Can anyone actually prove shaving, or are we all just guessing at a gap?

Three weeks of my tracker against the network's reporting, same offer, one source, postbacks firing server side.

Week	Mine	Theirs	Gap
1	311	302	2.9%
2	288	279	3.1%
3	402	389	3.2%

What interests me is not the size of the gap, it's that it's the same gap three weeks running. Random loss is not that well behaved.

But I can't get from there to proof and I've been trying for a month. Every mechanism I can think of that would produce a stable 3% is also consistent with something boring — bot filtering on their side, duplicate suppression, a postback timeout window. So: has anyone got a method that actually distinguishes shaving from ordinary loss, without a friend inside the network? Genuinely asking, I'd rather be told my table is meaningless.

A stable 3% is what a fixed rule looks like, and a fixed rule is more likely to be dedupe or a bot filter than a thief, because a thief with a fixed percentage is a thief who gets caught by exactly the table you just posted.

What's your postback failure rate — the ones your own system logged as sent but never got a 200 back on? If you don't have that number your table has a hole in it the same size as the effect you're measuring.

everyone shaves a bit. price it into the payout you negotiate and stop measuring it

There is one method that gives you a real answer and I'll give you the caveat before the method: if they catch you doing it, you are terminated that day and the balance goes with you. I've done it twice in six years and only when I was already leaving.

The method is known leads. You submit a small number of conversions yourself through the funnel — real details, unique phone number and email per lead so each one is individually identifiable, spread over several days and several IPs so they don't arrive as a block. Twenty is enough. Then you count how many of your twenty appear in their reporting.

What makes it work is that it removes every ambiguity in your table. Not "did 3% get lost somewhere" but "lead 14 was submitted at 09:12 and is not in the report". They can't attribute that to bot filtering without saying your own hand-typed lead was a bot, which is a conversation they'd rather not have.

What it doesn't do is get you paid. It tells you whether to stay.

Both times I ran it, the answer was that everything was there. Which was not the answer I expected and it changed how I read gaps like yours.

Did the same thing back in 2022 on a network that's since folded - twelve test leads over four days.

Two never showed up. Raised it, and the AM told me with a straight face that my pixel had probably misfired on those two, which was interesting because I'd submitted them by hand on my phone and there was no pixel involved anywhere.

Didn't get paid, didn't get an apology, did get an offer of a payout bump for the following month which I read as the closest thing to a confession you'll ever see in writing. Would I call that proof? Not to a third party. It was enough for me. How much are you still sending them a week?

Pulled it. 41 postbacks over the three weeks with no 200 back, out of 1,001 fired. So 4.1% of mine never confirmed and the observed gap is 3.1%.

Which means the gap is smaller than my own known loss and I've been staring at nothing for a month. Posting it rather than quietly deleting the thread.

That's the honest version, and it's also not quite the end of it. Your failures should be random and their gap was stable to a tenth of a point three weeks running. Those two facts still don't fit together.

I'd want a fourth week before I called it either way. Not accusing them — saying the table isn't finished.

So the summary is that even the guy with three weeks of server side data can't tell? Great, love that for us