Check whether the measures belong together.
An impression count and an install count may use different units or cover different populations. Identify the source definitions and filters before describing the gap as a conversion problem.
Use impressions versus page views and rate definitions. Do not assume every counted impression represents a person who then opened the product page.
Ask where the exposure came from.
A rise in impressions can reflect a different surface or audience. Determine which source and market segments changed, using only the dimensions actually available. High volume is not the same as high relevance.
Review search and browse context. If new exposure is broad, the aggregate response may differ even without a change to the listing itself.
Inspect the promise at the first contact.
Review the visible name, icon and message in the relevant context. Can a suitable user understand what the app does and why it fits? An unclear promise is a useful hypothesis to investigate, not a proven explanation for all missing installs.
Use value proposition guidance to clarify the argument. Keep the claim grounded in the product rather than adding exaggerated language to compensate for uncertainty.
Follow the evidence into the listing.
If the available data and review suggest a page-level misunderstanding, inspect the first screenshots, supporting proof and readability. Look for a concrete mismatch between the arriving user’s expectation and what the page shows.
Use opening screenshots and conversion diagnosis. Separate accurate corrections from ideas that need testing.
Choose a focused next question.
Record whether the strongest evidence concerns measurement, audience, message or availability. Assign the next check accordingly. Avoid changing several unrelated elements and then attributing a later improvement to one of them.
Use a written hypothesis where a creative experiment is appropriate. The purpose is to learn why relevant people choose, not to maximise impressions in isolation.
