Shoppers turn Amazon's editable About You page into a showcase of its strangest guesses

One shared profile noted that its owner reads Kindle books on a phone in the shower ^[12]. On Reddit, a user said Amazon had decided they do not know what truffle should taste like ^[13]. Neither entry names a product; each is a sentence about a person, written by a recommendation system and left in a settings menu for that person to find ^[1].

Amazon pitched the page as a way to improve recommendations. "Understanding your preferences is fundamental to providing you with high-quality, relevant recommendations," the company wrote when it launched About You on 13 May ^[16][15]. According to Amazon, the profile draws on five inputs: purchase history, searches, saved lists, product reviews and conversations with the Alexa for Shopping assistant ^[5][19].

The path a product team pictures for a page like this is short. The shopper opens the list, spots a wrong guess, deletes it, and the next recommendation improves. The shoppers in the Business Insider and Verge accounts did something else. They looked for the entry that sounded like a verdict on them and posted it, and one post set off the rest ^[2]. The Verge reported that the "flat buttocks" conclusion came from one woman's purchases ^[9]. Other entries read "cusses when scared" and "finds it hilarious when toys get stuck to ceiling tiles in the classroom" ^[8].

Those odd sentences did something the settings menu had not. They sent people looking for the page. It sits under Shopping preferences in account settings ^[6], and the route The Verge printed runs through three selections and two scrolls from the account icon ^[7][20]. The Verge wrote that "some of us didn't realize" the profile could be checked there at all ^[18].

By The Verge's account, most entries are dull. "While some of these descriptions seem like a roast from Amazon, most of them are pretty generic," The Verge's Emma Roth wrote ^[4]. Her own list covers aquariums, coffee and video games, and she said all of it is accurate ^[10]. The posts show which entries people notice: the ones about bodies, friendships and temperament. They do not yet show a trust problem in any form a product team could measure. Neither report includes figures on how many shoppers edited or deleted entries.

A team weighing its own version can put each inference on a two-by-two grid. The top row describes the shopping, the bottom row describes the shopper, and the columns split right guesses from wrong ones. Roth's aquarium list is a right guess about the shopping, useful and unlikely to be posted. A wrong guess about the shopping is the case the edit button was built for. An accurate line about a shopper's friendships or body is the box to fear, since being right is what makes it uncomfortable. Getting the shopper wrong looks like the profile that misnamed a Business Insider editor's partner ^[14].

I'd keep the whole bottom row off the screen and show those inferences as the product category they drive. The tradeoff is discovery. I'd expect a list of categories to draw no Threads wave, and a page nobody shares stays three selections deep in settings ^[20].