Where we start
This course runs on one real situation. Daybreak, a coffee subscription company, has a churn problem in one roast origin. The analysis is finished and correct. The deck about it is neither, and the gap between those two facts is the whole subject.
★ What you walk out with tonight. One written sentence for a deck you actually owe somebody, tested against four checks, plus the two-line version you say out loud when the meeting starts twelve minutes late.
This course is not about making charts. It assumes the charts exist and are honest. If they are not, that is learn-dataviz-with-phoebe, and it should come first. Here we take a pile of correct visuals and turn it into something that changes a decision.
A deck is a topic until somebody writes the sentence
7 min liveAsk most people what their deck is about and you get a subject area. Ask what it says and there is a pause. That pause is the entire problem, and it is almost never a communication problem: it is an unfinished thought wearing a deck.
LiveWhy the topic version survives so well4 min▶
Nobody defends a topic deck on purpose. It survives because it is safe in four specific ways, and being able to name them is most of what stops you writing one.
- A topic cannot be wrong. "Subscription churn" makes no assertion, so it cannot be contradicted in the room. A claim can be, and that fear is the real author of most vague decks.
- A topic does not commit you. If you say pause Guji auto-renew and it turns out badly, that was your call. If you present variation across cohorts, it was a group decision.
- A topic fits the calendar invite. Meetings get booked before the analysis finishes, so the title is written before the finding exists, and then nobody goes back and changes it.
- A topic lets you include everything. Once there is a claim, two thirds of your charts stop earning their place, and you have to throw away work you did. This is the one that actually hurts.
That last one is worth sitting with. The sentence is not hard to write because writing is hard. It is hard because writing it makes visible how much of the deck does not serve it.
What a usable sentence has in it
7 min liveA sentence that works has four parts, and dropping any one of them produces a specific, recognisable failure. This is not a style guide. Each part is load-bearing.
That seeded line is thirty words and contains none of the four parts. It has a subject area, a verb that means nothing ("covers"), a hedge ("a number of"), and an adjective doing the work a number should do ("interesting"). It is also, word for word, the kind of sentence that gets written when the analysis is finished and the author has not yet decided what they think.
Self-studyFour tests, and what failing each one means5 min▶
- The disagreement test. Could a reasonable, informed colleague disagree with this sentence? If not, it is a fact statement, not a claim, and you are reporting rather than recommending. That is sometimes correct - just know which one you are doing.
- The number test. Is there a figure in it? Without one, the room supplies its own sense of scale, and it will be wrong in whichever direction suits them.
- The Monday test. If the sentence is accepted, what is different on Monday? "We understand churn better" is not different. "Guji auto-renew is paused" is.
- The corridor test. Can you say it, from memory, in the eleven seconds you get walking to the lift? If it needs the deck to make sense, it is not the sentence yet.
The corridor test is the harshest and the most useful. A sentence that survives it will also survive the meeting being cut to fifteen minutes, the executive arriving late, and the deck being forwarded to somebody who was not there. All three of those will happen.
Self-studyWhen you genuinely cannot write it4 min▶
Sometimes the sentence will not come, and the honest reason matters because each one has a different fix.
- The analysis is not finished. The most common case, and the one people are least willing to say. The fix is not writing; it is going back to the data. A deck cannot conclude something the analysis did not.
- There are genuinely two findings. Then there are two decks, or one deck and one email. Two claims in one deck means the second one gets four minutes at the end and is never acted on.
- The finding is real but you are not allowed to say the consequence. Somebody else owns that call. Write the sentence with the consequence as a question: "...so should we pause Guji auto-renew until supply recovers?" A question is still a claim about what the decision is.
- It is a genuine status update. Fine. Say so at the top, drop the pretence of a narrative, and use a table. Status updates are a legitimate genre and they are badly served by being dressed as stories.
Write the sentence for a deck you owe somebody
★ 22 min · everyone writesNot a hypothetical deck. One that exists, or one you have been avoiding. The rule for the next twenty minutes: no slides, no charts, no tool other than a text box.
Bad version. Write the sentence badly and fast. Hedges allowed, length allowed. You are getting the shape out of your head, not producing the artefact.
Add the number. Find the single figure that carries the finding. Not three figures. If you cannot find one, that is a finding of its own about the analysis.
Add the consequence. What is different on Monday if this is accepted. If the answer is "we know more", you have a status update, and it should stop pretending otherwise.
Cut to one line. Delete every hedge, every "a number of", every adjective standing in for a number. Twenty-five words is generous. Fifteen is better.
Say it out loud to the person next to you, from memory, without the screen. The bits you forget were never load-bearing. Cut them.
Open a text box, not a deck. The moment a slide is on screen you start designing, and designing is a way of not deciding.
Write it in the form "X does Y, by Z, so we should W". The template is scaffolding. You will drop it once the sentence exists, but it stops you writing a topic by accident.
Run the disagreement test. If nobody could reasonably push back, you are reporting. Check that is what you meant.
Run the Monday test and write the answer down next to the sentence. Two lines total.
Say it to somebody, from memory. Watch where they frown. That is where the sentence is unclear, not where they are being difficult.
Now count your existing slides and mark each one as serves the sentence, sets up the sentence, or neither. The "neither" pile is the point of tonight, and it will be bigger than you expect.
If you take one thing from tonight. The sentence is not a summary you write at the end. It is the constraint you write first, and it is the only thing that gives you permission to cut a chart you spent a day on.
Before session 2
25 min- Write the sentence for three real decks, two of yours and one somebody sent you. The third is the easiest and the most revealing.
- Take a deck you received this month and try to write its sentence from the deck alone. Note how far in you had to read to find it.
- Mark every slide in one deck as serves, sets up, or neither. Bring the counts to session 2.
- Say one of your sentences out loud to a colleague who was not involved, and write down their first question. That question is usually the structure problem session 2 is about.
Three questions
Clear all three and session 1 gets stamped in your passport.
1 · "We reviewed churn across all cohorts and found meaningful variation." Which of the four parts is this missing?
It has a subject and a verb, which is why it feels like a sentence. But "meaningful variation" asserts nothing checkable, carries no figure, and implies no action. Three of the four parts are missing, and the charts cannot supply them because a reader who does not know what is being claimed cannot tell which chart is the important one.
2 · You have a real finding but the decision belongs to somebody else. What do you do with part four?
Dropping the consequence leaves the room to invent its own, which is worse than proposing one. Turning it into a question keeps the decision with its owner while still telling them what decision is on the table. Option D is the reflex, and it is how a finding becomes a two-hour discussion with no outcome.
3 · Why is the corridor test - saying the sentence from memory in about eleven seconds - the most useful of the four?
It is not about attention spans, and brevity is only a proxy. The test earns its place because it simulates the three most common ways a deck actually gets consumed, none of which involve somebody sitting through all of it in order. D inverts the point: the number is what makes it memorable.
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
You have a sentence. Session 2 is where it goes, and the answer is almost always first. Two structures that work, one that does not, why the detective story fails in a business meeting, and what to do with the evidence once the conclusion is no longer a reveal.