Going to production
The switch itself is one key. Everything on this list is about making sure switching it doesn't surprise you.
Every destination you actually use has been tested, not just one
A test-key request that succeeds for Google doesn’t tell you Meta will too. Check the per-destination status in the response for each one you’ve connected, not just the top-level status.
Your code handles a partial failure, not just success or total failure
It’s normal for one destination to accept an event while another rejects it, they validate independently. Decide what your own system does when that happens, log it, alert on it, retry your own call, rather than treating the whole response as pass/fail.
Retries use a stable id, verified
If your own client retries a request (a timeout, a dropped connection), confirm it resends the same id rather than generating a new one. See Handling retries for why that matters.
User data is hashed before it leaves your server, not after
Toplu never sees a raw email address or phone number if your integration is correct. Confirm hashing happens in your own code, not by checking Toplu’s side.
Switch the key, not the code
Swap tk_test_... for tk_live_.... Nothing else about the request should need to change, if it does, that's a sign the integration was built against test-mode behavior that doesn't hold in production.
Need a hand?
Coding is hard sometimes, and that's okay. Visit our help center or reach out directly, we're here to help.