Appealed six times, one came back and I don't know why

Account number four went down Tuesday morning. No warning, nothing changed on the landers, same creatives that had been running for nine days.

I've appealed six times now across three accounts this year. Five came back inside the hour with the exact same paragraph, word for word, which tells me nobody read a thing. One actually came back LIVE after eleven days and I genuinely have no idea what I did differently on that one!!

So what actually works here? Is it the wording? Is it just spamming appeals until one lands on a human? Someone on a call told me you should never use the word affiliate anywhere in an appeal, which sounds like superstition to me, but at this point I'll try anything.

the one that came back was read by a person. the other five weren't. that's the entire difference

Caveat first: I've had maybe fifteen appeals go through and I still can't predict which ones land, so treat this as pattern not method.

1. The instant reply is automated and appealing again the same day just refreshes the automated one. 2) The ones that came back for me were all submitted 24-48 hours later, in the morning their time, and were three sentences long. 3) Every long appeal I've written has failed. Every single one.

The word affiliate isn't the trigger. What you're selling is.

I've been through enough of these to have a process, so here it is. It is not a magic formula and I want to be clear up front that roughly two thirds of my appeals still fail.

What the appeal system actually is

There are two layers and people conflate them constantly. The first is an automated classifier that reads your appeal, matches it against the ban reason code, and replies. It resolves most appeals within minutes. If your reply arrives inside an hour, no human touched it, and nothing you wrote mattered.

The second is a human review queue. Cases reach it either by escalation rules on the account (spend history, tenure, prior clean record) or because the automated layer scored the case as ambiguous. Everything you do in an appeal is aimed at layer two. You cannot argue with layer one.

What moves a case into layer two

• Tenure and spend. An account that spent for six months gets a look. A three-week-old account that spent forty dollars does not. This is the biggest single factor and it is entirely outside your control at appeal time.
• Specificity. A short appeal that names one concrete thing — a policy section, a specific ad ID, a date — reads as a real complaint. Generic pleading reads as bulk.
• Ambiguity in the ban reason. Broad reason codes escalate more often than narrow ones. If the code says circumventing systems you are done, that one is terminal in my experience.

What to actually write

Three sentences. What you run, what changed on the date of the ban, and one question. No apology, no explanation of your business model, no mention of how much you've spent — they can see it and saying it reads as leverage, which annoys people.

Do not appeal twice in 24 hours. Every duplicate resets the automated pass and I am fairly confident, though I have not proved this, that it also drops your escalation priority.

What does not work

Rewriting the same appeal in different words. Appealing from a different email. Adding screenshots — the queue does not open them. Threatening a chargeback, which in one case I know of turned a suspension into a permanent ban on the payment instrument as well as the account.

One more thing I am uncertain about: I have never had an appeal succeed on an account that was banned within seven days of a billing card change. That might be coincidence, my sample is four.

Summary: three sentences, 48 hours later, one concrete detail, and accept that most of them are gone regardless.

eleven days for the one that came back, so patience is genuinely part of it then

Never appeal same day. Wait one day, is different queue. Anyway this is only my experience.

FWIW the thing nobody in this thread has said: keep a copy of every appeal you send and the timestamp of the reply.

I started logging mine in a spreadsheet in March and the reply latency alone tells you which layer answered before you even read the text. Under 90 minutes, automated, don't bother reading it. Over a day, someone opened the case and you should read every word of what they wrote because the phrasing usually tells you which policy they actually applied, which is often not the one in the ban notice.