Data corrector
Backfilling conversions a broken postback never delivered, and cost a source never passed — the two gaps that make a campaign read as the opposite of what it was.
Two things go missing in a tracker and neither announces itself:
- A network whose postback URL was wrong for a week sends no conversions, so the campaign reads as unprofitable.
- A source that does not pass cost on the click leaves every click at zero, so the same campaign reads as pure margin.
Both are wrong in the direction nobody checks. You do not go looking for revenue you did not know you earned, and you do not question a campaign that looks profitable.
They are also almost always fixed in the same sitting, because both fall out of reconciling one month. So conversions and cost go in together.
#One dialog, two file types
There is no type selector. The two formats do not overlap, so each file is read for what it is.
That is deliberate: picking a type first means two journeys for one intention, and a chance to choose the wrong one and be told row by row that every line is malformed.
#Nothing is invented
A conversion has to attach to a click this tracker issued. A row naming a click id that does not exist is refused.
Spend has to belong to a source that exists. A row naming one that does not is refused.
A refused row says why, by file and by line. The answer is the screen here — what happened to each row — rather than a success message that leaves you wondering which of four hundred lines did not take.
#Before you upload
Check the range first. The most common mistake is uploading a month that partly overlaps data already recorded, which double-counts the overlap. Conversions the postback did deliver are already here; the file should cover only what is missing.
Get the click ids right. They come from the network's own export, and they are the ids you passed the network — the same value the conversions log shows. A network exporting its own internal id gives you a file that will be refused in full, which is the correct outcome and an annoying afternoon.
#Afterwards
The figures move immediately, which is the point — but they move for the dates the rows carry, not for today. A backfilled month changes last month's numbers and leaves this month's alone.
If a campaign was paused on the strength of the wrong figures, this is the moment to look at it again.