9 Years of Content, 2,710 Affiliate Links: What Commission Bleed Looks Like at Scale

Clicks were up 18% year-over-year. Revenue had barely moved.

The creator assumed it was a conversion problem — maybe the audience had changed, maybe the algorithm had shifted. They'd been making content for nine years. They thought they understood their numbers.

They didn't. And what we found when we audited their full catalogue was worse than any of us expected.

---

They Had No Idea How Many Links They Actually Had

Before we ran the audit, we asked them for an estimate. "Around 1,500," they said.

We crawled every video description across 847 videos using the YouTube Data API. Every external link, every redirect, every affiliate URL — captured and classified.

The real number: 2,710 affiliate links.

Nearly double their estimate, spread across nine years of content, pointing at 1,124 unique products across four affiliate networks.

The oldest links in their catalogue were 9 years old. A buying guide published in 2018 for a product category was still pulling 40,000 views a month. Nine years of evergreen content — and nobody had checked what happened at the end of those links.

---

More Than Half the Catalogue Was Broken

We traced every redirect chain. Validated the affiliate tag at every hop. Checked stock status on Amazon and the software network. The results came back worse than the creator expected.

1,602 of their 2,710 affiliate links had a measurable problem. That's 59.1% of their catalogue.

The tag-stripped links are the ones that hurt the most to understand. The link works. The visitor lands on Amazon. The product is in stock. The sale completes. And the creator gets paid nothing — because their tracking parameter disappeared somewhere between their link and Amazon's server. No error. No signal. Just a missing commission.

The older the content, the worse it gets

Nine years of compounding link decay looked like this:

This creator had 634 links older than 5 years. 520 of them had at least one problem. Content that still ranked. Traffic that still arrived. Revenue that had quietly stopped.

---

Someone Was Taking 7.4% of Every Click at the Exact Moment It Happened

Six weeks before the audit, the creator had installed pa.js on their companion website. What it recorded over those 43,200 clicks changed how they thought about Honey.

7.4% of every affiliate click had its tag overwritten in the visitor's browser — after the redirect chain had already confirmed the link was healthy.

This isn't a server-side problem. No redirect scanner catches it. It happens in the split-second between the visitor clicking and the page loading. Honey — installed on the visitor's browser, recommended by creators to their own audiences — intercepts the click, removes the creator's affiliate tag, and inserts its own.

The creator's dashboard shows the click. The sale completes. No commission arrives.

---

48% of Their Audience Had Never Earned Them a Single Dollar

This was the number that made the room go quiet.

The creator's audience was global: UK (18%), India (14%), Canada (9%), Australia (7%). Over half their viewers were outside the US. Their affiliate links pointed to Amazon.com — one link, one storefront, one country.

Every UK viewer who clicked, browsed, and bought on amazon.co.uk? No commission. Every Indian viewer converting on amazon.in? No commission. Canada, Australia — same story.

This wasn't a marketing problem. They were doing everything right — creating content, driving traffic, building an international audience. The links were simply pointed at the wrong storefronts, and nobody had ever set up regional affiliate tags.

---

The Number That Made Them Go Quiet

Commission Bleed doesn't arrive as one dramatic event. It accumulates. Four streams, flowing continuously, invisible in the affiliate dashboard.

Source	Affected clicks/month	Monthly loss
Tag stripped links	18,400	$1,163
Dead links (still getting traffic)	23,100	$1,461
Browser extension hijacking	6,400	$405
Geographic revenue gap	76,800	$4,858
Total monthly		$7,887
Annualised		$94,644

$94,644 a year. Lost quietly, across nine years of content, with no single moment where it became obvious.

---

Four Weeks. $6,200/Month Recovered.

Week 1 — Geo-routing (11 minutes to set up) Amazon affiliate tags configured for UK, India, Canada, and Australia. Within 72 hours, commission notifications were arriving from amazon.co.uk, amazon.in, and amazon.ca for the first time.

Week 2 — Tag-stripped links repaired 291 of 387 stripped links replaced with pa.link redirects that preserve the tag at click time. The remaining 96 needed fresh affiliate links generated.

Week 3 — Dead links cleared 201 dead links replaced with fresh alternatives. 288 removed — products too outdated to replace.

Week 4 — Hijack prevention activated pa.js upgraded from detection to prevention. When Honey attempts to overwrite the tag at click time, the original is restored before the click completes. Hijack rate: 7.4% → 0.3%.

---

Before and After

"I assumed the drop in revenue-per-click was just the algorithm or audience behaviour changing. It never occurred to me that the links themselves were the problem. Some of those videos have been earning me nothing for years."

---

What This Means for Your Catalogue

This isn't an unusual case. A nine-year catalogue is more extreme — but the patterns hold at every scale. The tag stripping rate, the link decay curve, the geographic gap — these are structural issues, not individual bad luck.

If you've been creating affiliate content for more than two years and have never run a full audit, your numbers are likely in this range. Most creators don't find out until they look.

The first audit is free. Connect your channel and we'll show you exactly what we'd show them.

---

Methodology: All data reflects a real audit conducted using ProtectAffiliate's scan infrastructure. Channel-identifying details are anonymised at the creator's request. Revenue impact figures use the creator's self-reported commission rates and conversion data. Browser extension hijack data was collected via pa.js on the creator's owned web properties.