Template pack

§The body kept the mechanics out. This is the mechanics, kept light, because the constraint at one to five people is not rigor but maintainability: a system heavier than the team can carry is dropped in the first busy month.

§ A.11. Hypothesis card#

§One card per experiment, filled before it runs. Hypothesis: if we do X, then Y will happen, measured this way, because Z. Riskiest-assumption link: which assumption on the list this tests, and its consequence and uncertainty. Metric: the one number. Threshold: the value that means yes, with its reason from the business. Decision rule: what we do above, below, and in between. Guardrails: the metrics we watch to make sure nothing else breaks. Sample and duration: the sample the effect requires, computed, and the fixed stopping point, or a note that this is a qualitative read because the sample is too small.

§ A.22. Experiment backlog#

§One row per candidate hypothesis. Hypothesis in short form. Funnel stage it targets. Impact, confidence, ease, each scored. Score, and rank. Re-sorted after every result. Filtered to the funnel's weakest stage when choosing what to run next. Ideas enter as hypotheses, never as tasks: "dark mode raises night retention because users said so," not "add dark mode."

§ A.33. Decision-rule sheet#

§Kept with the hypothesis card, separated here because it is the part most often skipped. Above threshold: the action, named. Below threshold: the action, named. Ambiguous middle: the action, named, usually run longer or redesign. Signed off by the team before the experiment starts, so the result executes a decision rather than opening a debate.

§ A.44. Cohort and retention table#

§The honest view of after-launch behavior. Rows are cohorts, by start week. Columns are the periods after start, matched to the product's usage frequency. Cells are the share still active, or activated, or paying. Read down a column to compare cohorts; read across a row to see a cohort's curve. Tagged by acquisition channel where channel quality is in question. This one table underlies activation, retention, and most growth reading.

§ A.55. Weekly review format#

§The cadence, on one page. Launch: the experiments starting this week, from the backlog. Read: the experiments that finished, each against its pre-written threshold, decision executed per the rule. Learn: for each finished experiment, the insight, a durable statement about the users, added to the record. Decide next: the backlog re-sorted, next week's experiments chosen. Short, because the decisions were pre-committed; an execution meeting, not an opinion session.

§ A.66. Experiment record#

§The compounding asset. One entry per experiment, searchable by funnel stage and by assumption. Hypothesis, metric, threshold, result, decision, insight. Findable a year later, so the team does not re-run what it has run and can tell a new hire what it knows. This is innovation accounting at one-to-five scale: the memory that turns a series of experiments into a growing understanding of the market.