THE PLAYBOOKS / REVERSE-ENGINEERED OUTCOMES

Outcome experiment canvas

Connect the desired outcome to a customer behavior, the evidence, and a decision rule. Do not invent a baseline, expected lift, or certainty that your evidence does not support.

01What result are you trying to improve?

Define the customer benefit and relevant business outcome.

02Whose behavior matters in this test?

Specify eligibility and whether the unit is a person or account.

03What is observed, reported, hypothesized, or measured?

Keep each type of statement separate.

04What do you believe—and why?

Make the explanation capable of being wrong.

05What is the smallest useful change?

Include accessibility, error states, and constraints.

06What is the primary measure?

Document event, denominator, observation window, and data coverage.

07What is the comparison or research approach?

Explain how you will avoid attributing unrelated changes to this intervention.

08What harm or trade-off would make you stop?

Consider errors, complaints, workload, trust, and customer control.

09How will you decide to keep, change, or stop?

Set an appropriate review point; consult a specialist for statistical design.

10Who is accountable for the decision?

Include implementation and analysis owners.