribution. The correct response has never been to remove the tool. It has always been to develop the capacity to use it without being used by it.</p><p>Social media, for many children in Japan, is one of the few windows to perspectives outside their immediate community — including perspectives that might help them recognize when the adults in their lives are wrong. That is worth something. It may be worth protecting.</p><h3>What I Would Ask the Association</h3><p>If you are genuinely concerned about children’s safety in digital environments — and I believe you are — the most durable investment is not prohibition. It is building the skill to evaluate information before accepting it.</p><p>This skill is teachable. It is not age-dependent. It transfers across platforms, across technologies, across whatever comes after social media.</p><p>Design an educational program. Define what “a child who can navigate online information safely” looks like in observable terms. Test it. Publish the results. Build something that works when the next platform arrives, because it will.</p><p>An 18-year-old who has never been taught to evaluate information and encounters social media for the first time as a legal adult is not protected. They are simply delayed.</p><p>The goal should be children who can see the contamination — not children who have never seen it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c99526450029" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Art of Brewing Coffee with Excel]]></title>
            <link>https://medium.com/@yosatonet/the-art-of-brewing-coffee-with-excel-77f0f063bfdf?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/77f0f063bfdf</guid>
            <category><![CDATA[m5stack]]></category>
            <category><![CDATA[iot]]></category>
            <category><![CDATA[coffee]]></category>
            <category><![CDATA[coffee-culture]]></category>
            <category><![CDATA[excel]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Sat, 12 Sep 2026 13:37:20 GMT</pubDate>
            <atom:updated>2026-09-12T13:37:20.473Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/740/0*RIE_m8DtEWhPaTB-.png" /></figure><p>and-drip coffee is my obsession. Equipment keeps piling up. Before this, my only real hobby was Excel — then suddenly I fell headfirst into coffee. A full-on coffee drip swamp.</p><p>My approach to dripping is, I’ll admit, a little unusual. The guiding principle: never stress the beans.</p><h3>My Drip Process</h3><p>Every drip starts with beans — sourced from a roaster or café. Each batch is completely different: the origin, the variety, the roaster’s philosophy. Before the coffee reaches your cup, there’s a whole ritual to enjoy.</p><ol><li>Grind the beans with a coffee mill into a fine powder.</li><li>Bring water to a full boil, transfer it to a drip kettle, and cool it to the temperature that suits the roast level. Monitor with a thermometer.</li><li>Set a dripper on the server, and place a paper filter inside.</li><li>Pour a small amount of hot water first — just enough to bloom.</li><li>The bloom: CO₂ escapes from the grounds, and the bed swells up like a small hill.</li><li>Slowly pour in stages, taking care not to collapse the mound.</li><li>Finish with a bold final pour — water pressure drives the last of the extraction.</li></ol><p>That’s the flow.</p><h3>I Made the Mistake of Wanting to Record It</h3><p>There are recipes out there: pour X ml at Y seconds. They work. They taste good. But they don’t taste like <em>me</em>. And honestly, my preferred taste changes depending on my mood that day.</p><p>So a perfect cup isn’t something I engineer — it’s something the beans offer me on a given day, and I happen to catch it. Trying to reproduce it the next time? It’s never quite the same. And I never know when that perfect moment will strike.</p><p>The answer seemed obvious: do what a dashcam does. Record everything, always.</p><p>The simplest version: set a scale under the dripper, hit record on your phone. But then you’re left with hours of footage to transcribe, hard drives filling up, and eventually — you stop doing it.</p><p>What I actually needed was to capture the weight as raw numbers, embedded seamlessly into the act of brewing.</p><h3>No Such Product Exists</h3><p>Commercial scales show you the weight at this moment. That’s it. There’s nothing on the market that records weight over time. So close — and yet.</p><p>Fortunately, I had a background that made this problem feel solvable. I once worked as a measurement engineer in semiconductor manufacturing. I knew in my bones that any physical phenomenon can be recorded automatically. And it was there that I first met Excel — using it to organize the data I captured from instruments.</p><p>That job was contract-based (SES), so it ended when the contract did. I moved on to become a PC instructor, and that’s where I’ve been ever since.</p><p>Which means: taking a phenomenon, recording it, and wrestling it into Excel — that’s not a creative leap for me. Given my background, it was almost inevitable.</p><h3>The World Has Changed</h3><p>That said, I’d been away from engineering for a while. When I started researching what it would take to build this thing, something alarming happened.</p><p>I had mentally budgeted ¥100,000 (roughly $700) in parts. I expected soldering. I had already retrieved my soldering iron from storage.</p><p>I didn’t need it.</p><p>Individual components cost a few hundred yen each. Connections are made by pressing wires into connectors — no soldering required. Entire functional modules are sold pre-assembled, in blocks.</p><p>The load cell circuit, the ADC conversion, the serial transmission to a PC — almost none of that needed to be built from scratch. I needed a controller, a weight sensor kit (which came with its own housing), and that was basically it.</p><ul><li>Controller: M5Stack TAB5 (a fairly high-end model)</li><li>Sensor: Mini Weight Unit (HX711-based)</li><li>Load cell: a kit from Amazon — the kind used in school science labs</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/0*u3DR6Ln3BXJvu5NO.png" /><figcaption>The device at the bottom right is an M5Stack (specifically the TAB5, a fairly high-end model).</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/555/0*GHtQF3jvlfHlTctc.png" /><figcaption>Weight sensor</figcaption></figure><p>The only thing resembling “programming” was setting up the M5Stack to send weight data to the PC. And even that required no code — I used UIFlow2, a visual block-based environment. Drag, connect, done.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/740/0*LeySHST02U7daQja.png" /><figcaption>The program uses a block-based format called UiFlow2.</figcaption></figure><p>And on the Excel side? There’s a built-in add-in called <strong>Excel Data Streamer</strong>. Enable it, and Excel ingests the serial data automatically — and graphs it in real time.</p><p>I don’t know of a single other Excel user who has dug into Data Streamer. Not one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/740/0*P9XOB9VB5zEpbr8E.png" /><figcaption>Actual Excel screen</figcaption></figure><h3>The Machine Is Complete</h3><p>With this setup, I can record every successful brew alongside the bean variety — effectively building my own recipe database. What seconds passed, how many ml I poured. Captured.</p><p>But something unexpected happened. The goal was recording. The result was that I couldn’t stop watching the graph being drawn in real time.</p><p>I started paying more attention to the screen than to the coffee.</p><p>“A little unusual” turned out to be an understatement.</p><h3>The Swamp Continues</h3><p>As of September 2026, I have built five versions of this device. Still not satisfied. Version six is currently under construction.</p><p>The latest captures: temperature inside the kettle, water volume in the dripper, and water volume in the server — letting me see exactly how boldly or carefully I poured, and how much the kettle temperature dropped from the first pour to the last. The data feeds into a Power Query database, and I’ve been continuously loading it into a generative AI project. Someday, maybe, an optimal recipe will emerge.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/517/0*Z8UhoVpqcgWN3qnA.png" /><figcaption>Latest model (truly wireless)</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/751/0*ToFUi8To4Ct9_LI3.png" /><figcaption>Using Google Sheets to create graphs from CSV files found on the web.</figcaption></figure><p><em>Measuring coffee is, in a way, what I do professionally too — just with higher stakes.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=77f0f063bfdf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Trusting Excel and Copilot Too Much Breaks the Integrity of Your Formulas]]></title>
            <link>https://medium.com/@yosatonet/trusting-excel-and-copilot-too-much-breaks-the-integrity-of-your-formulas-c89dd4285749?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/c89dd4285749</guid>
            <category><![CDATA[co-pilot]]></category>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <category><![CDATA[error-message]]></category>
            <category><![CDATA[risk-management]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Wed, 15 Jul 2026 05:14:54 GMT</pubDate>
            <atom:updated>2026-07-15T05:14:54.337Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*_HHKlvBbCF35rNh9MMWrcQ.png" /></figure><p>There’s a lot of talk right now about how convenient Excel, generative AI tools like Copilot, and the Copilot integration built into Excel itself have become. That convenience is real.</p><p>But simply accepting that convenience at face value can silently destroy your formulas. This is a report on exactly that happening.</p><p>What follows is a straightforward walkthrough: I simply followed Excel’s own prompts, one step at a time, and recorded what happened.</p><h3>The starting sheet</h3><p>I have a worksheet like this. It contains confirmed 2024 sales figures and confirmed 2025 sales figures, with monthly, per-product averages calculated for each year.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/886/1*gJyoEn7fat2sWZ7bt99zPQ.png" /></figure><p>On top of that, as a 2026 sales forecast, I calculate a target annual sales figure for all products in each month, based on the growth rate derived from the two prior years.</p><p>The formula for the target annual sales figure is:</p><pre>= (2025 monthly average across all products / 2024 monthly average across all products) * 2025 monthly average across all products * 4</pre><p>This multiplies the prior year’s figure by the growth rate, and then by 4 — the number of product types. Setting aside whether that’s the “right” formula, the intent behind it is at least legible.</p><p>What I want to draw attention to here is that the cells holding the 2024 and 2025 monthly averages are all marked with a small green triangle — Excel’s error indicator.</p><p>Selecting one of these cells reveals the warning: <strong>“Formula Omits Adjacent Cells.”</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/525/1*d2_ZqgZkCPZZQzfSh3s1qA.png" /></figure><h3>The formula isn’t wrong</h3><p>Looking at the formula in this cell, it’s simply averaging four product columns. There’s nothing wrong with it. And yet Excel flags it as an error.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/721/1*xb_kk16BDB3vP9Bqffjheg.png" /></figure><p>Excel treats it as an error whenever a formula references only part of an adjacent range of cells. In this case, that means: because column A isn’t included, Excel considers it an error.</p><p>Looking at column A, what displays as a three-letter month abbreviation is, underneath, actually a date.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/475/1*YrcKfbLJI850E_D8xzh4hA.png" /></figure><p>A date, stripped down to what it really is, is just a number — a date serial value — and internally, Excel treats it as such. So what Excel is really asking is: “Should this date serial value, which looks like a date, also be included in the average range?” But instead of presenting that as a question, or even a warning, it presents it as an <strong>“error.”</strong></p><h3>Step one: trusting Excel itself</h3><p>If it’s labeled an “error,” it feels like something that needs fixing. So I click <strong>“Fix this error.”</strong> This is the first step — the act of trusting Excel itself.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/434/1*qGpvU0DEbURYtF72YiX-RQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/429/1*TY9uP6V3DEhcB9X8RvyErA.png" /></figure><p>Doing so launches Copilot inside Excel.</p><p>At the end, it asks me how I’d like to proceed. While this is pending, the worksheet cells turn pink — as if to say, this hasn’t been finalized yet.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/628/1*-ikB5QfdVolgmq7fBOrksw.png" /></figure><p>Here, to let it auto-correct, I click <strong>Done</strong> on the Copilot panel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/413/1*Ay5fzTjDpD0OPwXU49enpg.png" /></figure><p>The error disappears.</p><p>The formula itself hasn’t changed either, so it looks as though nothing was actually altered — in other words, I assume Copilot checked things over, found nothing wrong, and simply cleared the error indicator. I feel reassured.</p><h3>What actually happened</h3><p>But look closer. The month display is now left-aligned in the cell.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/621/1*ITc0M29lrx8VvGdkroL3Lw.png" /></figure><p>Why did that happen? What was originally the date 2024/1/1 has been rewritten as the text string <strong>“Jan.”</strong></p><p>This is not something I asked for. The month display had, at minimum, year information embedded in it — and that data was silently erased by Copilot, without any explanation. The cell can no longer be used for date-based calculation or recognition.</p><p>This is not what I wanted to happen.</p><h3>Trying the built-in error checker instead</h3><p>Excel also has a built-in error-checking and correction feature that doesn’t require Copilot at all. It’s accessible from the Formulas tab on the ribbon.</p><p>Thinking this might fix things properly, I click <strong>“Update Formula to Include Cells.”</strong></p><p>This time, Excel rewrites the formula to include column A in the calculation range.</p><h3>Two fixes, two different — and both wrong — results</h3><p>In other words, there are two ways to “fix” this error, and they produce two different results. And neither of them is what I wanted. Both are unrequested, unilateral reinterpretations of my data and my formulas.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/654/1*wXIm79Kvxyl9zjVskjcYFg.png" /></figure><p>If this worksheet is left as-is, then no matter how sound the calculation logic for the 2026 forecast might be, the underlying source data has already been silently destroyed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/795/1*D5pEOiUhwxNs3aXUE_43JA.png" /></figure><p><em>This behavior was verified on July 15, 2026, at 14:00 JST using Microsoft® Excel® for Microsoft 365 MSO (Version 2607 Build 16.0.20228.20014) 64-bit.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c89dd4285749" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What Happens When You Turn “Temperature Sells Ice Cream” Into an Excel Formula]]></title>
            <link>https://medium.com/@yosatonet/what-happens-when-you-turn-temperature-sells-ice-cream-into-an-excel-formula-3556cf7c7bb0?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/3556cf7c7bb0</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[predicted]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[formula]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Mon, 13 Jul 2026 03:42:12 GMT</pubDate>
            <atom:updated>2026-07-13T03:42:12.508Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/898/1*ZiSg-AEEwybV1FE-OmnxNA.png" /></figure><p>Everyone knows this: the hotter it is, the more ice cream sells. You don’t need a maths lesson to believe that.</p><p>Let’s build it in Excel anyway.</p><p><strong>Day A</strong>: 20°C, 50 units sold. <strong>Day B</strong>: 30°C, 80 units sold.</p><p>That’s it. That’s our data. Two facts, already settled — they won’t change no matter what we do next.</p><p>Now suppose tomorrow’s forecast says 25°C. We can turn our two data points into a formula:</p><p><strong>Predicted units = (unit difference ÷ temperature difference) × (forecast temp − Day A temp) + Day A units</strong></p><p>Plug in 25°C, and the formula says 65 units. Done. Set it up once in Excel, and every morning you just swap in tomorrow’s forecast and get a number.</p><p>It looks complete. It looks like a fact. It isn’t.</p><p>Two points on a graph always produce a perfectly fitting line — always. That’s not a discovery, it’s a mathematical guarantee, and it should make you suspicious rather than confident. A formula built from two points can’t be wrong about those two points. It says nothing about day three.</p><p>So: what happens with a third day? A fourth? A tenth?</p><p>Once you have three or more points, they stop lining up neatly. You can no longer draw one perfect equation through them — only an approximation, a “roughly this.”</p><p>There’s a method for that. You already know its name.</p><p>You spent a good chunk of secondary school plotting exactly this kind of data — two variables, a scatter of points, a line drawn through the middle that doesn’t touch most of them. You were taught to read the scatter for strength and direction, and — this part matters — to be suspicious of any estimate that falls outside the range of points you actually plotted.</p><p>Here’s the strange part: almost nobody brings that instinct into Excel. The spreadsheet skips straight from raw numbers to a formula. No scatter, no line, no visual check on whether the two points you happened to use were representative or a fluke. The habit you built for years, on paper, in a maths classroom, quietly doesn’t make the trip.</p><p>That gap is worth naming, because it’s exactly where things go wrong. A cell containing “65” looks identical whether it came from a well-supported trend across twenty data points, or from two data points that happened to be sitting near the edges of a much messier reality.</p><h3>Two different things are hiding inside that one formula</h3><p>Pull the formula apart and you find three separate ingredients:</p><ol><li><strong>What actually happened</strong> — 20°C→50 units, 30°C→80 units. Settled. Not up for debate.</li><li><strong>What you’re assuming will happen</strong> — tomorrow will be 25°C. That’s a forecast, and forecasts are wrong sometimes.</li><li><strong>What you’re assuming about the relationship itself</strong> — that sales rise in a straight line with temperature. Maybe they do. Maybe they plateau once it’s hot enough that people would rather stay indoors.</li></ol><p>Excel doesn’t distinguish between these. A fact, a forecast, and an assumption about the shape of a relationship all end up looking like the same thing: a number in a cell, feeding a formula.</p><h3>Why this matters for spreadsheet auditing</h3><p>Most discussion of “spreadsheet auditing” — and I’ve looked, this holds in both English and Japanese material — centres on one question: is the formula built correctly? Are the references right? Was the macro tampered with? Is there a version history?</p><p>That’s real, necessary work. But it only checks one of the three ingredients above — the mechanics. It says nothing about whether the <em>forecast</em> was reasonable, or whether the <em>assumed relationship</em> was the right one to assume in the first place. A perfectly correct formula, built on a bad assumption, passes every mechanical audit and still gives you the wrong answer.</p><p>Cell Styles — a feature that’s been sitting inside Excel for years — actually gestures at this distinction. There’s a built-in “Input” style and a “Calculation” style, meant to separate the numbers you typed in from the numbers Excel derived. In years of teaching, I’ve rarely seen anyone explain what that distinction is <em>for</em>. It gets taught as a formatting choice.</p><p>I think it’s closer to the beginning of an answer than a footnote.</p><p>If Excel education started here — fact, assumption, and the mechanical layer that turns them into a number, each named and kept visibly separate — I suspect the motivation to learn Excel properly would look very different. Not “learn these functions,” but “learn to see which parts of your spreadsheet you’re allowed to trust unconditionally, and which parts you’re not.”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3556cf7c7bb0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Where Are Japan’s Spreadsheet Horror Stories?]]></title>
            <link>https://medium.com/@yosatonet/where-are-japans-spreadsheet-horror-stories-2a2fcbba94fe?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/2a2fcbba94fe</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[horror-stories]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <category><![CDATA[japan]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Sat, 11 Jul 2026 12:34:06 GMT</pubDate>
            <atom:updated>2026-07-11T12:34:06.466Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*rancyJdO8YS_YgdkUl150A.png" /></figure><p><em>A search that started with almost nothing, and ended with 65 cases, a hidden pattern, and a question I still can’t answer.</em></p><p>The <a href="https://eusprig.org/research-info/horror-stories/">European Spreadsheet Risks Interest Group</a> (EuSpRIG) has been collecting spreadsheet horror stories for over two decades. Misplaced decimal points that cost millions. A hedge fund that lost track of a formula and shorted the wrong thing. Budgets built on broken links to cells that no longer existed. The list is long, and it is almost entirely Western.</p><p>I wanted to know: where are Japan’s stories?</p><p>I already had one confirmed case of my own — a recent incident where a broken VLOOKUP inside Japan’s Ministry of Land, Infrastructure, Transport and Tourism corrupted 510 public procurement records. I wrote about it <a href="https://medium.com/@yosatonet/vlookup-is-not-magic-how-japans-ministry-of-land-corrupted-510-public-procurement-records-and-7730b713935d?sharedUserId=yosatonet">here</a>. Beyond that one, a handful of personal-data leaks. That was it. That was everything I could find.</p><p>And that bothered me, because it didn’t match my memory. I could recall hearing about “an Excel mistake” on the news more than once over the years — vaguely, in passing, the way you half-remember a headline. If those memories were real, the stories had to be somewhere. Google’s AI Overview turned up only a few. A normal web search turned up almost nothing at all.</p><p>For a while I genuinely wondered if the explanation was simpler than I wanted it to be: maybe Japanese workplaces just don’t use Excel the way Western ones do. Maybe there was nothing to find because there was nothing happening.</p><p>I gave myself one hour, as a last attempt, to try different search terms before giving up.</p><h3>The query that worked</h3><p>Nothing worked until I stopped searching for the tool and started searching for the <em>word people actually use to describe the mistake</em>. In Japanese administrative and news language, “スプレッドシート” (supureddoshiito, the loanword for “spreadsheet”) is not how these things get described. “表計算” (hyōkeisan, literally “tabular calculation,” the standard term for spreadsheet software) is.</p><p>The query that finally worked was simple: <strong>“ミス 表計算 [prefecture name]”</strong> — “mistake, spreadsheet, [prefecture]” — repeated across all 47 of Japan’s prefectures.</p><p>Some prefectures returned nothing. Some returned three or four relevant hits from news outlets, government disclosure documents, and security-news sites. But it worked. The mistakes were there. Excel <em>was</em> sitting quietly inside Japanese government and institutional workflows, and it <em>was</em> occasionally failing in ways serious enough to be reported. I had just been asking the search engine the wrong question.</p><p>By the end, I had 65 cases. I compiled them into a corpus, later anonymized it, and deposited it on Zenodo along with a short preprint discussing what I found. Neither claims to be exhaustive — this is a convenience sample built from one search engine, one language, and roughly ten rounds of query refinement. I want to be upfront about that limitation before saying anything else.</p><h3>The dog that didn’t bark</h3><p>Here’s what surprised me most: the big-money disasters that fill EuSpRIG’s collection were largely <em>absent</em>. I found a few procurement miscalculations, a few refund errors, a few budget mix-ups. But nothing on the scale of the stories Western spreadsheet-risk researchers are used to cataloguing.</p><p>What I found instead, over and over, was something else entirely.</p><h3>The most common failure: the sheet nobody deleted</h3><p>The single most frequent failure mode in the corpus — 14 of the 65 cases — was almost embarrassingly simple: personal information sitting on a second worksheet inside a file, hidden or just forgotten, that got published or emailed along with the sheet everyone actually meant to send.</p><p>There seems to be a workflow habit behind this: build a working file that contains both the personal-data sheet and the public-facing sheet in the same workbook, then strip the personal-data sheet out before publishing. It’s a reasonable-sounding process. But it depends entirely on someone remembering to do the stripping every single time, and workbooks are very good at quietly preserving whatever you forgot to remove.</p><p>There’s a second risk hiding inside that same habit that I don’t think gets talked about enough. If the standard operating procedure is “delete the personal-information sheet, then save,” there’s a real chance that the <em>deletion itself</em> — followed by an overwrite — permanently destroys the only copy of that data, with no leak at all, just a quiet, undocumented loss. Leakage and loss turn out to be two faces of the same operational risk: a single file being asked to serve as both the private record and the public document, with a manual step as the only thing separating the two.</p><p>That’s a fixable problem. Separate the file that contains personal data from the file that gets published, structurally, so the mistake can’t happen — not “remember to delete the sheet,” but “the sheet was never in the same file to begin with.” I don’t think this one needs a philosophical debate. It needs a workflow fix.</p><h3>The one that actually stopped me</h3><p>What did stop me — what I keep coming back to — is a smaller cluster of cases involving entrance exams and elections.</p><p>A miscalculated total score turned a passing student into a failing one, discovered and reversed only after the fact. A broken formula merged one student’s data into another’s official transcript — copies of which had already reached employers before anyone noticed. A vote-counting spreadsheet reused a stale candidate list, and an election result had to be corrected after the fact because the wrong person had, briefly, been declared the winner.</p><p>These aren’t rare edge cases dressed up to sound dramatic. They happened. They were reported. And they share something the leak cases don’t: the damage isn’t information moving to the wrong place. It’s a decision — about who passed, who’s employed, who holds office — made incorrectly and then acted upon by people downstream who had no way of knowing it was wrong.</p><p>Money, I’ve come to think, is fundamentally recoverable. An overpayment can be clawed back. An underpayment can be topped up. It’s a hassle, sometimes an expensive one, but it resolves. Dignity doesn’t work that way. You can’t refund the days a student spent believing they had failed. You can’t undo the interval during which someone held office who wasn’t supposed to. The record can be corrected. The experience of it can’t be.</p><h3>What I think this means</h3><p>My instinct, when I first sat with this, was blunt: <strong>don’t use Excel for anything irreversible.</strong> The risk is too structurally mismatched to the stakes.</p><p>But that instinct doesn’t survive contact with reality either. Budgets are tight. Timelines are tighter. Every one of these organizations had a reason for reaching for a spreadsheet — it was fast, it was familiar, it was what was available. Telling a prefectural high school to build a custom grading system is not a serious answer.</p><p>So here’s where I’ve landed, at least for now: the reason doesn’t make the risk acceptable. A tight budget is a real constraint. It is not a justification for exposing dignity to a mistake that can’t be taken back.</p><p>What I think actually needs to happen is smaller than it sounds, and I don’t think it’s complicated:</p><p><strong>Treat this as its own category of spreadsheet risk</strong> — not a footnote to EuSpRIG’s financial-loss stories, but a distinct kind of harm that deserves to be named and tracked on its own terms.</p><p><strong>Don’t publish the result until verification is done.</strong> Not “verify quickly, then publish” — verify first, and let publication wait for it.</p><p><strong>Make that verification a shared responsibility</strong>, not one person’s sign-off. A second, independent party — someone who didn’t build the original file — checking the number before it becomes real.</p><p><strong>Move as fast as accuracy allows, not faster.</strong> If the schedule is the problem, move the schedule earlier. Don’t compress the one step that exists to catch the mistake.</p><p><strong>Hold the number back until the answer is actually settled.</strong> Not because delay is free — it isn’t — but because a wrong answer delivered on time is worse than a right one delivered a day late.</p><p>None of this requires new technology. It doesn’t require abandoning Excel. It requires deciding, in advance, that for a specific category of decisions — the ones that can ruin a day, a year, a chance — speed is not the thing being optimized for.</p><p>I don’t think that’s a hard problem to describe. I think it’s just a matter of someone deciding to do it.</p><p><em>The anonymized corpus (N=65) is available on </em><a href="https://doi.org/10.5281/zenodo.21307213"><em>Zenodo</em></a><em>. The accompanying preprint, “Spreadsheet-Error Incidents in Japan: A Convenience Sample and the Question of Dignity-Related Risk,” is available </em><a href="https://doi.org/10.5281/zenodo.21307600"><em>here</em></a><em>. Both are early-stage, non-exhaustive work, and I’d welcome anyone who wants to push on it, challenge it, or extend it.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2a2fcbba94fe" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Audit That Wasn’t There — Until It Was]]></title>
            <link>https://medium.com/@yosatonet/the-audit-that-wasnt-there-until-it-was-c0139671c9e5?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/c0139671c9e5</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[eusprig]]></category>
            <category><![CDATA[japan]]></category>
            <category><![CDATA[audit]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Fri, 10 Jul 2026 02:00:36 GMT</pubDate>
            <atom:updated>2026-07-10T02:26:20.399Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A note for EuSpRIG</em></p><p><a href="https://eusprig.org/">https://eusprig.org/</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/608/1*HV5O9O4tE8bg-1U4EK-Tww.png" /></figure><p>I was watching the conference remotely the day after my own papers were presented, still somewhere between jet-lagged and wired. What struck me wasn’t what I expected. I’d braced myself for horror stories — the forensic case studies, the decades of “here is exactly which cell, which formula, which assumption broke” that this community is known for. Instead, most of what I heard was about AI: how to build models that hold up under audit, how to design for auditability from the start. Less autopsy, more architecture.</p><p>And somewhere in that shift, a question settled in that I couldn’t shake: if people here are already this far into designing <em>for</em> auditability, why does this entire body of work — forensic or forward-looking, either one — have almost no equivalent in Japan?</p><p>It’s not that Japanese companies don’t use spreadsheets. I’ve spent thirty years watching them do it, badly and brilliantly, in roughly equal measure. It’s that nobody here seems to be able to <em>say so</em> in public, with evidence, the way EuSpRIG says so. I know this from the inside — I’ve seen things I couldn’t point to, because what I knew came wrapped in confidentiality, or in the vague shrug of “that’s just how it’s done here.” I had the observation. I never had the receipt.</p><p>So I went looking for the receipt.</p><p>Japan has a public audit-report search system, and I started asking it simple questions. How often do auditors, writing in their most detailed and most public disclosures, mention <em>systems</em> — the generic word for automated software — versus how often they mention <em>spreadsheets</em> by name? The gap was not subtle. Systems came up by the hundreds. Spreadsheets came up by the dozens, and I mean that literally — a number small enough to read through by hand in an afternoon.</p><p>That in itself wasn’t surprising. What surprised me was going deeper, into what auditors actually claimed to have <em>done</em> when a spreadsheet was mentioned. Most of the time, it wasn’t much. A comparison against last year’s file. A cross-check against numbers pulled from another system. The kind of verification you could do without ever opening the formula bar — without, as I found myself putting it, so much as double-clicking a cell to see what color the reference trail lights up. Real auditing happens in that click. Almost none of what I found got there.</p><p>I want to be honest about how that felt. For a while, it felt like confirmation of the worst version of my own hypothesis — that the culture I’d been trying to describe for years, the one where formulas carry intent nobody is required to examine, was simply absent from the institutional record. Not hidden. Not confidential. Just never built.</p><p>And then I found two.</p><p>Two companies, audited in the same month, by different engagement teams but the same firm’s office, where the auditors had done the real thing. They’d gone into the spreadsheet, checked the logic against the rule it was supposed to implement, and — in one case — brought in a dedicated technology specialist to verify that the formula logic was doing what it claimed to do across a genuinely complicated allocation problem. This wasn’t a footnote. It was written up with the same seriousness EuSpRIG papers give to the disasters — except here, nothing had gone wrong. It was prevention, documented before the fact, not an autopsy after one.</p><p>I sat with that for a while before I understood why it mattered so much to me. Not because two data points prove a trend — they don’t, and I want to be careful not to claim more than that. What they prove is narrower and, I think, more useful: that the capability exists. Somewhere in this country, at least one office of at least one firm has built the muscle to do what your community has spent decades arguing every auditor should be able to do. It is not evenly distributed. It may not even be <em>documented</em> as a method anywhere except in the two reports where it happened to surface. But it is real, and it is on the public record, in a way my own testimony, however experienced, never could be.</p><p>There’s a vendor-marketing genre here — I suspect you have its cousins wherever you are — that sells the fear of the unaudited spreadsheet as a reason to abandon spreadsheets altogether. Buy the platform, and the audit problem disappears. What I found argues something close to the opposite: the two audits that actually held up to scrutiny didn’t involve anyone abandoning the spreadsheet. They involved someone deciding to look inside it properly. The tool was never the variable. The willingness to check was.</p><p>I don’t know yet whether what I found in that one office is a method that could travel, or a fluke of two good engagement teams happening to land on two demanding clients. I don’t know why the appetite for this kind of scrutiny is so concentrated, or why the vocabulary auditors use to describe <em>failures</em>, once something has already gone wrong, seems to lose the word “spreadsheet” almost entirely — as though the moment a mistake becomes a story about people and process, the tool itself quietly leaves the sentence.</p><p>I’m not ready to turn this into a paper. The sample is too small, the search tool I used has its own blind spots I haven’t fully mapped, and I chased at least one promising lead this week that turned out to be nothing — a multi-year restatement that looked, for a hopeful afternoon, like the kind of spreadsheet disaster your literature is full of, and turned out to be an unrelated tax accounting error. I’m telling you that not to undercut the finding, but because I think you’d want to know the failed hypothesis too. It’s part of the same walk.</p><p>What I have, for now, is smaller and I think sturdier than a paper: proof that the thing I always suspected was true — that somewhere, quietly, someone here is already doing this work at a standard you’d recognize — was never just a feeling. It has a paper trail now. It’s thin. But it’s there.</p><p>I’d be curious what this looks like from where you sit, wherever you’re reading this from. Is the ratio — a great deal of noise about “systems,” and almost none about the tool where the real judgment lives — familiar to you, in your own audit literature? And when your case studies find something that <em>worked</em>, not just something that broke, does it read the same way to you as it did to me — quieter, harder to sell, but somehow more convincing than the disasters ever were?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c0139671c9e5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I can make these in Excel. May I make it?]]></title>
            <link>https://medium.com/@yosatonet/i-can-make-these-in-excel-may-i-make-it-7be4a7f5fa70?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/7be4a7f5fa70</guid>
            <category><![CDATA[grid-layout]]></category>
            <category><![CDATA[japan]]></category>
            <category><![CDATA[graph-paper]]></category>
            <category><![CDATA[grid-paper]]></category>
            <category><![CDATA[excel]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Sun, 05 Jul 2026 08:13:36 GMT</pubDate>
            <atom:updated>2026-07-05T08:13:36.550Z</atom:updated>
            <content:encoded><![CDATA[<p>I have a bit of Excel skills, so I can create various things with Excel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/647/1*rghCYS50dInEJlR6N1l8MQ.png" /></figure><h3>I can create various things by merging Excel borders with cells.</h3><p>For example, a graph that does not use the graph function. Indicate the first scale with a cell.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*WJ-u6a-QpjjJ-hldNWtE5w.png" /></figure><p>You can also make it like this. The cell is marked with a single scale, but it is displayed as a symbol without filling it in.</p><p>I can also draw drawings in Excel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/567/1*jvc9uQG_tS8t78lSBnP15Q.png" /></figure><p>You can also create something like a wiring diagram by connecting the cells.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/764/1*49F1xGnuT26d2mknsXpzXA.png" /></figure><p>I can also create a clean calculation sheet using cell widths and cell joining.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/885/1*OmAVMJWHTbBpj-DG61pG9Q.png" /></figure><h3>Since I can make them, I can make them too, right?</h3><p>In the example above, Excel’s features are fully utilized to achieve a clean layout. That’s wonderful.</p><p>Plus, it’s efficient to create and looks great.</p><h3>The government is also using this method</h3><p>Japan’s Digital Agency is a top-ranking government organization responsible for DX in Japan.</p><p>You can also download files using this method there.</p><p><a href="https://www.digital.go.jp/policies/local_governments/common-feature-specification">https://www.digital.go.jp/policies/local_governments/common-feature-specification</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/1*zNA9chk6iyZu5SBeDDmVTg.png" /><figcaption>住登外者宛名基本情報照会API仕様書, 2026年2月27日</figcaption></figure><p>There was a time when the Minister in charge at the Digital Agency was Diet member Taro Kono.</p><p>Before taking office, he had previously instructed the Ministry of Education on X to completely eliminate Excel documents with this structure.</p><p><a href="https://x.com/konotarogomame/status/793699820712054784">https://x.com/konotarogomame/status/793699820712054784</a></p><p>During his tenure and even in 2026, this document will be available for download.</p><p>Even under the knees of the minister who criticizes it, I can feel the power that continues to exist.</p><h3>Is it really okay?</h3><p>Is that really okay?</p><p>Not bad. But it’s not good.</p><p>Excel is software that performs calculations using tables. It is not layout software. Nor is it graphic software.</p><p>That’s all there is to it.</p><h3>Excel graph paper</h3><p>These features allow Excel sheets to function as graph papers or grid papers by making them narrower column widths and cells square-sized. To fully utilize this feature, cell coupling is necessary.</p><p>This usage is called Excel graph paper in Japan, and in Japanese, it is pronounced ‘Excel Houganshi.’</p><p>In Japan, you can see such usage everywhere — municipalities, administration, education, and business.</p><p>If we were to define it as a bad usage, it should keep decreasing, but it hasn’t decreased at all. In fact, the number of observations continues to grow.</p><p>So, is it a good way to use it? I can’t say for sure.</p><p>This is because it deviates from the concept of calculating with tables.</p><p>In Japan, there are many statements that view Excel graph papers negatively. However, those messages are often on the web or social media. Some people even deny it with anger and see it as a problem.</p><p>However, quantitative research is not conducted from a scientific or business perspective regarding what is evil, and while it can sometimes be observed in society, the actual number is far less or far more than imagined, or even unknown.</p><p>Moreover, within the anger, it seems to be less about dissatisfaction with Excel and more about the individual person with low information literacy who creates such things.</p><p>I feel there is danger that such statements resonate with Google, become scientifically based, and become memes.</p><p>What exists here is simply a fact: just a spreadsheet.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7be4a7f5fa70" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[VLOOKUP Is Not Magic: How Japan’s Ministry of Land Corrupted 510 Public Procurement Records — and…]]></title>
            <link>https://medium.com/@yosatonet/vlookup-is-not-magic-how-japans-ministry-of-land-corrupted-510-public-procurement-records-and-7730b713935d?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/7730b713935d</guid>
            <category><![CDATA[excel-vlookup]]></category>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[risk-management]]></category>
            <category><![CDATA[japan]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Fri, 26 Jun 2026 12:55:44 GMT</pubDate>
            <atom:updated>2026-06-26T13:31:12.552Z</atom:updated>
            <content:encoded><![CDATA[<h3>VLOOKUP Is Not Magic: How Japan’s Ministry of Land Corrupted 510 Public Procurement Records — and Nobody Noticed for Five Years</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/572/1*mDIGkXhKWCtOfdYTVQQhoQ.png" /></figure><h3>Why I Didn’t Know About This for a Year</h3><p>In 2026, someone in the EuSpRIG community asked me: “Are there any spreadsheet horror stories from Japan?”</p><p><a href="https://eusprig.org/">https://eusprig.org/</a></p><p>I searched. I found one. A data integrity incident at Japan’s Ministry of Land, Infrastructure, Transport and Tourism (MLIT), which came to light in July 2025.</p><p><a href="https://www.mlit.go.jp/report/press/kanbo08_hh_001241.html">https://www.mlit.go.jp/report/press/kanbo08_hh_001241.html</a></p><p>I have been involved with Excel for 30 years. I have presented papers at EuSpRIG. And yet I had no idea this incident existed until I went looking for horror stories.</p><p>That invisibility is not incidental. It is part of the story.</p><h3>Background: How Japan Procures Engineering Consultancy Work</h3><p>To understand what happened, a brief explanation of the Japanese public procurement system is necessary.</p><p>When government agencies in Japan commission engineering consultancy work — road design, river management, structural surveys — they use a competitive evaluation system called the <em>Comprehensive Evaluation Tender Method</em> (総合評価落札方式). This is not a simple lowest-bid system. Technical capability is scored alongside price.</p><p>A key component of that technical score is the contractor’s past performance rating: the <em>gyomu seiiseki hyotei-ten</em> (業務成績評定点), or business performance evaluation score. After completing a project, each contracting company and individual engineer receives a numerical score from the issuing agency. These scores are registered in a national database called TECRIS (Technical Consulting Records Information System), managed by JACIC (Japan Construction Information Center). When companies bid on future work, their historical scores from TECRIS are pulled and factored into the evaluation.</p><p>In short: your past scores determine your competitive position in future bids. They matter enormously.</p><p>The scores were wrong.</p><h3>What Happened</h3><p>On July 23, 2025, a company participating in a tender issued by the Kinki Regional Development Bureau (part of MLIT) looked at the bid documentation and noticed something. Their own score was lower than it should have been.</p><p>They raised the issue. The bureau investigated. They found errors — not just in that one tender, but throughout the historical records.</p><p>What followed was a nationwide halt. From August 4, 2025, all regional development bureaus suspended procurement procedures: no award decisions, no contract signings. Approximately 61,000 completed projects were reviewed.</p><p>The result: <strong>510 errors</strong> across the country. In the Kinki bureau alone — the origin of the problem — 242 projects were affected, with errors at 365 individual data points.</p><p>At 0.83%, this might appear to be a small error rate. It is not. In the context of public procurement integrity, the baseline must be zero. These were not rounding errors. They were incorrect performance scores that determined which companies won public contracts. The correct figure, if applied retroactively, could change who was awarded past work.</p><p>How it was reported in Japan: as a “database input error.” Some outlets mentioned Excel, briefly. None framed it as a spreadsheet risk incident. None called it what it was.</p><h3>Timeline: Five Years of Undetected Errors</h3><p>What makes this a horror story is not the scale. It is the structure of invisibility.</p><p>FY2020 (completed work) Scoring staff enter errors into internal system. Engineer scores not yet required for registration</p><p>June 2022 FY2021 data registered to JACIC via VLOOKUP. 4 errors introduced through duplicate project names</p><p><strong>November 2022</strong> <strong>Critical event</strong>: FY2020–2021 engineer scores retroactively registered. VLOOKUP used to merge data tables. 89 mismatches introduced by duplicate project names</p><p>June 2023 FY2022 data registered manually (no VLOOKUP). Transposition errors and omissions: 68 errors</p><p>July 23, 2025 A single company notices its score is wrong and reports it</p><p>August 4, 2025 All regional bureaus suspend procurement nationwide</p><p>August 18, 2025 Corrections complete; procedures resume</p><p>August 19, 2025 National total confirmed: 510 errors across ~61,000 records</p><p>September 29, 2025 Expert committee convened (Kinki bureau)</p><p>November 6, 2025 Expert committee report submitted</p><p>For up to five years, the errors were undetectable. Why? Because the companies whose scores were stored in TECRIS had no way to view their own records. There was no mechanism for a contractor to check whether their registered score matched the score they had been notified of.</p><p>The report offers no explanation for why this was the case.</p><p>Discovery was accidental. One company, on one day, happened to notice. If they had not, the errors would still be there.</p><p>One further question the report leaves open: after discovery, approximately 61,000 records were reviewed and declared corrected within two weeks. The verification method was cross-referencing TECRIS entries against the paper notifications previously issued to contractors. But whether those notifications themselves were accurate is a separate question. The report acknowledges this directly: “There is no guarantee that the review was complete. Errors that were not identified may still exist.” The figure of 510 is not a definitive count.</p><h3>What the Report Says</h3><p>The Kinki Regional Development Bureau convened an expert committee, which published its report in November 2025. The 365 errors in the Kinki bureau were classified by origin:</p><ul><li><strong>Scoring staff errors</strong>: 185 instances (manual input errors, transcription errors, draft values left as final)</li><li><strong>Registration staff transcription errors</strong>: 170 instances</li><li><strong>Errors introduced during JACIC matching adjustments</strong>: 10 instances</li></ul><p>The single largest cause was VLOOKUP misreference. In November 2022, registration staff used VLOOKUP to merge two Excel tables when retroactively entering engineer scores for FY2020–2021. They used project name as the lookup key. Multiple projects shared identical names. VLOOKUP returned the first match — silently, without error — pulling scores from the wrong project.</p><p>The report’s recommendations: establish written procedures, integrate CCMS and TECRIS so data flows automatically between systems, strengthen staff training, pursue DX (digital transformation).</p><p>These are reasonable. But notice what is missing.</p><p>The report identified VLOOKUP misuse as a root cause. It then recommended removing VLOOKUP from the process — through system integration. It did not ask <em>why</em> VLOOKUP was misused. “Remove the tool” and “understand why the tool was misused” are different interventions. Only one of them addresses the cultural substrate.</p><h3>What the Report Does Not Say</h3><p>Consider the composition of the expert committee:</p><ul><li><strong>Chair</strong>: Masaki Abe, Attorney (Aiman Law Office)</li><li><strong>Member</strong>: Atsushi Iizuka, Professor, Chuo University (construction management)</li><li><strong>Member</strong>: Masahide Hotta, Professor, University of Tokyo (civil engineering)</li></ul><p>Law. Civil engineering. Construction management. All appropriate for the domain. Not one person with expertise in spreadsheet risk.</p><p>This is why the report cannot ask the questions that matter most: Why was Excel operating as an unofficial intermediary between two official systems? Why was VLOOKUP being used without understanding of key uniqueness? What does it mean that the person responsible for data integrity was using a function they did not fully understand?</p><p>These questions require a different frame — one that does not exist in Japan as an established discipline. There is no Japanese equivalent of EuSpRIG. There is no recognized profession of spreadsheet risk management. The concept has no name in this context.</p><p>If the concept has no name, it cannot be on the agenda. If it is not on the agenda, no one with relevant expertise is invited to the table.</p><h3>VLOOKUP as Magic</h3><p>I have been teaching Excel in corporate settings for 25 years. What I am about to describe is not speculation. It is what I have observed repeatedly.</p><p>When I run corporate training sessions, participants rarely ask me to teach VLOOKUP. They already know it — or think they do. What they ask for are functions they have not heard of. When I teach ROUND, they are surprised and engaged. Basic numerical handling was unknown to them, but VLOOKUP was not.</p><p>When I probe their understanding of VLOOKUP, what I find, consistently, is this: they understand it as magic. Not as a lookup function with defined behavior. As a mechanism that “fetches something from another table” — a black box that Excel operates on their behalf.</p><p>The critical implication of treating VLOOKUP as magic: they do not know what happens when the key is not unique. They do not know that VLOOKUP returns the first match silently. They do not know that a duplicate project name will cause it to fetch the wrong row without warning.</p><p>That is precisely what happened in November 2022.</p><p>VLOOKUP is a powerful function. Properly understood, it implements relational lookup — conceptually similar to a SELECT in SQL, retrieving a value from one table based on a unique key in another. Used with that understanding, it is reliable and appropriate.</p><p>Used as magic, it introduces errors that look exactly like correct data.</p><h3>Why VLOOKUP Becomes Magic: Evidence from Job Postings</h3><p>This is not a story about individual incompetence. It is a story about a structural condition.</p><p>In Japan’s labor market, “can use VLOOKUP” has become a proxy for “can use Excel.” Job postings for office and administrative roles routinely list VLOOKUP proficiency as a required or preferred skill. Interviews for clerical positions include the question: “Can you use VLOOKUP?” One major career website published an article titled, without apparent irony: “VLOOKUP is the benchmark for Excel proficiency.”</p><p>In this environment, knowing VLOOKUP is the goal. Understanding why it works, and under what conditions it fails, is not part of the goal.</p><p>The result is a predictable inversion: workers who know VLOOKUP but not ROUND. Who can fetch data from another table, but cannot tell you what happens when two rows share the same key. Who have memorized a function without understanding the data model it requires.</p><p>This is the predictable output of rote-based Excel education. Teach the operation. Do not teach the concept. The operation can be memorized and demonstrated in an interview. The concept cannot be tested with a yes/no question.</p><p>Certiport describes the Microsoft Office Specialist (MOS) certification as offering “hands-on, live-in-the-application testing” and positions Microsoft Office proficiency as “more than a basic requirement in the workforce — it’s a powerful advantage.” The implication is that MOS certifies genuine, applicable understanding — not just the ability to recall a sequence of steps.</p><p>If that goal were genuinely achieved in practice, incidents like this one would be less likely. A workforce that truly understood the tools they were using — not just the operations, but the underlying concepts — would be less prone to treating VLOOKUP as a black box.</p><p><a href="https://certiport.pearsonvue.com/Certifications/Microsoft/MOS/Overview">https://certiport.pearsonvue.com/Certifications/Microsoft/MOS/Overview</a></p><p>But in Japan, MOS education has become oriented toward passing the exam rather than building the understanding the exam was designed to certify. Preparation materials focus on memorizing correct sequences of operations. When a certification becomes the goal rather than the evidence, what the certification was meant to prove disappears.</p><h3>This Is a Spreadsheet Risk Incident</h3><p>Let me be precise about what Excel was doing in this system.</p><p>Two official systems existed: CCMS (a contract management system used by all regional bureaus) and TECRIS (the national performance database managed by JACIC). These systems did not talk to each other. The transfer of data between them required manual intervention.</p><p>Individual staff members — different people each year, with no written procedure and no shared methodology — bridged this gap using Excel. Each developed their own approach. Some used VLOOKUP. Some transferred data manually. Some extracted from CCMS, some from TECRIS, some from older internal systems. The Excel files were the connective tissue between two official databases, operating entirely outside any formal control framework.</p><p>This is shadow spreadsheet infrastructure. Not a rogue actor creating an unauthorized tool — a systemic pattern in which Excel fills the gaps left by inadequate system integration, invisible to governance and uncontrolled by process.</p><p>VLOOKUP then operated within that shadow infrastructure, silently, for years.</p><p>The report concludes: “No deliberate or intentional errors were found. The primary cause is attributed to the complexity of the system from performance evaluation to TECRIS registration, and to the multiple points of manual entry involved.”</p><p>This is accurate. It is also incomplete. The complexity of the system did not create itself. The multiple points of manual entry existed because Excel was used as an unofficial bridge between systems that should have been integrated. And the use of VLOOKUP without understanding of key uniqueness was not a random individual failure — it was the predictable output of an education and hiring culture that treats VLOOKUP proficiency as a destination rather than a starting point.</p><p>Japan does not yet have a name for this. That is why a committee of lawyers and engineers could analyze the failure in detail, identify VLOOKUP as the mechanism, and still not ask the question that would lead to a cultural intervention.</p><h3>Conclusion</h3><p><strong>Update (December 2025): The rankings did change.</strong></p><p>A subsequent impact investigation by the Shikoku Regional Development Bureau confirmed that in one contract from 2025, the first and second-ranked bidders had been swapped due to the scoring errors — meaning a company that should not have won the contract was awarded it. The bureau apologized to the company that should have ranked first. “Could change” became “did change.”</p><p>510 errors. Possibly more.</p><p>Five years undetected, because the people whose scores were at stake had no way to check them. Discovered by accident, because one company looked at a document and noticed something was wrong.</p><p>When incidents like this surface in Japan, the instinctive response is often: “This is why we should stop using Excel.” That response misses the point. The problem is not Excel. The problem is Excel used unofficially, unconsciously, and without understanding — as magic — in a context where it was never designed to operate, without oversight, without documentation, and without anyone in the room who recognized what was happening as a category of risk with a name.</p><p>The category has a name. It is spreadsheet risk.</p><p>Japan does not know that yet. And until it does, the same failure will recur in different forms, in different ministries, on different spreadsheets — silently, without error messages, returning plausible-looking wrong answers to people who do not know to question them.</p><p><em>Primary sources: Kinki Regional Development Bureau, “Report on the Performance Score Input Error Incident” (Expert Committee, November 6, 2025); MLIT press releases, August 8 and 19, 2025. All official documents in Japanese.</em></p><p><a href="https://www.kkr.mlit.go.jp/n_info/c9us5e000005qzoh.html">https://www.kkr.mlit.go.jp/n_info/c9us5e000005qzoh.html</a></p><p><em>The author has presented research at EuSpRIG 2026. ORCID: 0009–0007–0992–8410</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7730b713935d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why Spreadsheet Risk Surveys Don’t Work in Japan]]></title>
            <link>https://medium.com/@yosatonet/why-spreadsheet-risk-surveys-dont-work-in-japan-39e25b573da6?source=rss-df612c2ee268------2</link>
            <guid isPermaLink="false">https://medium.com/p/39e25b573da6</guid>
            <category><![CDATA[surveys]]></category>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[risk]]></category>
            <category><![CDATA[spreadsheets]]></category>
            <category><![CDATA[japan]]></category>
            <dc:creator><![CDATA[Yoshihiro Sato]]></dc:creator>
            <pubDate>Wed, 24 Jun 2026 21:16:44 GMT</pubDate>
            <atom:updated>2026-06-24T21:16:44.640Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/572/1*uZjIC4kGsOL1Xsz1Ndpr4A.png" /></figure><h3>A note for EuSpRIG researchers considering Japan as a study target</h3><p>I recently read a 2021 EuSpRIG paper by Khwima Mckinley Mkamanga and Simon Thorne titled “Exploring Spreadsheet Use and Practices in a Technologically Constrained Setting,” which examined spreadsheet practices at a water utility in Malawi. The study revealed that employees were aware of spreadsheet risks but lacked systematic methods to address them — a pattern familiar to many EuSpRIG readers.</p><p>Reading it, I found myself asking: could this survey be conducted in Japan?</p><p>My answer, based on over 25 years of Excel education experience in Japan, is: probably not — at least not without significant redesign.</p><h3>The survey assumes a shared concept that doesn’t exist here</h3><p>The questionnaire in the Malawi study asks whether users test formulas (Q12), document their models (Q14), and back up critical files (Q20). These questions presuppose that respondents understand what testing, documentation, and version control mean in a spreadsheet context.</p><p>In Japan, most Excel users don’t share that conceptual framework. Formulas are things you create, not things you verify. Backups, if considered at all, mean hitting Ctrl+S before the computer crashes. The idea that a spreadsheet could fail silently — producing wrong numbers without any visible error — is genuinely unfamiliar.</p><p>Here is a concrete paradox: by the time a Japanese Excel user can understand and answer this survey, they have almost certainly already moved past the problems it is trying to measure. The survey selects for its own solution.</p><h3>How risk is understood in Japan — and how it isn’t</h3><p>When Japanese office workers talk about spreadsheet risk, they usually mean one specific scenario: someone left a shared file open, someone else overwrote the formulas, or a blank sheet was saved over the original. The file is broken. Work stops.</p><p>This is not framed as a risk management problem. It is framed as a software limitation — “Excel can’t handle simultaneous editing” — and the conclusion drawn is often “we should stop using Excel.” CRM vendors are particularly vocal about this framing.</p><p>The deeper issue is that there is no verification culture. Nobody checks whether formulas are correct before trusting the output. Nobody asks who is responsible for spreadsheet quality. When something goes wrong, individual blame is assigned — quickly and heavily — rather than examining the system that allowed the error to occur.</p><p>EuSpRIG readers will recognize the irony: Japan is the country that gave the world poka-yoke, the principle of designing systems to prevent errors before they happen. That concept is well-known in manufacturing. It has not crossed over into office spreadsheet practice.</p><h3>Replacing Excel doesn’t solve the underlying problem</h3><p>The common response to spreadsheet problems in Japan — switching to a CRM or SaaS tool — does not address the absence of verification culture. It moves the problem into a black box. The logic inside the new system is less visible than a spreadsheet formula, and the data is harder to interrogate.</p><p>CRM systems and spreadsheets serve fundamentally different purposes. A CRM records business activity and automates routine processes. Excel — used properly — is where you take that exported data and explore it: pivot tables, pattern recognition, hypothesis testing. These are complementary tools, not substitutes.</p><p>When the tool changes but the culture doesn’t, errors become harder to find, not easier.</p><h3>A note for researchers</h3><p>If you are considering Japan as a study site for spreadsheet risk research, the Malawi questionnaire is a useful starting point — but it will need substantial adaptation. More importantly, you may find that the preliminary conceptual work (establishing shared vocabulary, defining what “risk” means in this context) takes longer than the survey itself.</p><p>Japan is not a technologically constrained environment. Infrastructure is not the limiting factor. What is missing is the conceptual framework that makes questions like “do you test your spreadsheets?” meaningful to ask.</p><p>That gap, I would argue, is worth studying in its own right.</p><p><em>Reference: Khwima Mckinley Mkamanga, Simon Thorne, “Exploring Spreadsheet Use and Practices in a Technologically Constrained Setting,” EuSpRIG Proceedings 2021, ISBN: 978–1–905404–57–5. arXiv:2201.07696 [cs.SE]</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=39e25b573da6" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>