Investigate product changes
A metric moves. Several explanations sound plausible. Bring the evidence into a focused investigation and see which ideas hold up, which remain open, and what to check next.
Start with a precise observation
Confirm the metric definition, comparison window, and source before looking for an explanation. Then frame a question the available evidence can help answer.
Illustrative investigation using a supplied metric. Check additional numeric breakdowns in your analytics tool, then bring that context into the question.
[1] Sample checkout funnel · Weekly comparison
64% → 52%
320/500 sessions → 260/500 sessions
Same funnel definition in both windows. A decline of 12 percentage points.
A focused investigation question
Examine competing explanations
Investigations test possible explanations against project evidence. A supported observation can still leave the wider cause unresolved.
01 / Hypothesis
Supported as a customer difficulty
One customer describes reaching payment before seeing the total [2]. This supports a price-visibility problem for that customer, not a cause for the overall decline.
What would help test it?
Check whether other customers describe the same difficulty and inspect the flow.
02 / Hypothesis
Inconclusive
An error issue reports eight affected users [3]. These records do not establish which sessions abandoned checkout or how those customers experienced the error.
What would help test it?
Ask engineering to verify the failure mode and customer outcome.
03 / Hypothesis
Inconclusive
The supplied evidence has no segment breakdown. A plausible explanation stays open until there is evidence to test it.
What would help test it?
Inspect comparable segments and acquisition sources in the analytics tool.
[2] Illustrative Intercom excerpt: “I got to the payment step before I could see the full price.” One customer account, not evidence that all abandoned sessions had the same experience.
[3] Illustrative Sentry issue: PaymentTimeout in the checkout service, eight users affected. The error and support records have not been linked to the same customers or sessions.
Illustrative investigation answer
Customer feedback supports a difficulty seeing the total [2]. Payment errors are another issue to assess [3]. The available evidence does not establish why overall completion declined [1].
Confidence in a causal explanation: low
The evidence identifies questions to pursue. It does not establish the reach of the reported difficulty or connect it to the measured decline.
Keep the limits with the answer
Read the conclusion with its confidence, citations, and unresolved hypotheses. Ask to save the investigation when you want to return to it or develop a brief.
Your next step might be another evidence check, a technical assessment, or a customer conversation. Choose it based on what the decision still needs.
Put the workflow to use
Start with the analytics, feedback, and error signals available for your question.
Explore ConnectReview the context and send a focused question. Examine the hypotheses and evidence in the answer.
Explore SynthesizeBring the supported finding and open questions into a brief for your team.
Explore ActConnect the relevant evidence and bring your first question.
Turn feedback into a delivery brief