Define what low means.
Name the source metric and comparison. Is the rate lower than the app’s own comparable period, a similar segment or an external benchmark? Each reference has limits, and none should be left implicit.
Use benchmark guidance and rate definitions. A percentage detached from its denominator cannot tell the team what to fix.
Inspect mix before creative.
Check whether source, market or audience composition changed. A different mix can shift the overall rate while individual segment rates remain stable. Record what the reports can separate and what remains unknown.
The audience-mix example demonstrates this arithmetic. Do not label an aggregate decline a screenshot failure before investigating the comparison.
Look for a specific misunderstanding.
Review the actual listing with the intended user task in mind. Does it communicate the outcome, show the mechanism and address a meaningful concern? A broad statement that the design looks old is less useful than a clear mismatch between promise and proof.
Use storytelling and trust evidence. Correct misleading claims directly; do not make truthfulness a test variant.
Prioritise an interpretable change.
Choose the strongest supported question and define the proposed treatment. Keep a record of what remains stable. If the change bundles copy, order and visual style, acknowledge that the evaluation concerns the combined treatment.
Use test prioritization. Ensure the platform workflow and available traffic can support useful evaluation before producing a large set of variants.
Review outcomes without forcing a win.
Compare the planned evidence and record concurrent changes. An inconclusive result may require more information or a different hypothesis. Do not declare success from a short-lived movement or keep testing until a favourable slice appears.
Bring the decision into the report. The next action can be to keep, revise or defer, provided the reasoning is visible.
