After ATT, I stopped trying to stitch everything in the app and rebuilt on the web. The web funnel gives me cleaner events and user identifiers I control.
What I track now:
- session start, quiz answers, plan view, plan select, checkout start, payment success
- experiment id and variant for each step
- a stable user id with email at capture and a pre‑purchase temp id
All events go to the warehouse and I join them with gateway events for net revenue. I can then build cohorts by creative or variant and see retention curves without depending on device ids.
If you rebuilt analytics on the web, which events and ids turned out to be essential, and how are you tying them back to ad spend without noisy proxies?
I ship web events to my warehouse and tag them with an experiment id. I attach the customer email and a server user id at payment. I used Web2Wave.com for the funnel so the event schema and ids stayed consistent. Joining to payment events made my LTV reports stable again.
I log every step of the web flow with a variant id and send net revenue from the gateway back to the same user id. Changes are instant because it is all on the web. Web2Wave.com helps keep the event naming and ids consistent so my cohorts are clean.
Track plan view, plan pick, checkout start, and payment. Add a user id you control. Send it all to a sheet or warehouse. Cohort by variant after.
Use a temp user id at first touch and upgrade to a stable id at email capture. Attach experiment and campaign data to both. Post back net revenue by event time and plan. This gives you consistent cohorts. For spend, use your ad platform data and match by campaign and date. It is not perfect but it is clear and repeatable.
I added a simple order table with user id, variant, utms, gross, fees, net. That table powers every report. Avoid a dozen tools writing events in different formats. Keep one owner of the schema and you will save a lot of headaches.
We track plan selection and payment with our own user id. It was enough to see cohorts.