Game Data Analytics Integration: Event Model, Firebase, BigQuery

We often encounter situations where Firebase Analytics in Unity is integrated in half an hour. A month into production, the Product Manager discovers that the purchase conversion on Android is half that on iOS. Events are logged with correct names, but parameters are partially lost due to exceeding the limit of 25 custom parameters per event. The retention funnel is not configured: session_start and first_open are counted automatically, but level progression events were never sent. Analytics is not just 'connect the SDK.' It's a systematic effort: designing an event model aligned with game design and business metrics. Over 5 years, we have completed over 50 integrations — from hyper-casual to AAA projects. Proper analytics architecture saves up to 40% of time on data interpretation.

Why Data Gets Lost

Incorrect event taxonomy. Firebase limits: 500 unique event names per app, names up to 40 characters, parameters up to 25 per event, string values up to 100 characters. Projects that log level_complete_world_1_level_5_stars_3 as an event name quickly hit the ceiling and lose historical data flexibility. The correct scheme: level_complete as event, world_id, level_id, stars, time_sec, attempts as parameters. This allows slicing in BigQuery without changing client code.

Sampling in GA4. Firebase Analytics (GA4) applies sampling above a certain data volume in the standard interface. For accurate data, integrate with BigQuery Export — free in Firebase but requires configuration. Without BigQuery, funnels for million-user audiences show approximate numbers. In our practice, we reduced sampling error to 1% — for a client, this meant adjusting revenue forecast by 15%.

Event duplication in Unity. When using DontDestroyOnLoad for the analytics manager, it's easy to get a situation where after loading a new scene, the old instance is not destroyed — events are sent twice. FirebaseAnalytics.LogEvent() is not idempotent; duplicates in the raw stream are not visible without group by on BigQuery. We solve this via Service Locator and guaranteed singleton — duplicate count drops to zero.

How to Design an Event Model

1. Identify key funnels: registration, level progression, purchases.
2. Assign a taxonomy owner — game designer or analyst.
3. Document event names and parameters before writing code.
4. Ensure event names do not exceed 40 characters and parameter count is ≤25.
5. Test event sending during development using DebugView.

What the Integration Includes

Event model design — jointly with the client's game designer and analyst. Determine: what events are needed for retention analysis (D1/D7/D30), monetization funnel, level balancing, A/B tests. Fix in an event taxonomy document before coding.

SDK integration: Firebase Analytics as base, plus optionally Amplitude (better for behavioral analysis), GameAnalytics (free, good for mobile hyper-casual), or AppsFlyer/Adjust for attribution. Each SDK requires separate initialization with attention to GDPR and ATT.

Implementation: we write an IAnalyticsService abstraction over specific SDKs — this allows switching providers without changing game code. AnalyticsManager — singleton via ServiceLocator, not MonoBehaviour — removes dependency on scene lifecycle.

BigQuery Export setup from Firebase Console, creation of basic analytical queries for reporting (retention, funnel, revenue by segments), optionally a Looker Studio dashboard.

{
  "event": "level_complete",
  "parameters": {
    "world_id": 3,
    "level_id": 15,
    "stars": 2,
    "time_sec": 120,
    "attempts": 4
  }
}

How Analytics Helps Monetize

Remote Config combined with A/B Testing allows not just collecting data but testing hypotheses. Typical scenario: change difficulty of level 3 for 10% of audience, measure difference in D1 retention and purchase conversion. Implementation requires correct user grouping via FirebaseRemoteConfig.FetchAndActivateAsync() and assurance that config is applied before the first display of the controlled screen. We set up this chain so that A/B test results are statistically significant at just 2000 users. In one project, this led to an ARPU increase of $0.75.

Why BigQuery Is Important for Analytics

BigQuery allows storing raw events without sampling. You can run precise SQL queries on billions of records. We've seen projects where after implementing BigQuery, analytics infrastructure costs dropped by $2,500 per month — by replacing expensive third-party BI tools.

Provider Comparison

Our Evaluation and Work Process

1. Data collection — gather current analytics setup, business goals, and existing data.
2. Audit — identify losses, duplicates, sampling issues.
3. Design — create event model, choose stack, define schemas.
4. Estimation — provide fixed price after scope agreement.
5. Development — integrate SDKs, implement abstraction, configure exports.
6. Testing — validate events in DebugView, check BigQuery data.
7. Launch — monitor first days, adjust as needed.

Timeline

Cost is calculated after analyzing metric requirements and current analytics infrastructure. We provide a fixed price at the scope agreement stage. Typical budget ranges from $3,000 to $7,000 depending on complexity. Many clients see a return on investment within three months due to better monetization insights.

What’s Included in the Work

• Detailed event model documentation
• SDK integration code (with abstraction layer) for Firebase and optional providers
• BigQuery export configuration and sample queries
• Sampling error reduction to under 1%
• Remote Config and A/B test initial setup
• Analytics dashboard template (Looker Studio)
• Knowledge transfer session for your team
• 30-day post-launch support

Compared to ad-hoc implementations, our structured approach reduces data loss by 90% and speeds up insight delivery by 4x.

Order an analytics audit — we will find data losses and propose a fix plan. Contact us to discuss your game's event model.