Ten years on

I ran hotel affiliate campaigns from 2004 to 2016. Then I stopped, did other things for a decade, and came back to the industry this year to build ProfitTracker. Coming back after ten years away, the oddest feeling is that the dashboards look more sophisticated, but the underlying evidence is often worse.

In 2016 a third-party cookie lasted as long as you wanted it to. A click was a click. The user was usually one person, on one desktop, using a browser that mostly behaved as expected. If someone clicked your affiliate link and booked four days later, you knew. Not because of any sophistication on your end. Because the chain was unbroken.

Back then, I could look at a transaction report and know exactly why a commission had fired. It was not perfect, but it was legible.

That chain is mostly broken now.

The cookie was the spine

Affiliate networks ran on third-party cookies. Drop a cookie on click, read it on conversion, settle the commission. The whole industry was built on top of a one-line trick that worked the same way in every browser.

Safari started restricting third-party cookies in 2017. Firefox followed. Chrome was supposed to follow and has been kicking the can down the road for years. ITP, ETP, the Privacy Sandbox, whatever the acronym this quarter is. The practical effect is the same. The spine is gone, and what replaced it is a patchwork of platform-specific click IDs and probabilistic guesses.

You can still track. But you can't track the way the affiliate networks did, and nothing as universal has replaced what worked.

Cross-device was the second hit

Even before the cookies went, the device assumption broke. In 2016 a guest was mostly still a desktop session. Now a guest sees an ad on Instagram on their phone over breakfast, googles the property name on a laptop at lunch, books on a tablet that night. None of those three sessions know the others exist. Without a logged-in identity that travels between them, no client-side tracking can join the pieces up.

Hotels feel this the most. Travel is a long-consideration purchase with low repeat frequency. The booking window can be eighteen months. The journey is fragmented across surfaces nobody owns and almost no one logs into.

So you get a click on Tuesday and a booking three months later from a different IP and a different browser. Some attribution tool will assign credit. The question is whether you should believe it.

Platforms started judging their own work

This is the bit that wasn't really a problem in 2016, and is now everywhere.

Google Ads will tell you how many bookings your Google Ads campaigns produced. So will Meta. So will TikTok. They all over-count, because their job is to over-count. Add up the conversions claimed by every platform a hotel runs and the total comes out to something well over a hundred percent of the actual bookings. Everybody saw the customer last. Everybody gets the credit. The hotel gets the bill.

In affiliate-network days, the network was structurally independent of the publisher. Commission only paid out on a confirmed sale. Disputes went through arbitration. There were plenty of problems with that model, but a publisher could not mark its own homework. Now, every ad platform marks its own homework, and the IBE marks its own homework on top, and the agency marks its own homework above that.

Nobody is making it up. Everyone is grading themselves favourably, and that's a harder thing to fix than outright invention.

"AI attribution" is the new sleight of hand

In place of the verifiable signal that used to exist, vendors are selling models. The current vintage is "AI attribution," and you'll see numbers like 95 percent attribution accuracy in a vendor deck for an independent hotel doing fewer than a thousand bookings a month.

A thousand bookings is not a sample size. Google's own data-driven attribution model needs an order of magnitude more conversion volume than that before its outputs mean anything. The model is trained on Google's aggregate cross-advertiser corpus, not the individual hotel's. Most of what gets sold to small and mid-market hotels as "AI attribution" is DDA rebadged, a marketing-mix model running on the IBE's own bookings, or UTM aggregation with a confidence wrapper bolted on. None of those produce 95 percent accuracy on a hotel that books twenty-five a day. They cannot. The maths does not bend.

The accuracy figure in the deck is a sales number, not a statistical one.

What I missed, and what I came back to fix

The thing I missed about affiliate marketing was how little dispute there was about whether the booking happened. The chain was the chain. You could see it. You could audit it.

The thing that got me back into this industry in 2026 is that almost nobody is rebuilding that. The vendors are racing each other to claim more accuracy with fewer verifiable signals. The hotels are paying more for tracking that tells them less. The platforms are getting better at convincing everybody they did all the work themselves.

ProfitTracker is the small bet I'm making against that. Not a model that explains the unmeasurable. Just an honest spine for the part that is still measurable, with the gaps marked as gaps.

I'll write more about the gaps next. Specifically, where the modern referrer chain drops the ball, and what that means for hotels trying to figure out whether ChatGPT just sent them a guest.