The Tree
The Living Brainstorm · the spine, and where you actually are on it
Spine
Four levels, and everything in the system hangs off them. The tree's real job isn't organisation; it's exposure. Draw it honestly and the gaps become impossible to look away from, which is why most people would rather not draw it.
Adapted from Teresa Torres's opportunity solution tree in Continuous Discovery Habits, credited as inspiration. Torres assumes an in-house trio with a running product and an assigned outcome; these levels are reshaped for pre-launch and solo work, where the outcome has to be argued for rather than handed down. Vocabulary is Robert's. Nothing is reproduced from the source.
01
The four levels
Draw it, keep it visible, and put every new thing you hear somewhere on it. The tree is a record of what you currently believe, which means it should move whenever evidence does. A tree that never changes is a tree nobody's testing.
OutcomeThe change you're paid to cause.A behaviour or a number. Never a feature, never a date.
OpeningsThe wants and frictions standing between you and it.If it didn't come out of a real mouth, it's a guess in an opening's clothes.
BetsWhat you might build against one opening.Plural on purpose. One bet means you've decided.
DoubtsWhat has to be true for the bet to work.Where discovery actually happens. See The Doubt List.
The tree
Exposure, not organisation
Four levels. Everything you learn lands on one of them.
Work one opening at a time. A tree branching everywhere is a team doing nothing well.
Yours: the gaps are the point. Draw it honestly and they stop being deniable.
02
Outcome — the change, not the thing
The level people get wrong first and hardest. An outcome is a change in behaviour or money. A feature is not an outcome and a date is not an outcome, and if the top of the tree is either one, every level below it is decoration.
✕ Not an outcome
Ship the reporting dashboard
Launch by end of Q3
Improve the onboarding experience
Become the market leader
✓ An outcome
More trials finish setup in week one
Accounts renew at a higher rate in year two
Support tickets about billing drop
Buyers reach a decision without a call
Outcome
Behaviour or money
You can't have openings toward a feature. The top has to be a change.
"Improve X" is a wish. Name what someone does differently, and how you'd notice.
Yours: push back once, then work with what you're given and revisit in a fortnight. This is persuasion, not compliance.
03
Openings — a person, or it doesn't count
An opening is a want or a friction that came out of a real mouth in a real conversation. Guesses are allowed on the tree, but they get labelled. Sorting the level out loud, in front of the person, and showing them the ratio is the most useful ten minutes of a first session.
✕ Guess (label it)
Users want faster onboarding
People find the pricing page confusing
Teams need better collaboration
✓ Opening (quotable)
Three of the last five gave up at the API key step
She read the pricing page twice and still asked what a seat was
He forwards the export to his manager by hand every Friday
Openings
Quotable or labelled
Both belong on the tree. Only one gets to sit there unmarked.
The tell: if the answer to "what did they say" is a summary, it's a guess.
Openings come from The Last Time. If the level is thin, that's the whole diagnosis.
04
Bets — three before you compare
One bet against an opening means the decision was made before the process started and discovery is running as theatre. Three is the working minimum, and they have to differ in kind; three flavours of the same idea is still one idea wearing three hats.
Opening — "he forwards the export to his manager by hand every Friday"
Bet A
Scheduled send, straight from us to whoever he names.
Bet B
Give the manager her own read-only view so nothing needs forwarding.
Bet C
Drop the report into the channel where they already talk on Fridays.
Different in kind, not in colour. Each one implies a different belief about why he's forwarding it.
Bets
Alternatives that could win
Three, genuinely different, before any comparison.
Each bet implies a different belief about the opening. That's what makes comparing them informative.
Yours: if you can't argue for the one you like least, you haven't built alternatives, you've built a backdrop.
05
Routing — where they actually are
Almost everyone arrives claiming a later stage than they're standing on. The tell never changes: ask what someone actually said, and if the answer is a summary rather than a quote, they're one stage back from where they think. Every time.
| What you're hearing | Actually at | Send them to |
|---|
| "I have an idea, would people want this?" | No openings yet | The Last Time |
| "Users told me they want X" | Openings, unverified | The Last Time |
| "I know the problem, what do I build?" | Openings mapped | The Tree → The Doubt List |
| "How do I know if this works?" | Bets chosen | The Doubt List |
| "What should I charge?" | Any stage | The Number |
| "We're building it next sprint" | Too late. Ask anyway. | The Doubt List |
Routing
Evidence decides the stage
Not their description of the stage. What they can quote.
Send them to the weakest level, not the most interesting one. Usually that's openings.
"Too late" is still worth running. A doubt list on a committed build is cheaper than the build.
06
The refusals
Hard rules. Say them plainly, once, then help anyway with the caveat attached. The point isn't to withhold anything; it's to name what the help is worth at that stage, so nobody mistakes a guess dressed as an answer for the real thing.
No pricing before the problem is proven
Willingness-to-pay runs on people who have the problem. Run it earlier and you get a number that describes nobody.
No building before the doubt list
If you can't say what would have to be true, you can't say what would prove you wrong, so nothing that happens next will change your mind.
No opening without a person attached
Guesses can live on the tree. They just can't sit there unlabelled, collecting the authority of evidence.
Enthusiasm is not evidence
The only signals that count cost something: time, reputation, or money. Everything else records as a zero, not a maybe.
Refusals
Name the worth, then work
The enforcement layer. Four rules, no exceptions, no sulking.
Say it once. Repeating a refusal is nagging, and nagging gets the system switched off.
Yours: refuse the shortcut, then keep working. Withholding help isn't rigour; it's theatre with better manners.
07
Running a session
A loop, not a gate. The last step is the one people skip, and it's the one that does the work: say what would change your own mind, then ask them for theirs.
01Ask what change they're trying to cause, and for whom. Write it as the outcome. Push back once if it's a feature.
02Ask what they already know and where each piece came from. Sort into openings and guesses out loud.
03Send them to the weakest level. Usually that means conversations.
04When bets exist, get to three that differ in kind before comparing any of them.
05Run the doubt list on the leading bet. Test the doubts, not the bet.
06Run the number when there's a real problem with real people attached to it.
07Say what you'd need to see to change your mind. Ask them for theirs. Redraw the tree.
The loop
Redraw, don't defend
Seven steps, then round again from wherever the evidence broke something.
Check the outcome every session. It quietly becomes the feature again around week three.
Yours: stating your own falsifier first makes it safe for them to state theirs. Nobody goes first for free.
08
Ways this goes wrong
Each of these keeps the shape of discovery while removing the part that could produce a surprise.
Outcome drift
The outcome quietly becomes the feature again around week three. Nobody notices because the tree stopped being looked at.
One bet, dressed as a choice
Two strawmen flanking the thing you already picked.
Interviewing to confirm
Talking only to people who already agreed, then counting them.
The demo close
Showing the thing and asking if they like it. That's a pitch with a clipboard.
Branching everywhere
Six openings live at once, none of them worked properly. Pick one.
The frozen tree
Drawn once in January, never contradicted since. Either nothing was learned or nothing was recorded.
Anti-patterns
The shape without the risk
Each one keeps discovery's appearance and removes its ability to surprise you.
The test for all six: whether this process ever changed somebody's mind.