← Back to context

Comment by ssl-3

16 hours ago

They're different risk profiles.

When a thief takes a steak from the cooler and walks out the door, they don't know who that person is. And while they may have video of parts of this, they don't necessarily see enough to prosecute. (Acting like you're stealing a steak but not actually doing it isn't a crime. Shoplifting can be hard to prove; part of that proof means demonstrating that they didn't change their mind and just drop off the steak somewhere else in the store.)

When a thief takes a steak from the cooler and walks it up to self-checkout and pays for it as if it is a bunch of bananas with their credit card, they have identified that person. They have them on video at the self-checkout committing this crime.

It actually doesn't matter much if they leave the store with their bounty or not in this second case. The crime is already done by converting the steak into bananas.

In principle I agree with you but in practice I feel like this really misses the reality of the situation. If it was an isolated incident at an isolated store in the mid 90s I think you'd be right. But presumably the thief makes a habit of this (otherwise why worry), it's likely a major chain with centrally coordinated loss prevention, the thief has presumably made a legitimate purchase from this chain at some point in the past and will again at some point in the future, and facial recognition is a thing.

In that scenario it seems to me that the best the thief can do is to "accidentally" ring up a steak as bananas occasionally and hope that if someone ever takes note that the past events will remain undetected.

That said, I'm pretty sure all of the self checkouts I've interacted with over the past several years would automatically flag such a "mistake". There are some things they're still bad at and they generate plenty of false positives but they seem to be reasonably good at identifying obvious "errors".

  • The best a thief can do is just not ring up the steak at all and if confronted just act like it’s an honest mistake.

    Anyone just buying bananas and steaks is going to look suspicious fast. So rotating the banana for other things is key. Having a mixed bag of purchases where the steaks are always just accidentally unscanned or misscanned.

    The facial recognition thing isn’t really in the picture for this minor type of crime. Law enforcement doesn’t have easy access to it as Hollywood would lead you to believe. Its use is reserved for higher profile crimes. You’d have to really be running a steak stealing criminal syndicate for this to happen. Before that, retailers would have already started forcing different procedures for the steaks. Like locking them up or pay at the butcher stand.

    One thing facial recognition can do, if used properly by the retailer, is flag you and alert the store to put extra eyes on you every time you enter the store. There’s an increased chance they’ll confront you in the act or otherwise scare you off of it.

  • I mean, eventually the people working there would take notice. "Oh, look. Banana Man is doing it again." :)

    That said: I've never, ever weighed meat at self checkout. In fact, I've never had a cashier any weigh meat, either.

    I've bought plenty of steaks in regular grocery stores, but each of those steaks (even the ones that they wrapped in butcher paper just for me after I selected them from the glass case at the back of the store) had the weight coded into the UPC that was printed at the time it was wrapped up (and weighed by the meat cutter).

    There has no further weighing required for the register to know the price, so weighing a steak at the checkout is pretty bizarro-world behavior to begin with -- at least in my experience.

    https://en.wikipedia.org/wiki/Universal_Product_Code#Number_...

    • That's why you peel off the UPC and then ring it up as however many cucumbers would approximately correspond to its weight. Fooling the overhead camera and associated ML algorithm is left as an exercise for the reader.

      2 replies →