When an advertiser, a network and a publisher all report different numbers, the cause is rarely dramatic. It's usually a small detail in the postback setup. These are the five I see most often.
1. Trusting the network's postback doc
Network postback instructions contain mistakes. On one integration, the network's own doc had a missing separator before a parameter and the wrong token value. Copied as written, every conversion would have failed. Always read the postback URL character by character before using it.
2. Putting the click ID in the wrong sub ID slot
Some networks reserve the first sub ID for non-unique values, like a publisher ID, and expect unique click IDs elsewhere. Put the click ID in the wrong slot and conversions can't be matched. Confirm each party's sub ID rules before mapping anything.
3. Not testing with a real click
A test button in a network dashboard doesn't always behave like a real conversion, and on some platforms it doesn't show up in reporting at all. The only reliable test is a real click, a real conversion and a manual postback fire with that click's ID, then checking that it booked at the right revenue and payout.
4. A typo in the conversion type
If the postback sends a conversion type the tracker doesn't recognize, some trackers record it as a generic conversion instead of rejecting it. You get junk data without any warning. Check every new conversion type in your reports.
5. Leaving the redirect domain unset
In RedTrack, if a campaign's redirect domain is left unset, it falls back to a shared domain instead of yours. It still works, but you lose the benefit of your own branded tracking domain. Set it on every campaign.
The simple rule
Every postback gets tested end to end with a real click ID before any traffic is sent. It takes minutes, and it's the difference between numbers every party trusts and a week of reconciling reports.