- Published
- Nov 4, 2026
- Reading time
- 11 min read
Google Play Installs vs Ad Conversions: Choose Comparable Counts
Reconcile Google Play and advertising install counts: distinguish install clicks, acquisitions, devices, current install base and first launches.

In this guide (9 sections)
Before comparing Google Play with an advertising install report, replace the word “installs” with the exact metric, unit and population on each side. A store-listing Install click, completed acquisition, install event, current install base and first launch are different observations. Save the selected Play report and its configuration before calculating a discrepancy or alleging fabricated traffic.
An older TikTok advertiser question claims one campaign reported more installs than the app's lifetime total in Google Play. The opened original displayed an age of seven months. Its campaign figures and claimed Play total are self-reported; the exact Play card and counting configuration were not verified in this review. They establish the question, not a measured finding of deception.
A separate Google Ads discussion asks why reported install conversions coexist with no observed app openings in internal analytics. Changing an optimization goal is not a diagnosis of that report gap. The practical first task is to identify which stage and population each system actually counts.
Name the unit before choosing a reference#
Make a small dictionary using the metric selected in the actual account. This prevents one team from using “install” to mean a download while another means a person currently reachable on a device.
| Observation | What the selected count can describe | Comparison to avoid |
|---|---|---|
| Store-listing Install click | Intent expressed on a listing surface | Treating every click as a completed installation |
| User acquisition | Acquisition under the selected Play user definition | Assuming it equals devices or repeated install events |
| Device acquisition | A device acquiring the app | Treating every device as a different person |
| Install event | An installation occurrence, including repeat installations under the metric's rules | Calling the total first-time users |
| Installed audience or install base | An installed population at the report's reference point | Calling it cumulative campaign delivery |
| First open or first launch | Initial use after installation under that system's definition | Treating a download as observed use |
| Ad-attributed install conversion | Credit for the selected action under advertising rules | Treating it as every installation from every source |
Google's Statistics reference defines the Play user, device, acquisition, install-event and installed-population families. These are selectable report definitions, not synonyms. Write the exact selection beside every exported number.
The broader platform measurement comparison guide explains the surrounding discipline: event completeness and advertising credit need separate evidence. This guide establishes the store-side acquisition reference before that attribution comparison.
Choose the install metric before comparing totals
A store visit, acquisition, install event and first launch are not interchangeable. Discuss which metric answers your campaign question.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
1. Capture the actual Play card and report#
For the disputed number, retain the app package, report page, metric label, user/device selection, new/returning setting, date range, aggregation and filters. Record whether the screen is a current overview, a detailed Statistics report, an exported monthly file or a historical screenshot.
A screenshot caption saying “lifetime installs” is insufficient if the underlying card is unknown. Ask the account owner to open the card's detailed report and preserve its configuration. Do not silently reinterpret it as new users, active devices or cumulative downloads because one definition would make the numbers look plausible.
Compare like with like within Play first. Can the selected detailed report reproduce the card for the same period and filters? Does an exported file contain that same metric, or a legacy column with a different label? Keep an unavailable historical configuration as a limit on the investigation.
This is particularly important when a campaign predates a report change. A current definition can help prepare today's comparison; it does not automatically identify what an old screenshot showed or explain an earlier discrepancy.
2. Separate current listing intent from completed acquisition#
Google's current store-listing performance reference describes a 2026 shift toward unique clicks on Install, Open and Pre-register buttons. It directs completed-acquisition analysis to the Grow overview, Statistics and downloadable reports. The page uses June and July descriptions of that transition; do not infer a precise account rollout date from them.
A listing's Install click is therefore not evidence that installation completed. Nor does an Open-button click establish the app's successful launch. Record the exact action counted and use the relevant completed-outcome report when that is your question.
The listing report also has a surface boundary. Its scope is not every way an app can be acquired across Google Play. A report restricted to listing interactions should not be treated as a store-wide acquisition census.
For campaign analysis, retain both intent and completion when they are useful. If Install clicks rise but completed acquisitions do not, investigate that stage transition with its own evidence. Do not calculate a completion rate by dividing unrelated report totals with different populations.
3. Keep user, device, event and installed-population counts separate#
One user can have more than one device, and one device can experience more than one installation. A total of installation occurrences can consequently exceed a unique-user count without either number being arithmetically wrong.
Google's metric-change explanation distinguishes events, uniques and installed-population snapshots. Its example includes a user reinstalling within the same week: the event total and unique-user total differ. That is a definition example, not proof that reinstalls caused the original advertiser's gap.
Keep new and returning selections explicit. Also distinguish a period's acquisitions from the installed population at its end. A current population cannot be used as the maximum number of historical acquisition events. Do not add daily installed-population snapshots together and label the sum lifetime installs.
The Grow overview reference lets the reader select user or device metrics. Its “Paid and direct” traffic category includes ads, referrals and searches containing the app's name or associated brand. That category is broader than a named ad network's credited conversions. Likewise, “Not attributed” means the store could not assign the acquisition to the defined source category; it is not a verified organic or invalid-traffic label.
Preserve that source definition with the export. A campaign-specific advertising number cannot be matched to a broad store source bucket simply because both sound paid.
4. Compare downloads and first launches as a sequence#
Google Analytics defines first_open as first use after installation or reinstallation, rather than the download itself. It is therefore a later observation than downloading an application. A raw count of these events should not be renamed unique newly acquired people without establishing the corresponding identity and counting scope.
Google's app-conversion comparison reference explicitly distinguishes Google Play download conversions from first opens. It also discusses comparing app conversion actions inside Google Ads using their source, app, counting and window settings. That documentation does not establish how TikTok or another platform defines its selected column.
For an MMP comparison, use that provider's own definition. AppsFlyer's app-store discrepancy reference says its new-install recording follows first launch and can place the install on a different date from the store download. Inspect the actual report and reinstallation policy; do not substitute that description for a native Google Play or advertising conversion action.
An install on Monday and first launch on Tuesday can occupy different daily reports. More importantly, some period's opens can come from earlier installs. A same-day opens-to-installs ratio is not automatically a cohort launch rate.
If there are no observed opens, preserve both the acquisition evidence and the app's collection evidence. This guide does not diagnose a missing SDK event or downstream activity count from the store total alone. Use the tracking QA framework to define the expected collection check with the implementation owner.
5. Reproduce dates, filters and export maturity#
Prepare one comparison sheet before changing any goal or campaign:
| Contract field | Required detail |
|---|---|
| Application | Package identifier, operating system and distribution scope |
| Stage | Intent click, completed acquisition, installation or first launch |
| Counting unit | Users, devices, occurrences or installed population |
| Population | New/returning, store source and advertising campaign eligibility |
| Report | Page, exact column and saved configuration |
| Time | Event/date basis, timezone, period boundaries and aggregation |
| Availability | Extraction time, update maturity and inaccessible fields |
Play's Statistics help specifies Pacific Time for installation data. Preserve that basis when comparing another report rather than treating an advertiser's local calendar dates as identical. A fixed “PT” offset should not be hardcoded across daylight-saving changes.
Google's monthly-export reference distinguishes aggregated statistics from other report types and lists install CSV columns such as current, daily and total user/device installs. It says data captured daily is posted within three to seven days into monthly files and recommends avoiding dependence on an exact update time. A missing recent row in that export is not automatically a missing install.
Keep the source file, column names, filters and extraction time together. Do not assume that a legacy export name has the same definition as a current chart just because both contain “installs.” Prefer a reproducible configured report over a manually relabeled spreadsheet total.
A hypothetical unit and cohort audit#
Suppose an authorized audit independently establishes the following over a defined interval. Every number is invented for this example:
| Defined population | Count |
|---|---|
| Distinct newly acquired Play users | 10 |
| Distinct devices receiving the app | 12 |
| Completed installation occurrences, including repeats | 15 |
| Credited install occurrences under a separately verified advertising contract | 13 |
The ad-credit count exceeds the new-user count, but compares different units. 13 / 10 = 130% is not a meaningful installation conversion rate. The audit must compare eligible credited occurrences with a comparable occurrence population before interpreting a gap. The two-device difference cannot be read as two fake installs; the table records twelve devices and ten users.
Now consider a separate, properly linked acquisition cohort of 200 distinct new devices. Assume the investigation can verify that 170 of those same devices first launched within a specified seven-day observation window. The cohort's observed launch share is 170 / 200 = 85% under those explicit assumptions.
That calculation requires the cohort relationship and a mature observation interval. It cannot be reconstructed merely by dividing an aggregate opens chart for the week by an unrelated installation chart. If that relationship is unavailable, report the two period totals separately and keep the cohort rate unknown.
Neither example establishes an expected benchmark or explains the original advertiser's account. A remaining mismatch after comparable definitions deserves investigation; metric differences are not a reason to dismiss verified defects.
Common mistakes and a practical closing checklist#
Avoid replacing one ambiguous total with another. Common mistakes include treating “Paid and direct” as one network, using current install base as historical downloads, adding snapshots, ignoring returning acquisitions, treating an Install click as completion, and assuming every download has already produced a first launch.
Before closing the comparison, retain:
- The exact Play page, card and metric definition.
- User/device/event selection and new/returning scope.
- Distribution, package and country filters.
- The selected advertising action and its source-specific rules.
- Date basis, timezone and extraction maturity.
- Any cohort relationship required by a rate calculation.
- Confirmed differences, reproduced defects and unresolved evidence gaps.
For the original disputed totals, the next useful artifact is the actual Play configuration and comparable exports. Headline arithmetic alone cannot decide whether the cause is a different counting unit, missing collection, wrong report scope or invalid activity. It also does not establish billed waste or a refund entitlement.
Where AdProof already measures supported campaign delivery, retain a scoped delivery export alongside the separately defined store and advertising reports. The anonymized fashion-retail delivery-quality case establishes an impression baseline; it does not measure app acquisitions or prove incremental installs. Contact AdProof to confirm supported inventory and available export fields for the delivery investigation. Store, analytics and ad-platform owners retain responsibility for their own acquisition definitions and records.
Frequently asked questions#
Can campaign install conversions exceed a Play user count?
They can be non-comparable when one count covers credited installation occurrences and the other counts users. Verify the actual units and scope first. That possibility neither proves the campaign report correct nor establishes the cause of a particular discrepancy.
Is an Install-button click a completed acquisition?
No. The current listing-performance reference describes intent clicks and points to other Play reports for completed acquisitions. Retain the exact selected metric rather than relying on a chart's informal “conversion” label.
Does first_open mean a brand-new customer?
It identifies first use after installation or reinstallation under Analytics' definition. Establish the relevant user/device and returning-install scope separately before calling the count newly acquired people or customers.
Can I divide this week's opens by this week's installs?
Only call it a cohort rate if the numerator belongs to the same acquisition cohort and the observation interval is defined. Otherwise it is a comparison of period totals that may include different devices and installation dates.
Choose the install metric before comparing totals
A store visit, acquisition, install event and first launch are not interchangeable. Discuss which metric answers your campaign question.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
