D8A cheatsheets · 27 sheets
One page. One concept.
The reference sheets we post every day, in full. Read them here or grab the print-ready PDF, free, no signup.
Data Analyst
8 sheets#026 · free PDFHow can a chart mislead without using fake data?A 0.9 point gap shown as 4x growth, and the axis never lied. The five tricks that make honest numbers tell dishonest stories, plus the 5-second audit that catches all of them.August 18, 2026 · 4 min#022 · free PDFGROUP BY vs PARTITION BY: what is the actual difference?GROUP BY collapses 200 rows into 3. PARTITION BY keeps all 200 and adds group context to each. The two queries side by side, the rule of thumb that picks between them, and the window-filter trap.August 14, 2026 · 2 min#018 · free PDFWhy are aggregate functions not allowed in WHERE?WHERE AVG(salary) > 60000 throws an error every time. Why WHERE cannot see aggregates, the one-word fix with HAVING, and the performance trap of filtering rows in the wrong clause.August 10, 2026 · 2 min#015 · free PDFWhy does WHERE email = NULL return zero rows?The query looks right, runs without an error, and can never match a row. Three-valued logic, IS NULL, COALESCE, and the two NULL traps that empty your results.August 7, 2026 · 2 min#009 · free PDFSQL window functions: why does GROUP BY reject your column?A value per row, computed over a group, without collapsing anything. The anatomy, the ranking trio, and the top-N interview pattern.July 24, 2026 · 2 min#001 · free PDFSQL JOINs explained: why does your query return duplicate rows?The 4 joins you actually use, plus the 2 traps that silently produce wrong numbers. Free one-page PDF included.July 24, 2026 · 3 min#005 · free PDFWhy does your SQL alias error? The execution orderYou write queries top to bottom. The engine runs them in a different order. That one fact explains the alias error, WHERE vs HAVING, and more.July 24, 2026 · 2 min#013 · free PDFWhich chart should you use? The decision gridNine questions, nine answers. Pick charts by what you are asking the data, not by what looks impressive. Free print-ready PDF.July 24, 2026 · 2 min
Data Scientist
7 sheets#027 · free PDFShould I learn pandas or SQL first?Same job, two tools, one learning order. What SQL and pandas are each unbeatable at, the pipeline pattern that uses both, and why the jobs data settles the which-first question.August 19, 2026 · 2 min#023 · free PDFBias vs variance: which one is breaking your model?Models fail in two opposite directions, and the fixes are opposites too. How to read the diagnosis off your train and validation errors, the fix list for each disease, and why 'more data' cures only one of them.August 15, 2026 · 2 min#019 · free PDFWhy does your model ace the backtest and fail in production?A 0.94 backtest that becomes 0.51 live is almost always a split problem. The three sets and what each is for, the order of operations that prevents leakage, and the four split strategies that cover nearly every dataset.August 11, 2026 · 2 min#006 · free PDFPrecision vs recall vs F1: which metric should you use?Two questions, two metrics, one tradeoff. Plus the class-imbalance trap that makes accuracy meaningless. Free one-page PDF.July 24, 2026 · 2 min#002 · free PDFWhich pandas functions do you actually need? The 8 verbsFilter, select, sort, transform, group, aggregate, join, reshape. The pandas core in one page, with a free print-ready PDF.July 24, 2026 · 2 min#010 · free PDFOverfitting explained: why does your model fail the test set?The difference between learning the pattern and memorizing the noise, how to diagnose which one your model did, and the five fixes.July 24, 2026 · 2 min#014 · free PDFConfusion matrix explained: which model mistake costs you money?Rows, columns, the diagonal, the two error types, and the four formulas that fall out of one small grid. Free one-page PDF.July 24, 2026 · 1 min
Data Engineer
6 sheets#024 · free PDFBatch vs streaming: do you actually need real-time data?Nobody needs 'real-time'. They need a freshness number, and most have never been asked for it. The one question that settles the choice, the honest cost gap, and the micro-batch middle that covers almost everything.August 16, 2026 · 2 min#020 · free PDFData warehouse, data lake or lakehouse: which one do you actually need?Three storage architectures, one decision. What actually separates a warehouse from a lake from a lakehouse, the questions that pick between them, and the data swamp trap that eats unowned lakes.August 12, 2026 · 2 min#016 · free PDFWhy is my SQL query slow even though the column is indexed?The index exists and the engine ignores it. The sorted-copy mental model, the sargable rewrite that wakes it up, the leftmost-prefix rule, and when an index actually hurts.August 8, 2026 · 3 min#007 · free PDFStar schema vs snowflake: which one should you use?The two classic warehouse shapes, when each one wins, and the vocabulary interviewers expect. Diagrams included, plus a free PDF.July 24, 2026 · 2 min#011 · free PDFOLTP vs OLAP: why is your dashboard slowing production?Why the database that runs your app is the wrong place for your dashboards. Rows vs columns, the flow, and the prod-DB trap.July 24, 2026 · 2 min#003 · free PDFETL vs ELT: which one should you choose?Where the Transform step runs decides your tooling, your costs, and who gets to shape the data. The full comparison, plus a free PDF.July 24, 2026 · 2 min
Business Analyst
6 sheets#025 · free PDFHow do you write acceptance criteria that actually prevent rework?A ticket that says 'works correctly' guarantees a sprint review that says 'that is not what I meant'. Given/When/Then with real before-and-afters, the five-point review checklist, and the three cases every story needs.August 17, 2026 · 2 min#021 · free PDFHow do you prioritize when everything is marked critical?47 of 52 tickets marked critical is not a priority list, it is a queue. The MoSCoW method on one page: the four buckets, the 60% rule that keeps Musts honest, one test per bucket, and the 45-minute meeting script.August 13, 2026 · 2 min#017 · free PDFWhat is a RACI matrix and how do you build one that works?Responsible, Accountable, Consulted, Informed: the one-page grid that stops tasks dying unowned. The four letters, the golden rules, a worked example, and the smells to fix.August 9, 2026 · 2 min#008 · free PDFWhat makes a good user story? The three-line anatomyThe three-line template, the INVEST test, and acceptance criteria that make developers say thank you. Free one-page PDF.July 24, 2026 · 2 min#004 · free PDFWhat makes a KPI actually useful? The SMART testThe SMART test, the difference between a KPI and a metric, and the red flags that mean your KPI is decoration. Free one-page PDF.July 24, 2026 · 2 min#012 · free PDFThe 5 Whys method: how do you find a root cause?The cheapest root-cause tool there is: a worked example, the rules that keep it honest, and the trap that ruins it.July 24, 2026 · 2 min