Can we raise the minimum iOS version yet? Which feature should we fix first? Is the app still worth maintaining, or is it time to sunset it? In many companies these decisions are made on gut feeling — even though an analytics SDK has been running in the app for years. It's just that nobody looks at it, the events were last touched three versions ago, and the only person with access to the project is a developer who has since left.
This article isn't about marketing analytics. It describes what data you need for technical and business decisions when maintaining a mobile app, where to get it — often with no extra work — and how to collect the rest lawfully.
Five decisions that need data
Decision What data Where from Raising the minimum Android or iOS version Share of active users on each OS version Play Console, App Store Connect, Crashlytics Forcing an app update Share of users on old app versions Crashlytics or Sentry, backend logs Order of fixes and technical debt work Crashes and errors by screen or module, feature usage Crashlytics, Android vitals, event analytics Developing or removing a feature How many users use it and how often Event analytics Maintaining or sunsetting the app Active user trend, revenue, maintenance cost Store consoles, analytics, accountingMost of these questions don't require an elaborate analytics setup. They do require someone who knows the sources and can put them together.
What you already have without any SDK
Before you start adding events, check the data the stores and tools collect on their own.
Source What it shows Watch out for Play Console — Statistics Installs, uninstalls, "Installed audience" (users who have the app installed and used the device in the last 30 days), breakdown by Android version and country An install isn't active use Play Console — Reach and devices Install base by Android version, RAM, screen and country, compared with similar apps Device data, not in-app behaviour Android vitals Crash and ANR rates, with thresholds above which Google limits the app's visibility Android only App Store Connect Analytics Downloads, sessions, active devices, retention, and since 2026 also cohorts and detailed subscription data Usage data only covers users who agreed to share analytics with developers Crashlytics / Sentry Crash-free users, app and OS version distribution among active users Only if the SDK is configured correctly in the production build Backend logs Requests by app version and platform (if the app sends them in a header) IP addresses in logs are personal dataThe most underrated source is backend logs. If the app sends its version and platform with every request, you can check in minutes how many people still use a version from two years ago — with no SDK and no on-device consent. If the app doesn't do this, adding the header is one of the cheapest changes you can make in the next release.
We covered crash monitoring tools and Android vitals thresholds in more detail in the article on mobile app crash monitoring.
What the store consoles won't tell you
The consoles tell you how many people have the app and on what devices. They don't tell you what those people do in it: which features they use, where they drop out of checkout or sign-up, whether the new screen changed anything. For that you need event analytics — most often Google Analytics for Firebase, sometimes product tools such as Amplitude or Mixpanel, or self-hosted solutions.
The choice of tool matters less than the event plan. Older apps we take over rarely suffer from a lack of data — more often they have too much of it and nothing can be read from it.
Event plan: 20 good events beat 200 random ones
For an app in the maintenance phase, 15–30 events are usually enough. Rules that work:
- Each event answers a specific question. If you can't say what decision you'll make based on it, don't add it.
- One naming convention (e.g.
object_action:order_completed,invoice_downloaded) and a written event plan in the repository, next to the rest of the technical documentation. - Key processes as funnels: the start, the important steps, and the end of the process that earns or saves money.
- No personal data in parameters — no emails, phone numbers, or text entered by the user.
- Separate test data from production, so development builds don't inflate the numbers.
An example set for an app with user accounts and orders:
Event Question it answerssign_up_completed, login_failed
Do sign-up and login work
order_started, order_completed
How many orders are lost along the way
feature_used with a feature name parameter
Which features are worth developing and which to remove
error_shown with an error code
Which errors users see even though the app doesn't crash
Consent and privacy
Analytics in an app isn't just a technical matter. In the EU, storing or accessing information on a user's device is governed by the ePrivacy rules — in Poland, since 10 November 2024, by Article 399 of the Electronic Communications Law. As a rule, it requires the user's consent given after clear information. The exception only covers access that is strictly necessary to transmit a communication or to provide a service the user explicitly requested. Analytics usually doesn't fall under it, so in practice the analytics SDK should start with data collection disabled and enable it only after consent. It's worth confirming the interpretation for your specific app with a lawyer.
Other obligations:
- Store declarations — Data safety in Google Play, privacy labels and the privacy manifest in the App Store must reflect what the SDK collects. Details in our article on privacy requirements in Google Play and the App Store.
- App Tracking Transparency on iOS is required for tracking, meaning linking data with other companies' data for advertising or sharing it with data brokers. First-party analytics alone doesn't require it, but an SDK configured with advertising features does.
- Retention — in Google Analytics, event-level data can be kept for 2 or 14 months. For maintenance decisions, the shorter period and aggregated reports are usually enough.
The consequence of consent is simple: analytics data doesn't cover all users. Treat it as trends and proportions, and take absolute numbers from the store consoles and the backend.
How to read the data for typical decisions
Minimum OS version. If the oldest supported iOS or Android version accounts for a few percent of active users, and supporting it blocks library updates or requires separate workarounds in the code, dropping it usually pays off. We described how to do it technically in the article on forcing updates and a minimum supported version.
Features. A feature used by 1% of active users that generates a fifth of bug reports is a candidate for simplification or removal — not expansion.
Technical debt. Crashes and errors mapped to modules show where refactoring will have the biggest effect. More on setting priorities in Technical debt — when to pay it down and when to live with it.
Maintain or sunset. Look at the active user trend over 6–12 months, not total downloads. A shrinking active base with a constant maintenance cost is a signal to work out the options — from minimal maintenance to sunsetting the app.
The most common analytics problems in older apps
What we find most often when taking over an app:
- A Firebase project or analytics account on a former developer's private account — the data exists, but the company has no access to it. It's one of the first things to check when your developer disappeared.
- Events changed without a plan — the same action under three names, making period comparisons impossible.
- Analytics silently stopped working after a framework migration or an SDK update, and nobody noticed.
- Personal data in event parameters, e.g. an email used as the user ID.
- No consent — the SDK collects data from the first launch.
How much does it cost
Reviewing existing analytics, writing an event plan, fixing the app, handling consent, and setting up basic reports usually takes 16–40 hours. At an example rate of PLN 180 net per hour, that's roughly PLN 2.9k–7.2k net. Google Analytics for Firebase is free in its basic scope; product analytics tools have free tiers and paid plans based on the number of users or events. The rate and scope are indicative — they depend on the state of the app and the number of processes to measure.
Summary
You don't need an elaborate analytics setup to make good app maintenance decisions. First use the data you get for free: store consoles, Android vitals, crash reports, and backend logs with the app version. Add event analytics where the consoles aren't enough — with a short event plan, no personal data, and consent handling.
As part of mobile app maintenance we clean up analytics and put together the data needed for decisions every quarter: OS versions, stability, and feature usage. During an app takeover we start by recovering access to analytics accounts. If you want to know what state your app is in, request a free app analysis or get in touch through the contact form.