What do you need to find out about your app?
Start with a decision. Use traffic reports to compare where people arrive from, product events to see whether they finish a task, and session replay to inspect a screen where something may have gone wrong.
What would make you change something this week?
Pick the first decision that would change what you do. A booking app needs different measurements for a launch email, for the first useful thing people do, and for a troublesome screen.
| What you want to know | What to look at | What you might do next |
|---|---|---|
| Did the launch email bring people to the site? | Visits arriving from a tagged email link. | Compare it with another source using the same date range and visit definition. |
| Did people set up a booking page they can share? | Accounts created, followed by pages actually published within seven days. | Look for the point where the measured group stops progressing. |
| What did the Save screen look like when someone got stuck? | A replay of that screen, or a map of clicks and scrolling across recorded visits. | Try the same action yourself. Ask a person what they expected if the replay leaves the cause unclear. |
On a small screen, scroll the table sideways to read every column.
What should count as a finished action?
Record the thing that actually happened. A click on Publish says someone tried. A booking page saved and available at its shareable address says the app finished publishing it.
A page view counts a page opening or a tracked change of page. A visit, also called a session, groups activity during a stretch of use. An event is one recorded action, such as a completed signup. If one person publishes two pages, that can be two events but one account that published.
For a first-publish report, count each new account once and pick a window such as seven days from signup. If 5 of 20 accounts with a full seven days publish, that is 25% of those 20. A signup from this morning belongs in a later reading once it has had the same amount of time.
Compare tools
Scroll the table sideways to read every column.
| Tool | Question it answersThe decision you can make from its reports. | What you set up and countThe actions or visits that go into the numbers. | How it connects activityWhat can link visits and what you control before sending data. | Cost and collection limitsWhat the vendor counts, caps or charges for. | What you can take outWhat can still be used outside the tool. |
|---|---|---|---|---|---|
| Google Analytics | Compare where visits come from and which sources get credit for an important action. Build a funnel when you also send the product steps. | Install the Google tag. Use tagged campaign links and mark a real success event. Traffic acquisition counts visits by source. | A browser identifier can link visits made in the same browser. User-ID can connect signed-in activity if you set it up. Keep personal data out of page URLs and events. | Standard has no subscription fee. Analytics 360 has higher limits. Standard caps daily batch export at one million events. BigQuery can charge separately. | Export report results or send raw events to BigQuery. Keep your event definitions and report settings separately. |
| PostHog | See how many people create an account, publish a booking page or come back to use it. Open linked replays when you need to inspect a session. | Install its tracking library and send events when actions succeed. Choose the event order and time window in a funnel. | A stable account ID can connect a signed-in journey. Reset after sign-out. Choose which events and recordings are allowed before capture. | Cloud counts analytics events and replays separately. Monthly allowances are shared across projects. Extra events can be dropped at a limit. | Send events and identity mappings to storage you control. Keep report definitions and check replay retention separately. |
| Microsoft Clarity | Watch a reconstructed visit or compare where people click and scroll. Use it to look for a screen you can investigate. | Add its tracking code to selected pages. Filter recordings by page or device and use a heatmap to summarize several visits. | By default, different browsers or devices can count separately. Review masking and test what is collected when someone declines cookies. | Microsoft offers Clarity free. It documents up to 100,000 recorded sessions per project per day. A replay may be unavailable for an individual visit. | Download recording summaries with playback links or export recent dashboard data. Ordinary playback is kept for 30 days. |
How do you check one result before measuring real visitors?
Use an empty test copy and one thing you can count by hand. Choose an observed action, an expected number, and a case that should leave the number unchanged.
- Check where the data will goConfirm the app uses a disposable analytics project and fake records only. Keep real sign-ins, payments and customer data out. Start without replay or automatic capture you have not configured. If the destination is unclear, stop before enabling the tracker.
- Make one deliberate success and one failed attemptCreate a fake account. Try to publish with a required field missing, then fix it and publish once. If you only need traffic, open one safe landing page using a tagged test link and look for that visit.
- Compare what arrived with what you didCheck the live event view or wait for the report. For publishing, the failed attempt should have no success event. The saved page should give one success event. Reload the confirmation page and check it does not send another.
- Test what should stay privateTry a fresh browser with collection declined as you intend to configure it. If you choose to test replay, use a single fake screen and a made-up private marker in both an input and the page text. Verify the marker is absent in the recording. Ask your coding agent to check what the test browser sends so you can confirm the tracker leaves it out.
What can you leave out of the tracker?
Send the smallest amount of information that answers the question. A signup event can say it happened and carry a limited set of labels. Keep names, contact details, passwords, access links and message text in the app systems that actually need them.
Check full page addresses and titles as well as form fields. A tool can see a private value that the app puts in a link or prints on a confirmation page. If you need to connect signed-in events, use a stable analytics identifier that avoids direct contact details. That identifier can still connect activity and needs access controls.
Decide what should happen before someone accepts measurement, when they decline, and if they change their choice. A banner records a choice. The app and tracking code still have to act on it. Test browser and server collection separately if both send events. Keep capture off on private pages unless you can verify what leaves the page.
A replay is a reconstruction of what the tool collected. Some elements can be missing or hidden. One replay can suggest a test, and several can show a pattern worth investigating. Neither tells you what every visitor intended.
What might change the bill or leave gaps in your reports?
Find out what the plan counts and what happens when it runs out. A busy visitor might make one visit, send several page views and generate dozens of events. A replay can use another allowance altogether.
Start with a real estimate for one successful visit. Multiply the number of events you actually send by expected visits. Then check if an account ID, automatic clicks, replay, data exports or additional projects change the bill or the data you get.
Check limits that are separate from price. A free tool can stop recording new sessions at a daily limit, and a budget cap can permanently drop new events. Also check how far back you can still inspect the detail. A dashboard total can remain available after the underlying event or recording has expired.
What should you save before changing tools?
Keep the definitions that made the numbers useful. Write down when each event fires, which accounts are counted, the time window and the traffic rules. Save permitted exports that you will actually need.
Run the old and new setup against the same fake success and failure first. Before replacing production tracking, compare a limited window under the same collection rules. A change in session rules or in how people are counted can change the number even if the app is working the same way.
Check the file you can export. A report table, an event history and a link to a replay are different things. Keep access and deletion rules for any copy you retain. When a definition has changed, label the date rather than joining the two numbers into one unexplained line.