▸Example Google Analytics for Firebase call (the "before")
import com.google.firebase.analytics.ktx.analytics
import com.google.firebase.ktx.Firebase
// GA4 event with user_id and freeform parameters:
Firebase.analytics.logEvent("level_up") {
param("level", 7L)
param("character", "Aragorn")
param("user_id", userId)
}Most analytics SDKs accept dozens of custom parameters per event. Respectlytics's API stores exactly five fields per event: event_name, session_id (rotated every two hours), timestamp, platform, and country. Extra fields are rejected with a 400. The discipline is structural — engineers can't accidentally add PII over time because the API refuses it.
☑Remove Google Analytics for Firebase cleanly
-
1
Remove the Firebase Analytics dependency from your build (CocoaPods / SPM / Gradle / npm / pub)
-
2
Remove the GA4 property's data stream linkage from your Firebase console — the data stops flowing as soon as the SDK is removed, but the property remains until manually deleted
-
3
Replace
Analytics.logEventcall sites withRespectlytics.track("event_name") -
4
Re-audit your
Info.plistandAndroidManifest.xmlfor IDFA / AD_ID surfaces that GA for Firebase contributed -
5
Confirm
play-services-ads-identifieris no longer in the Android runtime classpath
⇋Google Analytics for Firebase vs Respectlytics — 5-field event schema
| Google Analytics for Firebase | Respectlytics | |
|---|---|---|
| Stored fields per event | — see tool note above (typically dozens of params) | Exactly 5 |
| API enforcement of schema | Lenient (extras stored) | Strict (extras rejected with 400) |
| Per-user state computable | Yes (people profiles, user properties) | No (use account system) |
| Custom event properties accepted | 25–250 depending on tool | 0 |
❓Frequently asked questions
How do we segment events without custom properties?
By using distinct event names. Instead of track('purchase', { product: 'gold_pack' }), fire track('purchase_gold_pack'). The aggregation buckets event names automatically; no manual configuration. Keep your taxonomy short (under ~50 distinct names per matrix axis) to stay navigable.
We need per-event price for revenue reporting — how does that work?
It doesn't — and that's the point. Authoritative revenue lives in your billing system (Stripe, RevenueCat, App Store Connect) with refund-aware totals and currency-conversion handling. Mirroring revenue into product analytics produces two truths that drift over time.
What if we genuinely need a sixth field for legitimate reasons?
Three options: (a) encode the variant into the event name (e.g., paywall_purchase_pro vs paywall_purchase_basic); (b) keep per-user state in your account system and don't mirror it; (c) use a different tool for the specific use case. Most product teams find (a) or (b) sufficient.
How does the API actually enforce this?
JSON validation at the API gateway. A POST /api/v1/events/ with any field outside the 5 allowed keys returns HTTP 400 Bad Request with a body listing the rejected field names. Your integration test catches the regression on the first commit that adds an extra field.