One customer can visit your website, install your app, and return on another device. Those actions do not automatically share an identifier or reporting date. Compare the event definitions before trying to join the totals.
Why web and app numbers don't match
The web and app implementations are often built by different teams, at different times, and event names and parameters drift between them. A “purchase” on the website may be a different event from a “purchase” in the app, with different parameters and different timing. Nobody has to be wrong for the totals to disagree.
| Mismatch | Web view | App view | Report rule |
|---|---|---|---|
| Users | Browser cookie | App instance ID | Count people by user ID after sign-in |
| Installs | No install step | Store installs and first_open | Report installs from the app stores |
| Sessions | Tab and page activity | App in the foreground | Compare trends, not raw session totals |
| Purchases | Web order events | In-app purchase events | Dedupe purchases by transaction ID |
| Attribution | UTMs and click IDs | Store referrers and attribution SDKs | Label missing or modeled app credit |
| Webviews | Can count as web traffic | Can count as app traffic, or both | Set one tagging rule for in-app pages |
Users are identified differently
A web user is identified by a browser cookie, and an app user is identified by an app instance ID. A person who uses both counts as two users until both platforms send the same user ID after sign-in. Without that shared key, a cross-platform user count will overstate the number of people, and filtering the report does not fix it.

Installs and first_open are not the same event
App stores report installs or downloads. GA4 records first_open when someone first launches the app. The gap between those two numbers is normal: some people install and never open, some reinstall, and some app versions were shipped without the SDK. Report installs from the store and treat first_open as a launch measure.
Late and queued app events land on different days
App events can be queued on the device and sent later, for example when the phone was offline. That means app data can arrive late and land on a different day than the action it describes, so a daily app total can change after the report was first pulled.
Sessions and attribution follow different rules
An app session depends on the app being in the foreground, while a web session follows tab and page activity. The same user habits produce different session counts on each platform. Attribution is also uneven: web traffic is credited through UTMs and click IDs, while app installs depend on store referrers, attribution SDKs, and, on iOS, Apple’s privacy rules and App Tracking Transparency consent. Check the coverage and modeling in each report rather than assuming either platform has a complete view of the journey.
Release lag keeps old events in the data
Tracking changes that require an SDK or application-code update reach users as they install the new release. Older versions keep sending the events they were built with, so a tracking change can show up in app data only gradually. Tag the app version on events and compare like with like.
Reconcile before the meeting
- Define cross-platform users by one user ID sent after sign-in, not by cookies or instance IDs.
- Dedupe purchases across the web and app streams by order or transaction ID.
- Report installs from the app stores and engagement from GA4.
- Tag the app version on events so older releases can be separated from current ones.
- Keep one shared event naming plan for web and app, and write down every expected gap.
Where web should win
Web should be the source of truth for website acquisition, landing pages, page engagement, and web sessions. It is the right place to see whether a web campaign changed volume or quality. Web should not be used to explain app installs or in-app behavior, because the browser cookie and page activity do not describe what happens inside the app.
Where app should win
App stores should be the source for install and download counts, and the GA4 app stream should be the source for launches, in-app engagement, and in-app events. The problem comes when a team uses app sessions to judge web traffic, or uses store installs to explain website conversions.
Go further with the platform mismatch guide, the Shopify and GA4 revenue guide, and the server-side tagging guide. If you want a second pair of eyes on the numbers, start a 14-day trial to see how Maven’s tracking audit and daily health checks cover the tracking side of the gap. Want it done for you? MetricMaven can take the disputed report and map the first workflow you can trust.
Part of: Why your numbers don't match
Also in Other sources: Why customer and audience counts don’t match across your CRM and ad platforms, Why ABM platform reports and web analytics don’t match, Dynamic UTM parameters: a generator, plus tracking template vs. Final URL suffix
Frequently asked questions
Why don't app store installs match GA4 first_open?
App stores report installs or downloads. GA4 records first_open when someone first launches the app. People who install and never open, reinstalls, and app versions without the SDK all make the two counts differ. Report installs from the app stores and use first_open for launch-based analysis.
Why is the same customer counted twice across web and app?
A web user is identified by a browser cookie, and an app user is identified by an app instance ID. The same person on web and app counts as two users unless you send the same user ID on both after sign-in.
Should we report web and app separately or together?
Report them separately for channel and platform diagnosis. Combine them only for people and purchases, and only after you have a shared user ID and a purchase dedupe rule by transaction ID. Name the rule on the report so the team knows which view they are reading.
Can Maven help reconcile web and app analytics?
Maven helps with the tracking side of the gap. Its tracking audit checks each client’s Google Tag Manager for broken or stalled pixels and events, tags with no trigger, and GA4 measurement-ID mismatches. Fixes are staged with the exact diff and QA evidence and go live only after you approve them, and daily health checks alert you when a tag stops working. The definitions and source-of-truth rules in this guide still need an owner on your team.
