Where we start
Most of what goes wrong in data work goes wrong in the first meeting, and it goes wrong quietly. The business person says what they want in the vocabulary they have, the data person translates it into the vocabulary they have, and both walk out believing they agreed. The running case for all six sessions is Harbourline, an invented chain of 38 pharmacy and clinic outlets across Singapore and Malaysia, and its operations director Adeline Tan, who has never heard the word pipeline and does not need to. She has a request. You have 45 minutes with her. Everything in the six sessions is about what you ask in those minutes.
★ What you walk out with tonight. A one-page Six Doors sheet for a request you actually have on your desk, two open questions behind each door written in the past tense, a plain teach-back line attached to each one, and a mark on the door your role should open first. Session 3 puts that sheet in front of Adeline and counts what it gets out of her.
Two people, one word, two meanings
5 min liveIn 1990 a Stanford psychology student, Elizabeth Newton, split people into tappers and listeners. Tappers picked a well-known song and tapped its rhythm on a table; listeners had to name it. Before tapping, tappers guessed the listeners would get about half the songs. Listeners got 3 of 120. The tapper cannot stop hearing the tune, so the taps sound obvious; the listener hears knocking. Every discovery meeting has a tapper on each side of the table.
LiveThe tune in your head, and why neither side can hear the other's3 min▶
Newton's tappers were not bad at tapping. They were unable to un-know the song while they tapped, so every knock arrived with the melody attached, and they judged the listener's chances from inside that experience. Economists had named this a year earlier: Camerer, Loewenstein and Weber called it the curse of knowledge, the finding that better-informed people cannot fully set their information aside when predicting what less-informed people will understand.
In a discovery meeting there are two tunes. The business person's tune is the operational worry: which outlets, which supplier, which Monday phone call. The data person's tune is the machinery: exports, joins, refresh cadences, the shape of a table. Each hears their own tune behind every word the other says. "Dashboard" is knocking. To Adeline it means "make the Monday problem go away". To the analyst it means a page with charts on it. Both leave the meeting sure they agreed.
The gap is not closed by either side learning the other's vocabulary. It is closed by questions that make the tune audible: about the decision, the current way of working and the real state of the data, asked in words that need no translation, with the reason for asking said out loud. That is the rest of tonight.
Self-studyThe translator role, and why you should not wait for one2 min▶
In 2018 three McKinsey partners argued in Harvard Business Review that analytics teams needed a named role, the analytics translator, whose job is to make sure the analytics work actually solves the business problem. The article is worth reading for the shape of the role. The number of translators the firm predicted the economy would need circulates widely and was not read at source, so it is not quoted here.
The practical point: most teams do not have one. The translation happens, or fails to happen, in the first meeting, and the person running that meeting is you. Treating the discovery conversation as part of the analyst's craft, rather than somebody else's job, is the position these six sessions take.
The stand-in request. A constructed case, but the shape is the most common one there is. Harbourline's Monday sales report is late and untrusted, so the request that reaches the analyst is "a dashboard". The worry underneath is stock-outs of chronic medication in six Malaysian outlets, and the decision the report is supposed to feed is next week's reorder quantity per outlet, due Tuesday noon. Nothing in the word "dashboard" carries any of that. It comes out in about four minutes if you ask what happens on Monday morning, and never comes out at all if you ask about the data warehouse.
Six Doors
6 min liveA discovery meeting is a corridor with six doors. Behind each one is a kind of fact you need before you can build anything, and each opens to a different kind of question. Most analysts walk straight to the fourth door, data state, because that is where the work lives, and leave the first two shut. The framework is the corridor drawn out, so you can see which doors you opened and which you walked past.
| Door | The question that opens it | What you walk out with | A teach-back line to attach |
|---|---|---|---|
| 1 · Outcome | If this worked, what would be different in six months? Which outlets keep you up at night? | The worry the request stands in for. Harbourline: stock-outs of chronic meds in six outlets, and a write-off target of 1.5 percent | "A dashboard is a tool. I need the outcome, because I can check a tool against an outcome and not the other way round." |
| 2 · Decision | Walk me through Monday morning. Who opens it, what do they do next, by when? When the number is bad, what changes that week? | The action, the person and the deadline. Harbourline: next week's reorder per outlet, three area managers, Tuesday noon | "The deadline tells me how fresh the data has to be. Fresher costs more, so I only build fresh where a decision needs it." |
| 3 · Current way | How is it made today? Can you show me the file and tell me who builds it? Where does each number come from, step by step? | The unwritten spec. Harbourline: Priya, six hours every Monday, an adjustments column nobody can explain | "The current file is the spec. Every column in it is a decision somebody once made, and I would rather inherit those than guess." |
| 4 · Data state | Which systems does one sale touch? Who can log in and pull an export? Do both countries use the same product codes? | Sources, owners, access, freshness. Harbourline: one POS vendor, per-outlet CSV, codes that differ by country, one IT person | "Every hand-off between systems is a place a number can change. I am counting the hand-offs." |
| 5 · Constraints | What is off the table? Budget, time, legal, people. Where do they read it, laptop or phone? When must you show something upward? | The shape of a yes. Harbourline: no budget until Q1, phones on Monday, a board slide in six weeks | "I ask for the limits first so I do not propose something you could never approve." |
| 6 · Success and partnership | How will you know it worked, and what would you stop doing? Have you tried this before? Who could spend two hours a week with me? | The test, the history and the partner. Harbourline: a quiet phone on Monday afternoon, a vendor pilot nobody opened, Priya | "If we agree on the test now, neither of us has to argue about it later." |
LiveDoors 1 and 2: start from the decision, because a number nobody acts on measures nothing3 min▶
Douglas Hubbard's definition of measurement is the whole of door 2 in one line: a measurement is a quantitatively expressed reduction of uncertainty based on one or more observations. Uncertainty about what? About a decision. His method begins by defining the decision before anything is counted, and if you cannot name a decision the number will change, you have not found a measurement yet, you have found a chart.
So the first door is the outcome and the second is the decision, and they come before any question about data. "What happens on Monday morning" gets you the decision (reorder quantities), the person (three area managers), the deadline (Tuesday noon) and, if you listen, the fact that today the number changes nothing because when it looks bad they phone the outlet. That last fact is worth more than any export schedule. It says the report needs to be different, not faster.
Teresa Torres's discovery work puts the same thing in product language: a desired outcome sits above the opportunities, which sit above the solutions. A dashboard is a solution. Asking for the outcome first is how you find out what it is a solution to.
LiveDoors 3 and 4: specifics in the past, and "show me the file"3 min▶
Rob Fitzpatrick wrote The Mom Test for founders talking to customers, and its three rules transfer whole: talk about their life instead of your idea; ask about specifics in the past instead of generics or opinions about the future; talk less and listen more. His verified line on the second rule is the one to remember: "The world's most deadly fluff is: 'I would definitely buy that.' But folks are wildly optimistic about what they would do in the future." For an analyst, "would you use a daily number?" is that fluff. "Tell me about the last time the report changed a decision" is the past, and the past is where the facts are.
Door 3, the current way, is opened by one request: show me the file. The Monday Excel is the spec nobody wrote. Every column is a decision somebody made once, the adjustments column is a fudge that encodes a real problem, and the six hours Priya spends are the cost you are being asked to remove. Door 4, data state, is the door analysts open first and should open fourth: which systems a sale touches, who owns each, who can log in, whether Singapore and Malaysia use the same product codes. DalleMule and Davenport's 2017 frame for data strategy asks whether a business needs a single source of truth or several fit-for-purpose versions; Harbourline, with three versions of the stock spreadsheet in circulation, has the second by accident.
Taiichi Ohno's five whys belong at door 3 as well. His example walks from a stopped machine through a blown fuse, a dry bearing and a failing pump to a missing strainer. "Why is the report late" walks from Monday to Priya to the POS export to product codes that differ by country. Stop at "Priya is slow" and you build the wrong thing.
Self-studyDoors 5 and 6: the shape of a yes, and the partner3 min▶
Constraints are asked for early because they define what a proposal can look like, not because they are interesting. No budget until Q1 means the first deliverable is built with what exists. Phones on Monday means a table nobody will scroll on a phone is not read. A board in six weeks means one honest number by then beats a suite after it. Malaysian loyalty data under PDPA with a lawyer who takes four weeks means anything touching it is a second phase.
Door 6 is two questions. The first is the test: how will you know it worked, and what would you stop doing? Adeline's answer is the best kind, observable and hers: the area managers stop phoning her on Monday afternoon. The second is the partner: who on her side spends two hours a week with you. Torres argues for weekly contact with the people you are building for; the analyst's version is a named person, not a stakeholder group. The person who works with you is the person who can run it after you leave. Fitzpatrick's test for whether a conversation went well is whether the other side gave up something, time, a name, an introduction; Priya, two hours a week, is that something.
Walk the doors in order the first time. Session 3 shows what the order is worth: on the bench, the same questions in a shuffled order average 28 points against 48 in door order, and the worst shuffle scores 8. Once you can walk them blind, session 3 also shows which door an analyst should open widest, and it is door 2.
Why I am asking, in one sentence
4 min liveEvery open question in the bank comes with a second sentence: a plain statement of why you are asking it. "The deadline tells me how fresh the data has to be, and fresher costs more." Said out loud, it does three things at once. It makes the question feel like a reason rather than an interrogation, it earns trust, and it teaches the business person one true thing about how data work is decided. By the end of a 45-minute meeting she has learned ten of them, without a slide.
LivePlain, true, one sentence, said out loud2 min▶
A teach-back line has three properties. It is plain: no word she would have to ask about, so "freshness" not "latency", "the file" not "the extract". It is true: a real principle of the craft, not a pleasantry. And it is one sentence, because two is a lecture and she stops answering.
Three that earn their place in almost any meeting. Before a data-state question: "Every hand-off between systems is a place a number can change, so I am counting the hand-offs." Before a current-way question: "The current file is the spec. Every column is a decision somebody made once, and I would rather inherit those than guess." Before a success question: "If we agree on the test now, neither of us has to argue about it later." Session 4 builds twelve of these and shows how to choose the one concept per meeting that is worth explaining properly.
Self-studyFive kinds of question, and only one of them earns trust2 min▶
Session 3 sorts every question in the bank into five kinds, and the simulator scores them differently. Open questions about the past or present uncover facts and earn a point of trust. Closed yes-or-no questions are cheap and get you one fact at most. Vague questions ("what are your pain points?") cost minutes and get a generic back. Jargon costs twice: she has to ask what you mean, and she is a little less sure you understand her shop. Leading questions are the dangerous kind: "you'd want this daily, right?" gets a "sure", and a belief that is wrong goes into your notes and later into your build.
The trust point is not decoration. On the bench, three of the 24 facts on Adeline's sheet only come out once trust exists, and one of them, that the only IT person is leaving in January, is the one that decides whether anything you build survives the quarter.
Your Six Doors sheet, for a request you have right now
★ 22 min · everyone writesOne page, six rows. Take a real request from your own inbox, or use Adeline's if you do not have one, and build the sheet you will carry into the meeting. Every line is written before the meeting; the meeting is for listening.
Write the request as it was received, in their words. Not your translation. If it arrived as "we need a dashboard", write "we need a dashboard". You will want this sentence back at the end of the meeting to see how far the real request was from it.
Two open questions per door, in the past or present tense. "Tell me about the last time", "walk me through", "show me", "where did that come from". Nothing with "would you". Nothing that can be answered with yes. Write them so a person who has never seen a database can answer every one.
Attach a teach-back line to each question. Plain, true, one sentence. Read each pair aloud once; if the second sentence has a word she would have to ask about, change the word.
Mark the door your role opens first. For an analyst it is door 2, decision: your output is a number somebody acts on, so the action, the person and the deadline are the facts you cannot build without. Session 3 measures what that choice is worth. If you are a scientist, an engineer or building with AI, the sibling course for your role marks a different door, and links back here for the corridor.
Rehearse the opener against Adeline. Her first line is on the page: "Thanks for coming. We need a sales dashboard." Say your first question out loud, with its teach-back. If it took more than twenty seconds, it is two questions; split it.
Keep the sheet after the meeting. Session 6 turns a filled-in sheet into the one-page brief you send her within 24 hours, and session 5 turns her answers into three candidate projects. The same page carries all of it.
Before session 2
30 min- Finish the sheet for your real request: twelve questions, twelve teach-back lines, one mark. Bring it printed; session 3 puts it on the bench.
- Find the tune. For the request you chose, write the one sentence you think the business person would say if you asked "what would be different in six months?" Then write the sentence you fear they would say. Both are guesses; the meeting replaces them.
- Re-read a past request you built something for, and mark which of the six doors you opened. Most people find two. Write down what you would have learned at each door you skipped.
- Read the current-way file for your request if one exists, the Excel or the deck. List every column you cannot explain. Session 2 starts there.
Three questions
Clear all three and session 1 gets stamped in your passport.
1 · Newton's tappers predicted listeners would name about half the songs; listeners named 3 of 120. What does the study say about a discovery meeting?
The curse of knowledge runs both ways. Adeline hears "stock-outs" behind "dashboard"; the analyst hears "charts". Neither is careless, and neither notices, which is why the doors are walked in order rather than trusted to instinct.
2 · Why do the outcome and decision doors come before the data-state door, when the data is where the analyst's work lives?
At Harbourline, "who makes the call and by when" yields Tuesday noon, three area managers, and the fact that today the number changes nothing. That last fact says the report needs to be different, not faster, and no export schedule would have surfaced it.
3 · What is the teach-back line, and why is it not just politeness?
On the bench, three of Adeline's 24 facts only surface once trust exists, and one of them, the only IT person leaving in January, decides whether anything built survives the quarter. The second sentence is how trust is built while you are still asking.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map.
Session 1 cheat sheet · pin this
Door 4, data state, is the one you know how to open and the one she cannot answer in your words. Session 2 is a five-rung ladder of plain questions that gets the real state of Harbourline's data out of a person who has never seen a schema: which numbers are typed by a person on a Friday, which come out of a till, who owns the product codes, who holds the only login. You draw the data-state map from her answers, and it is the first thing you send her.