Advertiser says my leads are fabricated. I know they aren't and I have no idea how to show it

Been running a form-fill offer into APAC since June, about 200 a week, approval rate steady at 78%.

On Friday the network forwarded me an advertiser complaint saying the September batch is "largely fabricated" and they're withholding the whole month pending review. Their evidence, as passed to me, is that a high proportion of the phone numbers do not answer and that some of the submissions share device characteristics.

The leads are real. I know the sources, I know the creatives, there's no incentive layer, nothing is being generated. But I've been staring at this since Friday and I genuinely can't work out what I would send that constitutes proof of a negative.

Has anyone been on this side of it and come out with their money?

you don't prove they're real. you make it cheaper for them to pay you than to keep arguing

The two allegations are separable and you should separate them in your reply, because one of them is nearly meaningless and the other might be your fault.

Numbers not answering. This is not evidence of fabrication and you should say so plainly. Contact rate is a function of when they call, how many attempts, and what number shows on the display. Ask for: the call attempt log per lead, the time of day, and how many attempts before they gave up. If they called APAC mobiles from a foreign landline once at 11am local, a low answer rate is the expected outcome and has nothing to do with you. I have seen an advertiser open exactly this dispute and close it within a day when asked for their dialler stats, because the dialler had been misconfigured for a fortnight.

Shared device characteristics. This one is worth taking seriously before you defend it. Shared fingerprints across a lot of submissions can mean fraud, and it can also mean everyone arriving through the same in-app browser on the same OS version from the same carrier, which is completely normal on mobile traffic in a single geo. Pull your own distribution: unique IP count, unique user agent count, carrier ASN spread, and time-between-click-and-submit histogram over the month.

That histogram is the single most useful thing you can send. Real form fills have a long messy distribution with a fat tail out past two minutes. Generated ones are a spike. If yours is messy, put the chart in front of them and the burden moves.

One caveat: if your traffic is heavily one source and that source is an app, you may genuinely have a homogeneity problem that isn't fraud and still can't be distinguished from it at their end.

Has the network actually seen the advertiser's evidence, or are they relaying a sentence from an email?

Relaying a sentence. I asked for the underlying data on Friday and got "we've requested it" and nothing since. That's part of what's bothering me.

Then your dispute is with the network and not the advertiser, and you should stop building a defence for a court that isn't sitting.

I've watched this pattern since about 2015 and it has a name it doesn't deserve. "Advertiser complaint" is the most useful phrase a network has, because it converts a decision the network made into an external fact you cannot examine. Sometimes it's true. Often the true version is that the advertiser's month came in under target, they asked the network to look at quality, and the network chose which affiliate's month to hold.

The test is simple and it costs you one email: ask for the complaint in the advertiser's own words, with a date, and ask whether other affiliates on the same offer are affected. A real complaint is a document. A manufactured one is a paraphrase that never quite arrives.

The time-to-submit histogram is the right artefact and I want to add the practical bit, because most people don't have it and won't after this either.

You can only produce it if you logged the click timestamp and the submit timestamp per lead at the time. If you didn't, you cannot reconstruct it now, and that's the whole game. Depends what you're running, but for any form-fill offer I'd log four things from day one: click time, first-keystroke time, submit time, and the referring creative. Costs nothing, sits in a table, and is worth a month's revenue exactly once every couple of years.

On this specific dispute I'd say your odds are better than even if the network is real and worse than even if it isn't. Which is it?

did they ever send the actual complaint

how long did they say the review takes

Goodwill settlement. Two words doing an enormous amount of work.

Six weeks on, so here's where it ended, which is nowhere clean.

I sent the histogram and the ASN spread. Time-to-submit was a long messy distribution with a median around 71 seconds and a tail out past ten minutes, which is not what generated leads look like. Nobody disputed it. Nobody engaged with it either.

I asked for the complaint in the advertiser's own words, as old_growth_seo suggested. Never arrived. Asked three times.

What happened in the end: they released 62% of September, described it as a "goodwill settlement", and closed the offer to me. No finding either way. I signed nothing but I also didn't push, because the remaining 38% was worth less than the two other offers I still run there.