Web2App · Published 2025-03-11 · 2 min read
How to send events properly for Facebook Web2App
The previous article explained how Facebook Web2App works. This one is about sending higher-quality events to Facebook so that campaigns learn faster.
The problem
If you run Facebook Web through AppsFlyer or Adjust you have already met the situation where the event score in Facebook is low — 3.0 at best — and Facebook is showing warnings that the events are not receiving FBC. Why does that happen?
The detailed answer from Adjust is attached.
The idea
- I usually set Adjust up together with Apphud.
- An Apphud purchase or trial goes into Adjust.
- Adjust forwards that event through the Conversions API straight to Facebook Web without adding any extra parameters to it. Adjust does not add the Facebook Click ID, which is the campaign identifier and matters a great deal.
Adjust behaves that way for any of its server-to-server integrations.
That is the cause of the low-quality events you see in the Facebook account.
What to do about it
Nothing. Attribution works fine on Facebook Web. It works even without the extra parameters — there is some sensitive probabilistic model inside Facebook.
How can event quality be improved?
- Do not use Apphud
Adjust does pass the extra parameters when it sends its own events directly. Track trials from the Adjust SDK and Adjust will send them to the server properly.
- Facebook Conversions API
Technically the hardest option: use the Facebook Conversions API directly from your app, where the Facebook SDK passes all the required event parameters itself.
You will have to take care of collecting the fbclid on your own domain and of building some equivalent of a probabilistic model, so that you can match click to install and send the event to Facebook.
- Do not pair Adjust with Apphud: use Apphud plus the Facebook SDK
The problem only arises with server-side events, so if you really want Apphud, Apphud has built a Conversions API integration for the pre-lander scenario. It is not entirely clear how that will behave when we send users straight to the store — possibly through a redirect. You will not have to build the probabilistic matching yourself; Apphud has taken that on.