Back to all posts
Analytics7 min read·October 25, 2026

What Should a New Mobile App Track?

TB
ThynkBlox Apps Team
App Engineering

A mobile app analytics tracking plan should connect each product decision to a defined user action. Start with activation and completed tasks, then add the few steps needed to understand where progress stops. For each event, record when it fires, which properties it includes and how consent affects collection. Leave out data that does not support a clear decision.

Which decisions should your tracking plan answer?

Before listing events, write down what you would change based on the answers. “Understand engagement” is too broad. “Decide whether to shorten the first booking flow” gives the tracking a purpose.

For a new app, start with questions such as:

  • Are new users reaching the outcome the app promises?
  • Which step needs attention when a task remains unfinished?
  • Do users return to complete another useful task?
  • Did a release improve the journey it changed?

Connect each question to an action. If people reach the booking form but cannot submit it, investigate that step before adding another reminder.

Your MVP scope should identify the core task. Use that task as the centre of your measurement plan. Before launch, test that each analytics event fires correctly.

What should count as activation?

Define activation as the first meaningful outcome you want a new user to reach. Avoid choosing account creation simply because it is easy to measure.

For an illustrative appointment app, activation could mean receiving confirmation of a first booking. For an illustrative planning app, it could mean saving a first plan. Choose an outcome that reflects the promise made to the user.

Write the definition precisely:

  • Who qualifies: a new user who has not previously completed the outcome.
  • What qualifies: a booking receives a confirmed status.
  • What does not qualify: opening the form, submitting an invalid request or receiving a rejection.
  • When it counts: when confirmation is recorded, rather than when the user taps the button.

Keep the activation definition separate from the underlying event. A booking can be completed repeatedly; activation describes the first qualifying completion. To see whether people keep returning after that, read how to read mobile app retention cohorts.

Choose an observation period before comparing results, and use the same period for each group being compared.

Which events belong in the first version?

Start with the core task and work backwards. Add intermediate events only where they could help explain an unfinished journey.

The following is an illustrative tracking plan for an appointment app.

Owner's questionEventExact triggerPossible decision
Do users begin the booking journey?booking_startedThe booking form opens for a new attemptReview the route into booking
Where do attempts stop progressing?booking_step_completedA defined step passes validationSimplify or investigate that step
Does the task succeed?booking_completedA confirmed booking is recordedReview the path to completion
What prevents completion?booking_failedAn attempt receives a defined failure resultPrioritise the relevant fix
Do users complete another booking?booking_completedA later confirmed booking is recordedReview reasons to return

Do not add a separate repeat-booking event unless it serves a distinct purpose. Define how reports identify later completions from the same underlying event.

Collect screen views only for named questions. A catalogue of every screen and tap is not a substitute for a decision.

How should you define completed tasks?

For each completion event, specify the evidence of success. A tap on “Save” records an attempt; your completion definition should require the save to succeed.

Record these rules beside the event:

  • Which part of the app confirms the outcome?
  • What happens if the user taps twice?
  • What happens if the request is retried?
  • Does an offline action count immediately or after confirmation?
  • How should a later cancellation appear?

For the illustrative booking app, count one completion per confirmed booking. Treat a cancellation as a separate outcome, with its own definition, rather than silently changing what completion means.

If offline actions remain pending, distinguish pending submission from confirmed completion. Do not give both states the same event name.

Define a rule for recognising duplicate deliveries without including the booking's customer details. Keep analytics identifiers separate from descriptive content wherever possible.

Which event properties are useful?

An event property adds context to an action. Include it only when that context could change your interpretation or next step.

For the illustrative booking app:

PropertyExample valueWhy include it?
entry_pointhomeCompare routes into the booking journey
step_nameslot_selectionIdentify the step being completed
booking_typeconsultationCompare predefined task categories
failure_reasonslot_unavailableDistinguish expected failure categories
app_versionrelease_aCompare behaviour across releases

These values are illustrative. Define an approved set of values for each property, including how missing or unknown values should appear.

Prefer a category such as slot_unavailable to a raw error message. Keep names, email addresses, phone numbers, appointment notes and other free text out of this plan.

Be especially cautious with search text, precise location and sensitive categories. If a broad category answers the question, use that instead. Do not attach every available account field to every event.

What naming rules should everyone follow?

Choose one naming convention and document it. For example, use lowercase words separated by underscores, with the object followed by the action: booking_started and booking_completed.

Apply the same rules throughout:

  • Use completed only for the defined successful outcome.
  • Keep event names stable across button-label changes.
  • Put categories in properties rather than creating a name for each value.
  • Give each property one meaning and one value type.
  • Document changes that affect comparisons with earlier results.

Avoid vague names such as button_clicked or success. Someone reviewing a report should understand what happened without guessing which screen produced it.

Your tracking plan should record the event name, business question, trigger, properties, exclusions and person responsible for its definition.

How should consent and data minimisation shape the plan?

Decide which collection is optional and what choices the app will offer before adding analytics. Confirm the applicable privacy requirements for your audience and data uses with qualified advice.

For optional analytics, make your collection rules explicit: what can be sent before a choice, what changes after acceptance or refusal, and what happens when that choice is withdrawn.

Treat each property as a separate decision. Ask:

  • What product question requires this value?
  • Could a less detailed value answer it?
  • How long do we need it?
  • Who needs access?
  • What should happen when it is no longer needed?

Set a retention period and a review point. Include automatically collected properties in that review.

Decide whether reporting needs continuity across sessions, devices or accounts. Avoid linking those contexts by default. Document what happens to queued events when consent changes, and test that behaviour.

How should you test the tracking plan before launch?

Include analytics checks in your Android app release testing. Verify the recorded events against actions you can observe.

Use this checklist:

  • A successful task produces the expected completion event.
  • A rejected or pending task does not appear as completed.
  • Repeated taps and retries follow the duplicate-handling rule.
  • Each property uses an approved value and type.
  • Refusing optional analytics follows the collection rules.
  • Withdrawing consent handles queued events as specified.
  • Event payloads contain no prohibited personal or free-text fields.
  • Test activity can be excluded from product reports.

Then walk through a complete journey and explain the resulting record in plain language. If the record says something happened that you did not observe, resolve the mismatch before relying on it.

What common mistakes should you avoid?

Do not treat app opens as proof that users received value. Pair return visits with the meaningful task you want people to complete.

Do not call every unfinished journey abandonment. Define how long an attempt remains open, and investigate before assigning a cause. An event record alone may leave the reason unclear.

Do not compare releases after changing an event's definition without explaining the change. Keep a record of measurement changes alongside product changes.

For a simple offline utility, a detailed account-based journey may be unnecessary. Start with the smallest plan that supports a decision.

Review tracking during ongoing app maintenance, and remove collection that no longer serves a purpose.

Frequently asked questions

How many events should a new app track?

Use the smallest set that answers your launch decisions. Begin with the core outcome and add steps only when they help you decide what to improve.

Should sign-up count as activation?

Only if creating an account is itself the promised outcome. Otherwise, measure sign-up as a step towards the first useful result.

Should analytics record every error?

Define the failure categories needed for product decisions. Keep detailed diagnostic records separate, with their own access, retention and data rules.

When should the tracking plan change?

Review it when the core journey, consent choices or product questions change. Preserve definitions where possible, and document any change that affects comparisons.

Ready when you are

Planning an app of your own?

Tell us where your project is today. We will help you work out the scope before you commit to a budget.