The Costco Test: Fighting Enshittification
I have an admission to make. I own Kirkland sweatpants. My wife has made it clear that if I ever wear them outside the house, she has a divorce lawyer on speed dial.
So yes, everyone knows I love Costco Wholesale. Lately I’ve been thinking about why, and it turned into a question about product management.
When I was at Univision, we had a running joke: A/B test anything long enough and you end up with a porn site. Seth Godin has made the same point. Every test rewards the “sugar high”.
It feels like the majority of features that ship in mature products are some form of dark pattern. Must we always end up here? Pick a KPI and you’ve picked where you end up if you chase it long enough. Clicks end in clickbait. Time spent ends in doomscrolling. Revenue per user ends in hidden fees. Retention ends in the cancel flow from hell.
Nobody sets out to do evil. In a mature company, dozens of teams each own a slice of the product and a number for the quarter. Each ships its winning test. Nobody owns the whole experience.
Law professors Jamie Luguri and Lior Strahilevitz argued that dark patterns spread because companies’ own A/B testing showed they made money. Harry Brignull, who coined the term “dark patterns,” has been cataloguing the results since 2010. Cory Doctorow gave the whole thing a name: enshittification.
Then there’s Costco. Most of Costco’s profit comes from membership fees, and about 92% of US and Canadian members renew every year. Its core metric is retention, but without the cancellation maze. They make leaving easy and will refund your fee. When you can’t trap people, the only way to keep them is to be worth coming back to.
It backs that up with hard limits. Markups are capped at 14% on brands and 15% on Kirkland. And when Craig Jelinek told Jim Sinegal they were losing money on the $1.50 hot dog, Sinegal said: “If you raise the effing hot dog, I will kill you. Figure it out.” Jelinek later admitted people would have paid $1.75. A test would have shipped the price hike. They refused the winning variant.
AI raises the stakes: test variants are nearly free to generate now, and prices and patterns can be tuned to each shopper. Companies without guardrails will reach the bottom of the chart faster. So when Varadh Jain sees $79 from Uber to his house, but $26 to his neighbors house (source), we can all expect much more of this.
This is the real founder mode - someone who is focused entirely on the long term, the big picture - and has the voice to kill any experiment, no matter the pressure on that quarter’s number.
- ClicksGrab attentionClickbait, pornGodin's endpoint
- Time spentNever let goDoomscrollOutrage feed
- Revenue/userSqueeze walletHidden feesDrip pricing
- RetentionBlock the exitCancel trapRoach motel