Excellent because it is explainable
It does not just deliver “fraud”: it exposes motive, version, intensity and evidence.
Customer operations
What the customer needs to prepare
Property GA4 and Firebase
Set assignees, Android/iOS streams, and Editor permissions to configure reports.
Channels and naming
Standardize source, medium and campaign; document Google, Meta, TikTok, affiliates and DSPs.
Consent
Choose which signals can be used across attribution, quality, and media.
Business Meta
Define valid installation, activation, retention, and decisions that the report will support.
Create the installs report by channel
Record required dimensions
In Admin → Custom definitions, register only really necessary Bridgee parameters, such as method and trust. Use native dimensions when they exist.
Create detail report
Reports → Library → Create new report → Detail report. Main dimension: First user source/medium or the approved Bridgee dimension. Metrics: New users, first_open and engaged sessions.
Add comparisons
Compare Android/iOS, campaign, country and period. Do not mix user acquisition with session acquisition without explaining the scope.
Create exploration
Explore → Free form. Lines: source/medium and campaign. Columns: platform. Values: installs, active users and engaged sessions.
Save and share
Suggested name: Bridgee | Installations per channel. Document filters and window.
The official GA4 documentation requires the Editor or Administrator profile to create detail reports. Custom dimensions may take 24–48 hours to appear.
How to measure results
| Verification | Expected | Action if failed |
|---|---|---|
| Blink → click ID | a valid click ID | review redirect, cache and replay |
| Click → install | method and confidence explained | examine reason codes and window |
| GA4 delivery | one delivery, without duplication | check native preservation Google |
| Daily Channel | media-coherent variation | verify UTMs, IDs, and consent |
| Score coverage | explicit observed/unavailable status | do not attribute absence as quality |
Why use
Anti-fraud protects the decision before invalid traffic contaminates reports, optimization and models. Extreme CTIT, replay, click ID reuse, partner conflict, automation, and journey incoherence become auditable evidence.
Protects without changing the channel
Risk qualifies traffic; does not silently reassign the installation.
Starts at shadow mode
Calibration with labels avoids blocking legitimate traffic by a new rule.
How to interpret and use
| Component | Question | Usage |
|---|---|---|
| Attribution confidence | How solid is click → install? | audit of the assignment |
| Traffic quality | Did the session have coherent activity? | channel/cohort qualification |
| Fraud risk | Is there automation or manipulation? | investigation and controlled suppression |
| Measurement integrity | Is the collection complete? | distinguish bad traffic from technical failure |
Score is not absolute truth. Use components, coverage and reason codes. Don't optimize budget just for the aggregate number.
Using Bridgee in Marketing Mix Modeling
The MMM receives aggregates by channel, date and geography: investment, facilities, trust, risk, quality, physical traffic and external controls. The trust score must be quality variable or sensitivity weight — not automatic revenue multiplier.
Recommended starters
spend, impressions, trusted installs, score coverage, fraud rate, store/OOH, seasonality, price, promotions and availability.
Good practices
consistent granularity, holdouts when possible, adstock/saturation, uncertainty ranges, backtesting and documentation of changes.
Customer best practices
- do not edit UTMs in the middle of a campaign;
- separate attributed origin from observed quality;
- monitor coverage, latency and consent;
- investigate mix changes before changing investment;
- compare GA4, media and Bridgee in the same window/timezone;
- maintain owners for marketing, analytics, app and privacy.