← All guides

Polar rejected my SaaS over two words — the full teardown

Updated August 2026 · a first-person account; this is the rejection that led to Audeza

In July 2026 my merchant-of-record application to Polar was denied by automated review. I appealed. The appeal was denied the same day. The product was a legitimate SaaS with real users — and the whole problem came down to two words on my landing page.

The rejection

The denial email was generic: the product didn't meet the acceptable-use policy. No phrase highlighted, no line number, no category named. That's typical — rejection emails almost never tell you which words tripped the wire, and that information gap is why most founders conclude the review is arbitrary. It isn't.

The two words

My landing page described the product as helping with "lead generation".

That phrase sits on or near the prohibited list of essentially every merchant of record, in the same neighbourhood as gambling and CBD. Not because the products are equally risky, but because "lead generation" correlates — in an underwriter's data, not in your intentions — with spam complaints, chargebacks, and unsolicited contact. The reviewer never sees your codebase. They see your words, and they price the risk of your words.

The frustrating part, in hindsight: the reviewer was right to flag it. What the product actually did was monitor public conversations and draft replies that a human reviewed before anything was sent — social listening, in industry terms. Same product, different two words. One framing is bannable; the other is fine.

Reading your copy like a risk analyst does

The mental shift that fixed it: an underwriter doesn't read your landing page as marketing. They read it as a claims document. Every phrase gets scored against one question — "if this merchant misbehaves, what will the complaint look like?" Words that imply contacting people who didn't ask ("leads", "outreach", "cold", "blast", "scraping") predict spam complaints. Words that promise outcomes ("guaranteed", "100%") predict chargebacks. See the full list in the vocabulary guide.

The four changes that passed human review

• Audited every reviewer-visible word. "Leads" became "mentions and conversations". "Outreach" disappeared entirely. The product description changed to what it factually was: monitoring public conversations and drafting replies for human review.
• Added real legal pages. Terms, privacy, and refund policy, linked in the footer. Reviewers check for them; most indie landing pages don't have them. (Minimum contents: the legal-pages checklist.)
• Made the human-in-the-loop explicit. "No auto-send. You review every reply." Underwriters care about spam potential, not your feature list — say the sentence that removes the risk scenario.
• Escalated past the bot. Automated denial plus a denied appeal is not the end. I requested human review — and with the site fixed before the request, the same product was approved the same day.

Before and after

What this generalizes to

Since then I've collected every merchant-of-record rejection story I can find, and the pattern repeats: most indie SaaS rejections are fixable wording and legal-page problems, not business problems. But the processors won't tell you which words tripped the review, so founders either give up or reapply blind. The fix order that works: clean the copy, add the legal pages, make the product inspectable, then ask for human review. The full pre-application version of that process is in the merchant-of-record approval checklist.