Why it matters
Ad platforms optimize on events they can tie back to an impression or click. When attribution data is missing, truncated, or stored in a different system than revenue, three problems stack up: under-reported conversions in the UI, weak Event Match Quality on server paths, and broken joins in your data warehouse between spend and outcomes.
Performance marketers feel this as "CAPI is live but results look flat." Engineering sees high send volume with low match rate. Analytics cannot reconcile platform cohorts to order-level LTV because the click ID never landed on the purchase row.
Attribution data is also the bridge between first-party data and platform learning. Your data warehouse knows who bought and who returned. Platforms know which ad click started the session. Attribution fields are the join keys that let you train models, send values back on the correct user, and read out incrementality with defensible grain.
Attribution data
Attribution data flows through the full pLTV stack:
- Data warehouse (input): Persist click and campaign identifiers on signup, install, and order events alongside stable user keys and net revenue fields.
- Model (Churney): Train user-level pLTV on labeled outcomes joined to the same attribution fields used at score time.
- Signal design: Choose anchor event timing and signal freshness so values arrive while identifiers are still valid for matching.
- Activation (output): Churney sends conversion and value payloads directly to ad networks via Meta CAPI or Google Ads Conversion API, including GCLID, fbc/fbp, and hashed customer parameters where supported.
- Readout: Compare matched event volume, Event Match Quality, and incremental ROAS vs BAU once cohort maturity allows.
The data warehouse is modeling input, not the delivery pipe. Attribution data must be captured at the edge (web, app, checkout) and stored in the data warehouse for training, then passed again on the server event at activation time.
Category variants
| Channel | Key attribution fields | Common gap |
|---|---|---|
| Google Ads (web) | GCLID, gbraid/wbraid on consent-limited traffic | GCLID not stored on order or CRM row |
| Meta (web) | fbc, fbp, campaign/ad set IDs from pixel or CAPI | Missing fbc on server-only implementations |
| Mobile app | MMP click ID, install referrer, SKAdNetwork postbacks | Web-to-app handoff drops identifiers |
| TikTok / other | ttclid, network-specific click tokens | UTMs used as substitute without click ID |
Common mistakes
- UTM-only attribution. Campaign labels help analytics but do not replace click IDs for API matching.
- Storing IDs only in the ad platform. Data warehouse models cannot join revenue without the same keys locally.
- Truncated or hashed GCLID in logs. Breaks Google Ads Conversion API upload matching.
- Late ID capture. Click ID collected after redirect chain loses the original parameter.
- Cross-domain gaps. Checkout on a subdomain without shared cookie or server forwarding of fbc/fbp.
- Treating MMP postbacks as data warehouse truth. Postbacks activate platforms; the data warehouse still needs unified user grain for pLTV.
Advertiser lens
| Role | Cares about |
|---|---|
| UA / performance | Match rate and attributed volume after CAPI go-live |
| Data engineering | ID capture contracts across web, app, and OMS |
| Marketing analytics | Join coverage between spend, clicks, and revenue tables |
| Engineering | Server-side forwarding of click parameters on conversion endpoints |
FAQ
What is attribution data?
Identifiers and metadata that connect an ad touch to a later conversion event, used for platform matching, reporting, and data warehouse joins.
What fields matter most for server-side value signals?
Google: GCLID (and gbraid/wbraid where applicable). Meta: fbc, fbp, plus hashed email or phone when consented. Mobile: MMP device and click identifiers aligned to install time.
Why does pLTV need attribution data in the data warehouse?
Modeling and calibration require the same join keys you will send at activation. Training on orders without click IDs produces scores you cannot match back to campaigns.
Does attribution data replace hashed customer identifiers?
No. Click IDs tie events to ad delivery. Hashed PII improves match rate on top of click parameters when users consent.
How do I audit attribution data quality?
Measure fill rate on key fields at conversion, match rate after CAPI launch, and data warehouse join rate between orders and click ID tables.
What if iOS ATT reduces visible attribution?
Data warehouse-first modeling and modeled/consented signals still benefit from every identifier you can capture; readouts lean on holdouts and mature cohorts, not platform totals alone.
Not the same as
| Term | Difference |
|---|---|
| UTM parameters | Campaign labeling for analytics; not a substitute for network click IDs |
| Multi-touch attribution (MTA) | Statistical model across touches, not the raw identifier payload |
| Media mix modeling (MMM) | Aggregate channel inference without user-level click keys |
| Hashed identifiers | Privacy-safe customer matching fields, complementary to click IDs |
