<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Skills Matrix on Ghost in the data</title><link>https://ghostinthedata.info/tags/skills-matrix/</link><description>Ghost in the data</description><generator>Hugo -- gohugo.io</generator><language>en</language><copyright>Ghost in the data</copyright><lastBuildDate>Sat, 21 Nov 2026 09:00:00 +1100</lastBuildDate><atom:link href="https://ghostinthedata.info/tags/skills-matrix/index.xml" rel="self" type="application/rss+xml"/><item><title>Nobody Reads 7,000 Words the Night Before an Interview</title><link>https://ghostinthedata.info/posts/2026/2026-11-21-night-before-the-interview/</link><pubDate>Sat, 21 Nov 2026 09:00:00 +1100</pubDate><guid>https://ghostinthedata.info/posts/2026/2026-11-21-night-before-the-interview/</guid><author>Chris Hillman</author><description>Seven thousand words on data engineering interviews won't help you the night before. Why I wrote the short version.</description><content:encoded>&lt;p>Across two posts, I&amp;rsquo;ve written close to 7,000 words on data engineering interviews. &lt;a href="https://ghostinthedata.info/posts/2025/2025-02-22-the-interview/">Mastering Data Interviews&lt;/a> walked through the formats and how to prepare for each one. &lt;a href="https://ghostinthedata.info/posts/2026/2026-01-11-guerrilla-interviews/">The Guerrilla Guide to Data Engineering Interviews&lt;/a> made the case that you get hired on what you&amp;rsquo;ve built far more than on what you&amp;rsquo;ve memorised.&lt;/p>
&lt;p>All of that is no help the night before an interview, when there&amp;rsquo;s a tickle of anxiety about how you&amp;rsquo;ll do and you&amp;rsquo;re trying to remember how you&amp;rsquo;d explain a window function (it always comes up). It&amp;rsquo;s even less help to the person interviewing you, who opens their notes five minutes before an interview they forgot was in the diary.&lt;/p>
&lt;p>Hardly anyone reads interview advice calmly. So I&amp;rsquo;ve written the short version: the &lt;a href="https://ghostinthedata.info/resources/interview-kit/">Data Engineering Interview Kit&lt;/a>, a set of one-page cheat sheets.&lt;/p>
&lt;/br>
&lt;hr>
&lt;h3 id="built-for-the-moment-youll-use-it">Built for the moment you&amp;rsquo;ll use it&lt;/h3>
&lt;/br>
&lt;p>Pilots have a version of this problem. Nobody wants to be paging through the full aircraft manual when a warning light comes on, so the flight deck carries a quick reference handbook: the procedures for abnormal and emergency situations, laid out to be used under pressure.&lt;/p>
&lt;p>The kit borrows that idea. It&amp;rsquo;s organised around five moments: a week out, the day before, in the room, when it goes sideways, and afterwards. You work out where you are, then pick up the sheet for it.&lt;/p>
&lt;p>So a week out, the candidate is ideally starting their story bank, while the interviewer levels the role and plans short interviews split into exercises. In the room, the candidate&amp;rsquo;s sheet says to ask before you build, and the interviewer&amp;rsquo;s says to help with the process but never with the answer. When it goes sideways (and I&amp;rsquo;ve been in plenty), there&amp;rsquo;s a line for that on both sides, whether it&amp;rsquo;s a candidate who blanks or an interviewer facing a story that&amp;rsquo;s all &amp;ldquo;we&amp;rdquo; and no &amp;ldquo;I&amp;rdquo;.&lt;/p>
&lt;/br>
&lt;hr>
&lt;h3 id="same-map-both-sides">Same map, both sides&lt;/h3>
&lt;/br>
&lt;p>Both of my earlier posts were written for candidates, and so is most interview prep. The person asking the questions often gets little more than a calendar invite and a CV.&lt;/p>
&lt;p>So the kit is written for both sides of the table, and both sides work from the same skills matrix: eleven skills, each described at three levels, from Junior Data Engineer through Data Engineer to Senior Data Engineer. Candidates rate themselves against it before the interview. Interviewers score with it in the room.&lt;/p>
&lt;p>Every scenario comes as a pair: a candidate page with the shape of a good answer and the traps to avoid, and an interviewer card for the same problem with what to read aloud, the follow-up questions and what to listen for at each level.&lt;/p>
&lt;p>The obvious objection is that I&amp;rsquo;ve published the sheet interviewers score with, so candidates can read exactly what they&amp;rsquo;re looking for. I&amp;rsquo;m fine with that. The matrix describes things you can only show by doing them: checking the grain before you write SQL, catching your own bug, owning an incident instead of blaming the vendor. Reading about them won&amp;rsquo;t make you do them with someone watching. Practising will. And if a candidate prepares against the matrix and clears the bar, that&amp;rsquo;s exactly the person you were hoping to hire.&lt;/p>
&lt;/br>
&lt;hr>
&lt;h3 id="if-youre-the-one-asking-the-questions">If you&amp;rsquo;re the one asking the questions&lt;/h3>
&lt;/br>
&lt;p>The SQL exercise shows what asking the right questions looks like. Every candidate gets the same brief: an orders table loaded from the source system&amp;rsquo;s change feed, so there&amp;rsquo;s one row per change event, and a daily report that needs each customer&amp;rsquo;s most recent order and its current status.&lt;/p>
&lt;p>The follow-up questions do the levelling:&lt;/p>
&lt;ul>
&lt;li>&lt;strong>Start&lt;/strong>, with everyone: &amp;ldquo;What should one row of your output represent?&amp;rdquo;&lt;/li>
&lt;li>&lt;strong>Push&lt;/strong>, for Data Engineer roles: &amp;ldquo;Order 103 has two events at the same time. Which one wins, and why?&amp;rdquo;&lt;/li>
&lt;li>&lt;strong>Stretch&lt;/strong>, for Senior roles, or when someone&amp;rsquo;s flying: &amp;ldquo;This runs hourly on 3 billion rows. What changes?&amp;rdquo;&lt;/li>
&lt;/ul>
&lt;p>A junior who gets the joins and aggregation right and checks the result against the sample is doing well. A senior should end up talking about an incremental design with a watermark that&amp;rsquo;s safe to rerun, and a test for the output grain. It&amp;rsquo;s the same brief with very different conversations, and everyone gets a fair go at getting as far as they can.&lt;/p>
&lt;p>A side note for candidates: if you&amp;rsquo;re going for a junior role and find yourself flying through the Stretch questions, that doesn&amp;rsquo;t mean you&amp;rsquo;re ready for a senior role. It means you understand the context well, which is a step in the right direction. But you still need the experience of being in the deep end over and over, and seeing different problems come up. Think of a basketball player. When you&amp;rsquo;re starting out, you might already shoot hoops amazingly. What makes you an NBA player is the practice, and doing it under pressure, game after game. The same applies here.&lt;/p>
&lt;p>For interviewers, the card for that exercise also has a few rules I&amp;rsquo;d want anyone running an interview to follow, kit or no kit. Give every candidate the same brief, the same sample data and the same answers to their questions. Tell them at the start that asking questions is part of the exercise, and let them run the query and look up syntax, because nobody writes SQL blind at work. If they freeze, offer a reset and note the hint on the scorecard, because using a hint well is evidence too. And my favourite line in the whole kit: &amp;ldquo;Don&amp;rsquo;t mark people down for typos or slow typing. You&amp;rsquo;re scoring how they think, and how they recover when something&amp;rsquo;s wrong.&amp;rdquo;&lt;/p>
&lt;p>None of this is my invention. Organisational psychologists call it a structured interview: the same questions for everyone, and a scoring guide written before anyone walks into the room. A &lt;a href="https://www.siop.org/tip-article/is-cognitive-ability-the-best-predictor-of-job-performance" target="_blank" rel="noopener">2022 re-analysis of the hiring research&lt;/a> by Paul Sackett and colleagues put structured interviews at the top of the list for predicting job performance. The same work found that most methods predict less than we used to think, which is a good argument for writing your scores down before the debrief starts and anyone&amp;rsquo;s gut gets a vote.&lt;/p>
&lt;/br>
&lt;hr>
&lt;h3 id="if-youre-the-candidate">If you&amp;rsquo;re the candidate&lt;/h3>
&lt;/br>
&lt;p>Start with the skills matrix. Pick the level of the job you&amp;rsquo;re applying for, read that level&amp;rsquo;s description for each skill, and tick one box: not yet, meets or exceeds. Then write down one real example for each tick. If you can&amp;rsquo;t think of one, don&amp;rsquo;t tick it.&lt;/p>
&lt;p>The sheet has a rule of thumb that works for your CV as well: count a skill if you&amp;rsquo;ve done it for real three times and can explain each choice. If it&amp;rsquo;s your first role, one project you can explain properly is enough for &amp;ldquo;meets&amp;rdquo;.&lt;/p>
&lt;p>That&amp;rsquo;s the Guerrilla Guide&amp;rsquo;s argument in a smaller box. I&amp;rsquo;ve watched strong engineers fumble interviews because they gave a generic answer where a specific story was needed, while weaker engineers with a well-rehearsed story sailed through. A one-page story bank and one scenario per skill rehearsed out loud will do more for you in the week before than another 7,000 words of reading. Mine included.&lt;/p>
&lt;p>It also helps to know what&amp;rsquo;s on the other side of the card. &amp;ldquo;Spots their own bug and says so calmly&amp;rdquo; is on the interviewer&amp;rsquo;s list of good signs. If you catch one halfway through, say so.&lt;/p>
&lt;/br>
&lt;hr>
&lt;h3 id="what-it-costs">What it costs&lt;/h3>
&lt;/br>
&lt;p>&amp;ldquo;Why are you charging for it?&amp;rdquo; Fair question. There&amp;rsquo;s a lot of time and effort in it. I&amp;rsquo;ve been putting it together over the last few years, in different formats, mostly from the interviewer&amp;rsquo;s perspective. That said, the two posts I mentioned at the start cover similar ground, and they&amp;rsquo;re free.&lt;/p>
&lt;p>What it really costs is preparation on one side of the table and honesty on the other. Interviewers have to do their thinking before the interview instead of during it: level the role, pick the exercises, change the numbers, the names and the twist so the exercise isn&amp;rsquo;t one a candidate has already read, and score before anyone speaks in the debrief. Candidates have to rate themselves against a matrix that will tell them where they actually sit, which isn&amp;rsquo;t always where they hoped.&lt;/p>
&lt;p>In return you get a shorter, fairer process. Instead of one long day of interviews, each interview is built from short exercises, and each exercise scores only the one or two skills it was set up to measure. Both people in the room are working from the same picture of what good looks like.&lt;/p>
&lt;p>It stops at senior, so staff, principal and lead roles aren&amp;rsquo;t in this edition. And the risk I&amp;rsquo;d keep an eye on is the kit turning into a script. A cheat sheet can stop you underselling what you&amp;rsquo;ve done, but it can&amp;rsquo;t give you experience you haven&amp;rsquo;t had, and an interviewer who reads the card aloud without looking up has missed the point of the follow-up questions.&lt;/p>
&lt;p>The long version is still there, all 7,000 words of it. The &lt;a href="https://ghostinthedata.info/resources/interview-kit/">short version&lt;/a> is for the night before, and for five minutes before the interview you forgot was in the diary.&lt;/p>
&lt;p>&lt;/br>&lt;/br>&lt;/p></content:encoded><category>Career</category><category>Data Engineering</category><category>Leadership</category><category>Interviews</category><category>Hiring</category><category>Career Development</category><category>Technical Assessment</category><category>Skills Matrix</category><category>Structured Interviews</category><category>SQL</category></item></channel></rss>