> and marrying that to what Amazon already knows about you.

Amazon is always putting up ads for me based on my purchase history, but they're never what I wind up buying. I have no explanation for what is wrong with their algorithm.

But you just bought a toilet seat, surely you want five more toilet seats, you toilet seat enthusiast you.

Frequently purchased together: toilet seat, and the exact same toilet seat from a duplicate listing at $20 higher price

I bought a new phone in a store, then bought a cover and a screen protecton on amazon. It sent me a notification: "We think you might be interested in a new phone".

It's pretty obvious that people don't go around shopping for screen protectors and then buy phones to match.

Just realized all the info I’m giving Amazon by using the app. Time to change up habits for 2020

I do all my product searching in incognito mode

I've always assumed that was an artifact of free returns; if you bought one thing, maybe you're actually still shopping and want to know other options. I've noticed when I get these they're frequently more expensive than the one I already bought.

I've always assumed it was repairmen. Some people really do need to buy toilet seats again, and again, and again, each time they do a job that requires replacing one (without ever needing a bulk order of 100.) They also might switch brands depending on whichever one is cheapest at the time; or order several different ones for different concurrent jobs to see which turns out best (like Backblaze does with hard drives.)

I have this experience every time I buy something. Monitor? Needs more monitors. Yoga mats? Clearly opening a studio. Trash bags? This man hasn't learned about dumpsters. Of course this pays off eventually, but it seems like they could come up with a better strategy when they have a bunch of purchase history.

Nope. The fact is after you purchase something you are still in market for that product. So keeping you in a marketing segment will always be more profitable to the advertising team. Yes, you might be finished shopping now that you have finally decided on the best toilet seat, but the marketing segment that you fall into will always convert higher than a control group. There is just way too much data to confirm this.

It may make sense statistically, but for an individual customer it's just a strange experience. Look around this thread for all the comments pointing out the same thing, and wonder if a shopping app that was truly good would have all that people pointing out such strange behavior

And this is one of the hundreds of things that are very, very wrong with marketing these days.

They stopped free returns on Prime a long time ago. Now only a few select items offer it.

This was about the same time they stopped shipping things in cardboard boxes and went with bubble packs, and most recently, paper wrap with no padding whatsoever.

what are you talking about. i buy things from amazon weekly and i have never once received something in paper wrap with no padding and have never paid for a return.

Ordered an alarm keypad it came in a box with zero padding.. just sliding around. A lot of my wife’s orders have been in blue bubble wrapped packaging

Other things however come just as you would expect (paper wrapping or bubble wrapped). So it all depends on the amazon worker doing his or her job..

Source? Literally everything I've returned on Amazon over the last couple years has had at least one free return option (usually via UPS / Amazon Hub dropoff).

Clothing comes to mind, that's also the only case I've ever received something in just a white plastic bag.

Basically, you choose why you're returning. In some cases if you say something other than "the item was defective", it'll only have options that cost money.

It's super rare, but I've seen it before.

Ha, for me it was hammocks.. for years!

Just call me The Hammock Man!

Matter of fact, they're all in the same complex; it's the hammock complex on third.

That's over by Spatula City right?

No, the Hammock District.

Upvote for a random Simpson's reference.

When I was in college Amazon was perpetually convinced I would want multiple different textbooks for the same subject.

The worst was when I searched for gorilla glue once. Good ol' Jeffy B thought I'd also want a $100k watch, a racecar/nascar (themed) Bible, and an anatomical model of some gonads. Pretty sure I still have the screencap somewhere

Wait, I don’t understand. Why did you even need the gorilla glue if you weren’t going to to tell time on your Bible-mounted truck nuts?

Sometimes you can learn a domain-specific life-hack from that kind of thing. I'm guessing, for this example, that gorilla glue is the secret weapon of taxidermy—for all your mounting-a-deer-to-the-wall needs.

It's just about the ROI.

Turns out people who have just purchased something are far more likely to purchase that thing again than a random person.

This makes them the perfect person to advertise to.

Even if the percentage is quite low this can be better than advertising to the general population. Say the numbers were 1 in 10 000 impressions leads to a toilet seat sale, but 1 in 1000 people who have just bought a toilet seat buy another one soon after (first one was broken/distributor/toilet seat collector/remodelling whole home/etc).

If those were the numbers it would be silly to not advertise to recent buyers.

Essentially, the self selection of 'I just bought <x>' very often puts you in a category that is far more likely to buy <x> again than almost anyone else.

There is a simple answer. Once you have bought a turtle necklace, you are identified as a "turtle necklace buyer" as opposed to any other general consumer. Unintuitive as it may be, you are far more likely to buy another turtle necklace than any of 330 million Americans picked at random. Hope that makes sense.

Yeah, it's like their algorithm misses that once you've got X, you probably don't want X (at least not for a while).

But I wonder why they don't just flip the prediction? If you bought X, then advertise anything but X.

Like if you bought a toilet seat, they might advertise toilet paper instead (which would be good, because all the stores around here just ran out...)

Data is useful for a lot more than ads.

Very helpful for identifying what’s hot and releasing your own version. Figuring out seasonal trends. Or getting competitive price information to match/beet. Walmart brags about their ability to presciently stock stores, which keeps turnover & capital efficiency high.

Following up on this: "Amazon promises not to use customer purchase data for anything other than emailing receipts, but that leaves all of the data entailed in actually making the system work." (via Stratechery). Changes the equation a bit.

Maybe their real algorithm is so scary good that they can only use it for internal use. Didn't they also show pregnancy-related items to women who didn't even know they were pregnant yet?

Or maybe showing a wild selection of weird products results in people "brainstorm" other shopping needs, as a kind of anchoring.

https://en.wikipedia.org/wiki/Anchoring#Anchoring_in_negotia...

I believe you’re referring to a story about Target from a few years back:

https://www.forbes.com/sites/kashmirhill/2012/02/16/how-targ...

Yesterday I walked into target with my brother and said something along the lines of "I'd rather go to jury duty than shop at target."

I wonder if their creepy surveillance tech knows I said it. I was going there to pick up my brother's medicine from the pharmacy. It's a real shame. I used to LOVE shopping at Target. But the dang store has gotten greedier and super obnoxious over the years.

I was positive they used to use their tracking data to design store layouts with minimal clashing, such that they distribute the shoppers across the store. It seemed to really work, too.

In their newer layouts, I dunno what it is, but everything seems crammed together all in the same spot. They replaced the popcorn, icees, and hot dogs, with food I don't like that costs too much. They got rid of the handbasket things that were handy and all over the store, and got rid of the benches too. Then they sold their pharmacy to CVS.. which immediately threw out the kickass containers Target used to use and added a super obnoxious phone system.

Just overall, I hate shopping at Target now.

No, that's a misconception. They sent letter(s) to at least one person who had been purchasing products commonly purchased by [knowingly] pregnant women. In at least one case, family members "discovered" the pregnancy this way.

Turns out that math is hard and people are difficult to predict.

I suspect Amazon would have better success with me (i.e. better than 0%) if they simply showed me a random ad.

And in this way Amazon gets a lot more data and avoids the "monopoly" issue. Better optics too. But this is the only way that Amazon can get tons of data from smaller companies that they can't (don't want to) compete with (niche products or physical spaces).

Like how they rolled out call center in a box on AWS which both on boards more to AWS and helps train speech and linguistics models.