One bet. Stated so the market can settle it.
One growth experiment: a bet the market can settle, stated so it can be disproved. Every Signal Sprint runs on this template. The required half fits in one sentence; the optional half is every question a past run wished it had asked first.
Five fields that compose into one sentence the market can settle. An experiment cannot be approved with a blank among them.
- titleTitle
- hypothesisHypothesis
State it so a result could prove it wrong.
- leverGrowth leveracquisitionactivationretentionrevenuereferral
- verbDirectionincreasedecreasemaintain
- metricMetric
- current_valueCurrent value
The baseline. A number without one is not evidence.
- goal_valueGoal value
- target_dateTarget datedefaults to +30 days
The questions a past run wished it had asked first.
Optional depth — except the kill condition, which approval requires. Each field here was the expensive omission on a real run. Collapsed by default, shown in full once filled.
- problemThe problem
What hurts them, in their words. Not what we sell.
- who_has_itWho has the problem
Title, seniority, and whether they can spend.
- who_we_reachWho we can reach
Often not the same person.
- characteristicsCharacteristics
Push every trait to something a stranger could look up. If it cannot be stated numerically, say so plainly.
CharacteristicStated asObservable?Numeric threshold - calibration_setCalibration set nudged if empty
Real companies you already know the answer for — at least one that should be a yes and one a no. If your targeting would reject a happy client, the targeting is wrong, not the client.
CompanyExpected (yes / no)How we know - offerThe offer
What you’re putting in front of them, what you’re asking them to do, and who it’s written for — the same offer usually needs rewording for a different role. Test it against the value equation: how big the outcome, how likely they believe it, how fast it arrives, how little effort it takes.
- kill_conditionKill condition required to approve
What would make us stop. Agree it while it is still cheap. Required before an experiment can enter the queue.
- open_questionsOpen questions
What nobody answered. On paper beats a silent assumption.
one line per entry - quotesVerbatim quotes
Precise phrasing carries what paraphrase loses.
one line per entry
Proposed, approved, locked, concluded.
Anyone on the account can propose. The operator approves into an ordered queue and locks one in as the active experiment. One at a time. A miss is a result.
In the backlog. Anyone on the account can propose.
In the queue, in order. The bet must be complete and a kill condition agreed.
Locked. One per account at a time. This is the sprint.
Outcome recorded: what actually happened, and what we learned.
Same record. A miss is a result.
Stopped on purpose, with the reason.
Never run. Stays visible so it is not proposed twice.
- Approval requires every bet field and a kill condition. No blanks.
- Exactly one active experiment per account. Activating a second is refused.
- Activating locks the bet. After that, edits are operator-only and logged as a typo fix with before and after.
- An outcome can only be recorded on the active experiment.
- Dropping requires a reason, and the row stays in Past.
- Every outcome asks for one line of what we learned. Nudged, not blocked.
Run one to a conclusion.
One experiment, locked in, read honestly. The next one is the correct response to the result, whichever way it went.
Book a Signal Sprint