First, read conversion by source and platform
Use each store console's acquisition reports, which the app store optimisation guide lists as the place to see which sources produced listing visitors and installs. Compute the conversion rate (CVRShare of visits or clicks that convert.) as first-time installs over product page views for each source separately, using the definitions in the conversion rate calculator walkthrough for app store listings, and keep the App Store and Google Play apart, because their reports, indexed fields and rating rules differ. The Google Play Console Help and Apple's developer documentation describe the current reports; the breakdowns available change, so check them before building a series.
Then lay the per-source series side by side with three dates: each app release, each listing change and each campaign launch or pause. Four patterns cover most cases. Every source falls at once: suspect the page itself, through the screenshots or the rating. Only browse and campaign traffic fall while search holds: suspect the first screenshots, which those visitors judge without a typed need. Only paid traffic falls: suspect where the ads land. No source falls but the blended rate does: the mix moved, and nothing on the page needs fixing. If conversion holds but fewer people reach the listing in the first place, the diagnosis of organic app installs falling starts from store visits instead.
What happens after the install is out of scope. If installs hold but new users never reach the app's useful moment, the diagnosis of installs that do not become active users applies; if the concern is what each install costs, start from the store's share of that cost in the app marketing troubleshooting section.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| The blended rate fell, but each source's own rate is flat | Share of page views by source this period against the last | A traffic mix shifted towards sources that convert lower, read as one blended rate | Report conversion per source and platform, and stop deciding on the blend |
| Browse and campaign visitors convert worse after new screenshots; search holds | First screenshots and opening text against what each source was promised | Screenshots and first lines that no longer show the promise people came for | Restore an outcome-led opening sequence and test it in the store's experiment |
| Every source falls together soon after a release | Rating and review count by version and country, with tagged complaints | A rating that fell, or recent reviews that turned negative after a release | Fix the complaint, reply to the reviews, and track the recovery by version |
| Paid visitors convert far below paid visitors to an older campaign page | Which listing each active campaign points at, and its first screenshots against the ad | Paid campaigns sending visitors to the default listing instead of a matching custom one | Point each campaign theme at a custom listing whose opening matches its ad |
| The fall begins on the date a listing update went live | Whether the change ran as a store experiment, and its result | A listing change shipped without a store experiment or a measured result | Roll back or test the change against the previous set in the store's own experiment |
Fix, verify and prevent each cause
A different crowd, not a worse page
Store search visitors arrive with a typed need; browse visitors are curious; campaign visitors saw one promise; a feature or a press mention brings a burst of people who were never looking. The ASO guide explains that search and browse work differently, and the custom store listings tutorial warns that paid traffic must be compared with paid traffic, never with search.
Fix: replace the blended store optimisationImproving an app's visibility in app store search and the share of store visitors who install it. headline with per-source, per-platform rates, and annotate the series with every campaign start, feature and release.
Verify: for the period in question, each source's own rate is stable, and the change in the blended figure is explained by the shift in page-view shares.
Prevent: never report a single store conversion figure to leadership; always report the sources that make it up, so a bigger campaign is not mistaken for a broken page.
Screenshots that stopped selling the job
The ASO guide treats the first screenshots as the conversion asset, seen by far more people than the description, and recommends a caption-led sequence whose first frame answers what the app is and why someone would want it. Redesigns drift from that: a new feature takes the first frame, the captions become brand lines, or the interface appears with no statement of benefit.
Fix: rebuild the opening frames around the outcome the most valuable sources came for, keep the text within the App Store Review Guidelines and the Play policies the ASO guide cites, and run the new set against the current one.
Verify: in the store's listing experiment, conversion for the priority source improves, and installer retention in the cohort reports holds.
Prevent: run the app store listing checklist before every listing update, so the promise in the first frames is checked against the traffic that will see it.
A rating that turned after a release
The rating sits beside the app in search results, so it works on conversion before anyone reads a word. A release that introduces a crash, removes a loved feature or changes pricing can turn recent reviews quickly, and the ASO guide notes that Google weights Play ratings towards recent reviews and shows them per country. The tutorial on app rating prompts and review replies tracks the rating by version and country, tags every review, routes bugs to product with the version, and treats a summary reset as a decision to earn.
Fix: ship the fix, reply publicly to the affected reviews with the version that resolves them, and keep the rating prompt wired to a moment of completed value.
Verify: the new version's rating and tagged complaints recover in the weekly table, and conversion across sources climbs back as they do.
Prevent: read reviews of each release in its first days, by version, and make a souring rating a reason to pause paid campaigns that land on the listing until the fix ships.
Ads that land on a page written for everyone
The default listing leads with the app's broadest promise. An ad built around one feature or audience that sends people there breaks the continuity the visitor expected, and the install campaigns guide warns that such campaigns convert below organic. The custom listings tutorial builds variant pages per campaign theme, links each campaign to its page, and keeps a register of which campaigns point where.
Fix: for each theme that genuinely differs, build a custom listing whose first screenshots repeat the ad's promise, and point that theme's campaigns at it.
Verify: paid conversion on the custom page rises above paid conversion on the default for the same kind of traffic, and, because more of the same clicks install, each paid install costs less, which the CPIAverage ad cost of each app install. for those campaigns should show.
Prevent: check the register whenever a campaign launches or a page is retired, so no ad points at a page that no longer matches it.
A redesign judged on taste
The ASO guide calls the stores' own listing tests the only reliable way to know what moves store conversion, because the counterfactual cannot otherwise be observed. A new icon, a new preview or a new screenshot set shipped straight to everyone leaves no way to separate its effect from the season, the campaigns and the rating. The app store screenshot test sets out the design: one asset set changed, the store visitor as the unit, a priority source named in advance, and retention, ratings and campaign costs as guardrails.
Fix: if the fall began with an untested change, restore the previous set or test it against the new one in the store's experiment.
Verify: the experiment's result for the priority source decides which set stays, with the guardrails inside their limits.
Prevent: make the store experiment the only route by which the icon, screenshots or preview change, and record each result where the next designer will read it.
A prevention routine for the store page
Treat the listing as the landing page of every install the app receives, because that is what it is. Each week, read conversion per source and per platform from both consoles, beside the rating and review count by version and country, and annotate the series with releases, listing changes and campaign launches. Hold one owner responsible for the listing as a whole, with the review routine, the campaign register and the experiment calendar under the same person, so that a design change, a release and a new campaign cannot each be blamed on another team.
Before any listing update, run the listing checklist and route the change through a store experiment. Before any release, agree who reads the first days of reviews and what would pause campaigns. Before any campaign, confirm the page it will land on and record it in the register. Once a quarter, retire custom listings whose campaigns have ended and repeat the per-source read for each localised storefront, since a market that drifts on its own is easy to miss in a global figure. The app marketing hub explains why store and campaign work have to be planned together.
Pemblequist Star Maps after its redesign
Pemblequist Star Maps, an invented stargazing app, sees store conversion on the App Store fall from 34% to 25% in the four weeks after a redesign that changed its screenshots and shipped a new version. By source, store search holds at 41%, while browse falls from 22% to 13% and paid traffic from 29% to 18%. The new first screenshot leads with a premium telescope-control feature; the old one showed the night sky identified through the phone's camera, which is what the campaign ads also show. The release also introduced a crash on older phones, and the rating for the new version stands at 3.1 against 4.5 before. Pemblequist ships a crash fix, replies to 140 reviews naming the fixed version, and runs the old opening screenshot against the new one in the store's experiment with browse as the priority source. The sky-identification set wins on browse at 23% against 14%, with day-seven retention level. Paid campaigns move to a custom listing built on the same frames. Six weeks later blended conversion is 33%, and the new version's rating is 4.3.
Common mistakes
Redesigning the page when only the traffic mix moved. Comparing paid visitors with store search visitors. Resetting the rating after a release that fixed nothing. Leaving a campaign pointed at a custom page that no longer matches its ad. Shipping new screenshots to everyone and calling the next month's figure the result.
