A decision brief and experiment card for PM work

Write an artifact that helps someone decide

A useful template makes missing reasoning visible. Fill these in with what you know, mark assumptions as assumptions, and keep the document short enough that the decision owner will read it. The worked examples are synthetic and may be copied or adapted with the blank templates.

Copy this decision brief

Decision to make:
Decision owner and review date:
User and task affected:
Observed problem (with source and date):
What we assume but have not verified:
Constraint (time, capacity or risk):
Options considered:
Recommended action and why:
What we will defer:
What would change this recommendation:
Next action, owner and due date:

Filled example: improve first sharing

Decision: whether to spend this week’s six engineering days on an invitation improvement or a reporting export. Owner: the product lead; review Monday.

Observation: in a synthetic cohort of 100 new workspaces, 40 created a note and 12 shared one. Five interviews suggest invitation confusion. We have not verified that confusion explains the drop-off or that sharing predicts retention.

Recommendation: spend one day observing the sharing task and validating events. If invitation friction is confirmed, scope the smallest fix within the remaining capacity. Defer the export until we verify the prospect’s required workflow and commitment.

Reversal condition: if the main issue is that creators have nobody to collaborate with, pause the invitation fix. Next action: the PM recruits five recent creators for observation; the engineer checks invitation event coverage. Both return findings to the product lead before scope is committed.

Copy this experiment card

Hypothesis:
Target cohort and eligibility:
Change or intervention:
Comparison and known confounders:
Primary outcome (numerator / denominator):
Observation window:
Guardrail and stop condition:
Minimum evidence needed for the next decision:
Instrumentation check:
Owner and review date:
Result: observed facts / interpretation / next action

Filled example: a clearer invitation step

Hypothesis: creators who want to collaborate are failing to invite teammates because the invitation action is unclear. A clearer step could help them reach first sharing.

Cohort: new team workspaces with a full seven-day observation window. Primary outcome: workspaces that share a note within seven days divided by eligible new workspaces. Guardrail: failed invitations and support contacts about unexpected invitations.

Comparison: collect the same events before and after, while recording source mix and rollout dates. A small before-and-after cohort is directional evidence, not a causal experiment. Stop if invitations are sent to unintended recipients or a serious access issue appears.

Decision rule: inspect task observations and cohort data together. Continue if users complete the intended task and no guardrail worsens; revise if they still get stuck; investigate value if they share once but do not return. Agree on the actual evidence threshold with the team before seeing results.

Review before sharing

These documents are starting points. Expand them only when the decision requires more detail. A completed template is not proof that a hypothesis is true.

• Can a reader identify the actual decision in the first sentence?
• Are observations distinguishable from assumptions?
• Does each metric have a denominator and a time window?
• Is the deferred option explicit?
• Can someone disagree using the stated reversal condition?
• Are the owner, next action and review date clear?
• Have you removed customer details, credentials and confidential material from a public or AI-assisted version?

Practise with a synthetic casePrepare for stakeholder disagreement