Choose a question you can answer

“Is marketing working?” combines several questions. Are campaigns attracting relevant visitors? Do those visitors activate? Do they become retained customers? Is the additional contribution worth the spend? Choose one decision and identify the evidence needed for it. A channel’s attributed revenue does not automatically measure its causal effect.

Use one coherent funnel

In a fictional seven-day cohort, 1,000 new visitors include 200 sign-ups, 80 activated users and 20 paying users. Visitor-to-sign-up conversion is 20%; sign-up-to-activation is 40%; activation-to-payment is 25%; overall visitor-to-payment conversion is 2%. These rates are coherent only if the stages refer to the same cohort and ordered progression.

Do not divide this week’s payments by this week’s sign-ups if many payments come from older users. That ratio may be operationally useful, but it is not cohort conversion. Specify the observation window and whether users had enough time to reach the next step.

Define activation in terms of useful behaviour

A button click is easy to track but may not represent value. For an analytics learning product, completing an exercise and checking the explanation could be a candidate activation event. This is a hypothesis to validate against subsequent behaviour, not a universal definition. Track event quality and duplicate events before treating a dashboard as evidence.

Keep marketing economics separate from attribution

A campaign with $2,000 spend and $6,000 attributed revenue has a ROAS of 3. If contribution margin before advertising is 40%, the attributed contribution after advertising is $400: $6,000 × 40% − $2,000. This still does not show incremental profit because some purchases may have happened without the campaign.

Build a defensible recommendation

1. State the decision and eligible population.
2. Define the primary metric and guardrails.
3. Check event completeness, duplicates and cohort maturity.
4. Segment only when there is a reason and sufficient data.
5. Separate observed patterns from causal conclusions.
6. Recommend a measurable next step with a stopping condition.

Practice prompt

Activation increases while paid conversion falls. List three explanations: lower-quality acquisition, a tracking change, or a product change that helps early usage but hurts checkout. For each, name a specific comparison that would distinguish it. Avoid choosing a story before checking the data.