Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Mobile & App Marketing › Troubleshooting

Organic App Installs Falling

Why organic app installs fall and how to fix it: lost store keywords, cut paid and brand demand, a spent launch, ratings, broken routes and counting.
Troubleshootingintermediate13 min readUpdated

First, split the fall into store visits, conversion and counting

Start from both store consoles and the measurement partner, read side by side. The app store optimisation guide sends you to the acquisition reports in each console, which break listing visitors and installs down by where they came from, and the reconciliation tutorial explains why the console, the analytics tool and the partner never agree: each counts a different event under different rules. That matters here, because "organic" is not one number. In the partner it is the installs left over once paid ones are matched to campaigns; in the consoles, installs are split into source categories, for instance search, browse and referrals, as the reconciliation tutorial describes, rather than reported as one organic figure. Write down which figure fell before asking why.

Then split it. If store page visits fell and conversion held, people stopped arriving, and the first five causes below are the candidates. If visits held and conversion fell, hand over to the conversion diagnosis, which reads the rate by source and platform. If the partner's organic figure fell while the consoles' totals held, suspect the count before the store. Read iOS and Android separately, the countries that matter separately, and the matching weeks a year earlier beside the period just before, so that Seasonality is compared rather than mistaken for a fault. Where weekly figures are small, the guide to reading small numbers advises pooling them and reading a range, and waiting for a move to repeat.

With the fall located, put a date on it and work backwards with the five whys: fewer search visits on iOS, because the app appears for fewer searches, because the subtitle changed, because the redesign release rewrote it, because nobody checked the indexed fields before submission. Each answer needs evidence behind it, and the chain ends at a process the team can change. The app marketing hub sets store presence, paid installs and launch side by side, which is why one fall often has more than one cause.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Store search visits fall on one platform from the date a release went liveThe name, subtitle, keyword field or Play descriptions before and after that releaseA release changed the app name, subtitle, keyword field or Play descriptions and dropped words the app was found forRestore the dropped words in the indexed fields and record every future change with its date
Organic installs and branded store searches fall soon after a campaign or awareness push stoppedCampaign and awareness dates against branded search visits and organic installs, market by marketPaid campaigns or awareness work elsewhere were cut, and the store searches and visits they were generating went with themTest what the paused work was adding by market or period, then fund what proves incremental
The fall is measured from a launch, a feature or press weekWhether the comparison period is a one-off peak, and installs from people who already knew the businessA launch, a store feature or press coverage set the baseline, and the borrowed attention has run outReset the baseline to the settled level and grow from repeatable sources
Visits fall gradually across sources while the rating slips or releases have stoppedThe rating by version and country, recent reviews and the date of the last releaseThe rating fell or updates stopped, and both stores count ratings and freshness among what decides where an app appearsFix what the reviews complain about, keep releasing, and run the reply routine
Referral visits or installs from tagged owned routes fall, or arrive without a sourceEach web, email and printed route: whether it still opens the matching screen and keeps its tagsRoutes from the website, emails and printed codes broke or changed, so the customers who arrived through them stopped or are now counted elsewhereRepair and retag each route, and test it in the contexts people use
The partner's organic count falls while console totals hold, from the date of a changeNetwork integrations, attribution settings and SDK releases near the start dateThe count moved rather than the installs, because a new network integration, attribution setting or SDK release moved installs between organic and paidCorrect or record the change, rerun the reconciliation and restate the baseline

Fix, verify and prevent each cause

The words the app was found for

The app store optimisation guide sets out what store search reads. Apple's documentation, on its App Store page for developers, names the app name, subtitle and keyword field as the text App Store search reads, while the description is there for human readers. The Google Play Console Help treats the title and both the short and the full description as text that shapes discovery, so on Play the long description counts for search in a way it does not on the App Store. A redesign that shortens the subtitle into a brand line, a rewrite of the Play description that drops the plain terms people search for, or a keyword field rebuilt from memory can each remove the app from searches it used to appear in. The listing still converts the people who arrive; fewer of them arrive.

Fix: compare the indexed fields before and after the release that coincides with the fall, restore the store search terms that describe the app's job, and keep brand terms out of fields they would waste. Messages that change often belong in promotional text, which the ASO guide notes can be edited with no new build.

Verify: search visits on the affected platform recover in the console's acquisition reports compared with the matching weeks a year earlier, and conversion from search holds.

Prevent: use the app store listing checklist on every listing change, so the indexed fields are read against the term list and the change is recorded with its date.

Demand that a paused campaign was creating

Some of the people an ad or an awareness push reaches find the app later through the store, by searching for its name. The attribution guide for apps warns that installs which would have happened anyway, and searches for the brand, are often handed to the last ad a person saw, and the measurement plan for app installs adds the other side: organic installs go up while paid campaigns are live, and no counter gives the campaign credit for that lift. Cut the campaign and the paid line falls as expected, and the organic line can fall too, with branded searches in the store falling with it.

Fix: do not restart everything on a hunch. Test what the paused work was adding, as the attribution guide describes, by comparing markets or periods with and without it; the brand search holdout test sets out a matched-region design that reads paid and organic together. Keep the work whose Incrementality holds up and drop the rest.

Verify: in the test, markets with the work running show more total installs, organic and paid together, than matched markets without it.

Prevent: before cutting any campaign or awareness budget, record the date and read organic installs and branded searches for the following weeks beside paid spend, so the knock-on effect is expected and measured.

A launch peak used as the yardstick

The guide to launching a new app is direct about the spike: its main value is the lesson it gives on how a large crowd converts and stays, and the people installing in launch week are, more than later, people who already knew the business. Store features are editorial decisions, made on quality, design and newsworthiness rather than on request, and press coverage fades. Each brings a crowd once. Measured against that peak, every month afterwards looks like a fall.

Fix: set the Baseline to the level installs settled at once the launch, feature or coverage had passed, label the peak in every report, and plan growth from the sources that repeat: store search, paid campaigns tested against payback and owned channels.

Verify: against the reset baseline, organic installs are flat or growing, and the remaining gap to the peak is explained by the dated event that created it.

Prevent: annotate every launch, feature and press mention when it happens, and never let a target be set from the weeks they covered.

A rating or update rhythm that slipped

The ASO guide records that both stores name an app's history of installs and engagement, its ratings and how recently it was updated among the things that affect where it is shown, without publishing how much each counts. It also notes that the rating is displayed next to the app in search results, so a slipping score works against both being shown and being chosen. A release that annoyed users, or months with no release at all, can therefore show up first as fewer store visits across search and browse. The ratings and review replies tutorial follows the rating for each version in each country, which is how a turn after one release is found, and notes that Play leans its rating towards the latest reviews and reports it country by country.

Fix: ship the fix for whatever the recent reviews complain about, reply to them with the version that resolves it, keep the rating prompt on a moment of completed value, and keep releasing on a rhythm users can see.

Verify: the weekly table shows the latest version's score climbing and its tagged complaints thinning, and store visits follow over the weeks after.

Prevent: read reviews by version in the first days of each release, and give each release release notes written as the guide to release notes and a changelog describes, leading with the change, so the dated history shows what each version did to the rating.

Owned routes that stopped sending people

Some of the people counted as organic came from the business's own channels: an app link on the website, a link in a shipping email, a printed code on a receipt or a counter. The guide to web-to-app journeys builds those routes to open the matching app screen, tags every one with campaign parameters so the route can be read, and tests them in the contexts people use. A website redesign that removed the app link, an email template that lost it, a printed code pointing at a retired address or a route that now opens the browser instead of the app can each stop the flow. A route that still works but lost its tags keeps sending people, who are then counted under a different source.

Fix: list every owned route with its destination and tags, repair the broken ones, restore the tags, and retest each through the store and the deferred link on a clean device.

Verify: tagged sessions and installs by route reappear in analytics and the partner, and referral visits in the consoles recover.

Prevent: add the route list to every website, email template and print change, so a redesign cannot remove a route without someone noticing.

A count that moved, not the installs

The reconciliation tutorial reports the partner's organic and paid installs as two lines for exactly this reason: an integration that breaks or is added can shift installs between the two while the total stays where it was. A newly linked network starts claiming installs the partner used to call organic; an attribution window is lengthened; an SDK release changes how first opens are recorded or handles consent differently on one platform. The partner's organic line falls, the paid line rises or holds, and the store consoles, which do not take part in attribution, barely move.

Fix: check SDK releases, integrations and Attribution settings first, as the tutorial advises, correct anything that is a fault, and rerun the monthly table for the affected months.

Verify: organic and paid together track the store totals within the tolerance set from the app's own history, and the date of the step matches the change.

Prevent: record every integration, window and SDK change in the measurement change log with its date and platform, and restate the baseline when a change is deliberate, so nobody chases the same step twice.

Prevention: a dated log of listing changes, the indexed fields checked on every release and organic installs read beside paid spend

Three habits prevent many of these falls or make them quick to explain. The first is a dated log. Every change to the name, subtitle, keyword field, descriptions, screenshots or promotional text, every campaign start and pause, every awareness push, every launch, feature and press mention, every integration and attribution change goes into one dated record, beside the release history. When organic installs move, the log turns the five whys from guesswork into a lookup: find the date, find what changed on it.

The second is a release gate for the indexed fields. A listing often changes as part of an app release and rides along with the build, which is how a shortened subtitle or a rewritten description reaches the store without anyone reading it for search terms. Make the listing checklist part of every release that touches the listing, with the term list beside it, and treat a change to the name, subtitle, keyword field or Play descriptions as a change that needs a reason and an owner. The app marketing operations playbook already ties a listing check to each release; this is where the indexed fields get their own line in it.

The third is reading organic installs beside paid spend, never alone. Each week, show organic installs by platform and important country next to paid installs, paid spend and branded store searches, against the matching weeks a year earlier. A report that shows organic installs on their own invites the wrong conclusion whenever a campaign starts or stops. Read each month's reconciliation in the same review, so a counting change is caught before it is mistaken for lost demand.

Set targets from the settled level, never a peak. When the cause is cost rather than volume, the diagnosis of a rising cost per install applies, and other app symptoms are routed from the app marketing troubleshooting collection.

A bird identification app finds why its organic installs fell (illustrative)

Illustrative example — not a real result

Wrenquilo Birds splits a quarter-long fall into three causes

Wrenquilo Birds, an invented bird identification app, sees organic installs in the measurement partner fall from 21,600 in the first quarter to 15,800 in the second, a drop of 27%, and the team assumes the store has changed something. Splitting the fall tells a different story.

Measure, second quarter against the same weeks a year earlieriOSAndroid
Product page visits from store searchDown 24%Down 2%
Conversion from search visits31% against 32%28% against 28%
Branded store search visitsDown 18%Down 15%
Partner organic installsDown 33%Down 19%
Store console first-time downloadsDown 17%Down 6%

Conversion is flat, so the store page is not the problem. Three changes explain the rest. The version 4.0 redesign, released on iOS only, shortened the subtitle from a line about identifying birds by song and photo to a brand line, and search visits on iOS fell from that week. A paid social campaign of £9,000 a month had been paused in early April; branded searches fell on both platforms soon after, and restarting it at £4,000 a month in 3 of 6 markets for 6 weeks lifts branded store searches by 14% and total installs by 9% against the other 3. And a new network linked to the partner in May began claiming about 1,200 installs a month that the partner had previously counted as organic.

The team restores the song-identification words to the subtitle, logs every listing change with its date, keeps the smaller campaign in all 6 markets, and records the network change in the measurement change log with a restated baseline. Over the following 4 months, iOS search visits return to within 3% of the same weeks a year earlier, and organic and paid installs together end the period 6% above them.

Common mistakes

  • Blaming the store's algorithm. The weights are not published, so the theory cannot be tested. Check the listing, campaigns, rating, routes and counting first.
  • Treating organic as one number. The partner's organic line and the console's sources measure different things. Say which figure fell before looking for a cause.
  • Redesigning the page when visits fell. A listing that still converts needs no new screenshots.
  • Cutting a campaign and reading only the paid line. Some of what the campaign drove was counted as organic. Read both, and test before deciding.
  • Measuring from the launch peak. Borrowed attention does not repeat. Set the baseline from the settled level.
  • Leaving listing changes unlogged. Without dates, a shortened subtitle and a paused campaign look like the same fall. Log every change.
  • Missing a counting change. A new integration can move installs from organic to paid with no change in demand. Check the change log and the reconciliation before the store.

Frequently asked questions

Does cutting paid install campaigns reduce organic installs?

It can. The app attribution guide notes that organic installs and brand searches are regularly credited to whichever ad came last, and the measurement plan for app installs adds that organic installs rise while paid campaigns run with no counter crediting the campaign for it. When a campaign stops, some of the store searches and visits it was prompting stop too, and they disappear from the organic line rather than the paid one. Only a test that pauses the campaign in some markets or periods and not others shows how much it was adding.

Is a fall in organic installs a store algorithm change?

Usually the explanation is closer to home. The stores do not publish how they weigh the inputs to search and browse, so an algorithm change can rarely be confirmed or ruled out, and blaming one stops the investigation. Check first what the team controls: listing changes in recent releases, campaigns and awareness work that stopped, the rating and update rhythm, owned routes into the app and any change in how installs are counted. Each of those leaves a dated change to look for.

How should organic installs be compared over time?

Like for like. Set each period beside the matching weeks a year earlier and beside the weeks just before it, so a seasonal swing is not read as a fault, and read iOS and Android, and each important country, separately, because a fall confined to one of them points to a different cause. Where weekly counts are small, pool them or read a range rather than a single figure, as the guide to reading small numbers advises, and wait for a move to repeat before acting on it.

Terms used on this page

Frameworks

Sources

  1. App Store — Apple Developer
  2. Google Play Console Help — Google

Related