My Affiliate Site at 90 Days: $72.90, a 4x Search Curve, and a 17-Cent Problem

Every post in this series ends with the same promise: the next one reports the numbers, either way. MeowTested has had real search traffic for about three months now, and a team of AI agents has run the day-to-day for seven weeks. Here are the numbers.

Lifetime Amazon earnings: $72.90. In my last update on August 24 the figure was $17.36. September is running at about $100 a month.

That's still small, and most of this post is about why. The reason isn't traffic, which is doing fine. Each click through to Amazon earns me about 17 cents, and that number sets the ceiling on everything else. The charts below are live: hover them, tap them, and drag the sliders at the end.

The curve

In the latest 28 days of Search Console data, Google sent 151 clicks, against 39 in the 28 days before. The site ranks in the top 10 for 249 search queries, up from 50 a month ago, and its average position has moved from the mid-30s in June to about 11.

The flat stretch in July is the part I'd have panicked about if I hadn't seen it before. For five weeks, clicks sat between 7 and 11 a week while impressions hovered between about 480 and 650. Nothing was broken. New pages enter Google's results at position 20 or 30, where nobody clicks, and it takes weeks for some of them to climb. In mid-August a batch of them did, and the weekly count went 11, 28, 30, 37, 49.

Some of the new impressions weren't people

Switch that chart to Mobile vs desktop and a second story shows up. Over the last four weeks, mobile impressions roughly tripled and mobile clicks doubled, which is what healthy growth looks like. Desktop impressions went up almost sevenfold while desktop clicks stayed at 4 to 7 a week.

One search explains most of the gap. "best easy to clean cat fountain" logged 1,536 impressions in four weeks at an average position of 7.4. Of those, 1,530 were on desktop, and none produced a click. A real search at position 7 gets clicked. I can't prove what it is, but the pattern looks like rank-tracking software running the same query over and over, and every automated check counts as an impression for whoever appears in the results.

It matters because impressions are the number everyone watches. Site-wide click-through fell from 1.6% to about 1% this month, and taken at face value the obvious fix is to rewrite titles on pages that are working. On mobile, click-through over the same stretch stayed around 1.7% to 2%. The honest measure of growth here is clicks, and above all clicks through to Amazon. Impressions are a hint.

Where the money comes from

Tap the bars for the rate between checkpoints. From late July to late August the sites sent about 5 clicks a day to Amazon and earned about 40 cents a day. Since August 31 it's been about 20 clicks and $2.50 a day.

The clicks convert well. For every 100 clicks, Amazon records about 19 items ordered, which is high for Amazon. The limit is what people buy: the average item is about $24, the commission on it is around 3%, and nothing in the product report has cleared $100 yet.

Three things in that breakdown changed my plans.

One page earns half of it. The guide to easy-to-clean cat water fountains produced 101 of the 203 clicks, and fountains account for about $23 of the $53 in US commission. It's the site's best asset and its biggest single point of failure.

People click words, not buttons. Two-thirds of clicks come from a product name in the article text. On September 2 I added a sticky bar that follows you down the page with the recommended product. In its first eight days it got 2 clicks. Today it gained an event that records when it's actually shown, and it now hides while the product cards are on screen, so in a few weeks I'll know whether nobody sees it or nobody wants it.

ChatGPT is a search engine now. It sent 30% of the site's visits and a quarter of its Amazon clicks over those four weeks, and I've done nothing aimed at it. Bing matters for the same reason: ChatGPT's web search has leaned on Bing's index, and Bing sent 9 Amazon clicks from 2 visits.

A correction

In the last post I said VanTested's traffic drop was seasonal, because van air conditioning goes quiet in August. That was wrong, or at least badly incomplete. In the week of August 17, VanTested and BrewTested both lost about 95% of their search impressions, and neither has recovered. The pages that fell hardest were the ones that had been ranking in the top 10. Search Console shows no manual action against any of the five sites, so this looks like an algorithm change rather than anything I did to those pages.

The response was to stop spreading effort thin. MeowTested is 94% of the revenue, so until the end of October every content run goes to it and the other four sites get maintenance only.

The 17-cent problem

Revenue is Amazon clicks times earnings per click. Across every Amazon store, my earnings per click are $0.171. At that rate a $1,000 month needs about 5,850 clicks, ten times September's pace. A $5,000 month needs about 29,000, which is 48 times.

Ten times is a traffic problem, and traffic has been compounding. Forty-eight times is a different business. The way to shorten both is to earn more per click. A Litter-Robot 4 bought through Amazon pays me about $21; through the manufacturer's own affiliate program, at the 7 to 10% that direct programs typically pay, it would be roughly $50 to $70. At 30 cents a click instead of 17, the $5,000 month needs about 16,700 clicks instead of 29,000.

The catch: in July, at around 200 visits a month, I applied to several direct affiliate programs and every one turned me down. The rule now is to reapply only once the site holds about 1,000 visits a month for two months running, because a second rejection comes with another six-month wait. September is pacing around 630. I haven't applied to Whisker, the Litter-Robot maker, and won't until the site clears that bar.

In the base case the first $1,000 month lands around June 2027, and $5,000 doesn't arrive in 2027 at all. In the push case, where growth slows gently and direct programs come through in February, $1,000 arrives in February and $5,000 by the summer. The conservative case levels off around $300 a month. All three rest on about three weeks of breakout data, which is why the sliders are there.

The next 90 days

The plan comes straight from that arithmetic: make the pages that already earn earn more, and make each click worth more.

• Put the litter box roundup on page one. It sits at position 14 with 3,890 impressions in four weeks, and litter boxes are the one category on the site where products cost $150 to $700. We've owned a refurbished Litter-Robot 3 for five years; one of our two cats uses it happily and the other refuses it. That's the sentence a buyer needs, and only an owner can write it.
• Move the fountain guide from position 8 into the top three, with photos of the ceramic fountain we clean every week or two, taken apart.
• Put tools inside those pages, not on new URLs. Today I shipped a litter box cost calculator inside the article comparing automatic boxes with scooping. I started building it as its own page, then stopped halfway: a standalone calculator would have competed in search with the exact article it was meant to help.
• Build a deals page before Amazon's fall Prime sale and Black Friday, fed by the scripts that already check every linked product's live price and deal status.
• Set up Bing properly, because of ChatGPT.
• Reapply to direct programs, starting with Whisker, in the month the traffic clears the bar.

The next checkpoint is October 1, when the calculator and the sticky bar will have three weeks of data behind them. I'll report those, and the numbers, either way.

About the figures: earnings are from Amazon Associates (the $72.90 includes international stores; the per-checkpoint chart is the US store). Search data is Google Search Console. Visits and click-throughs are from Google Analytics, which misses some very short visits, so Amazon's own click counts run about twice as high.