Define the conversion you are discussing.
Write down the report, numerator, denominator, store, country and date range. Different reports can describe different events and populations. Two percentages that look similar may answer different questions. Preserve those definitions when comparing periods.
The conversion service page includes a scenario calculator. It demonstrates arithmetic; it does not establish that two real reports are comparable.
Check whether the audience changed.
Campaigns, market mix and discovery sources can change the visitors reaching a listing. A lower aggregate rate may coincide with a new audience even if the listing itself did not change. Inspect relevant segments where the available data supports them.
The illustrative chart shows why an overall percentage can move without a within-group change. It is a fictional example designed to explain interpretation, not a claim about a client’s performance.
Inspect the promise and the product evidence.
Ask what the visitor expects on arrival and what the listing demonstrates. A promise of a quick first result should be supported by a clear starting experience. A dense settings screen may show real functionality while failing to answer that first question.
Use value proposition and opening screenshots to identify the mismatch before choosing a visual treatment.
Separate accuracy fixes from experiments.
Correct misleading or outdated information through the appropriate publishing workflow. When two accurate messages could plausibly work, write a hypothesis and confirm a suitable evaluation method. Do not describe every copy revision as an isolated experiment.
Use the hypothesis builder and test prioritisation to make the question and next step explicit.
Evaluate the result in context.
Record the published version, concurrent changes and relevant observation conditions. A stronger install rate is not the entire business outcome if it attracts people who misunderstand the product. Consider downstream quality where reliable evidence is available.
A result can support a rollout, reject an idea or leave uncertainty. Preserve that distinction. If the underlying cause remains unclear, start with an ASO audit instead of declaring the latest design a universal winner.
