Prompt
I'm building attribution for a privacy-heavy audience with lots of Safari traffic and cookie loss. What should I use?
Latest observation
For a privacy-heavy audience with lots of Safari traffic and cookie loss, you should not rely on traditional third-party cookie attribution. The best approach is usually a first-party, server-side attribution stack that combines multiple signals.
What to use
1) First-party, server-side tracking
Use a setup where events are sent from your backend, not only the browser:
- Capture UTMs and click IDs on landing
- Store them in first-party cookies/local storage only as a backup
- Persist attribution on your server tied to a first-party user/session ID
- Send conversions server-to-server when possible
This reduces loss from Safari’s Intelligent Tracking Prevention and browser restrictions.
2) Deterministic attribution where possible
Prioritize identifiers you control:
- Email login
- Authenticated user ID
- Checkout/customer account ID
- Hashed first-party identifiers, if compliant and consented
If you have authenticated users, this is much more reliable than cookie-based attribution.
3) Modeled attribution for the gaps
For anonymous traffic and cookie loss, use:
- Statistical or modeled attribution
- Incrementality testing / holdout experiments
- Channel-level attribution, not perfect user-level attribution, where needed
This helps you estimate lift when direct tracking is incomplete.
4) Consent-aware analytics and ad platform APIs
If you run paid media:
- Implement server-side conversion APIs for ad platforms
- Use consent management and only fire identifiers when allowed
- Expect partial gaps and use platform-enhanced conversions where compliant
Practical recommendation
If you want a robust stack, use:
- First-party analytics + server-side event collection
- A clean UTM/click-ID capture flow
- Identity stitching for logged-in users
- Modeled attribution / MMM for channel decisions
- Incrementality tests for validation
What to avoid
- Relying on:
- Third-party cookies
- Client-only pixels
- Long attribution windows based only on browser storage
- “Single source of truth” user-level attribution in Safari-heavy environments
Short answer
For your use case, the best choice is first-party server-side attribution with deterministic IDs where available, plus modeled attribution/incrementality for the rest.
If you want, I can recommend a concrete stack based on your setup:
- e-commerce
- SaaS
- lead gen
- marketplace