Is that really a primary growth factor? I've never once acted on an Amazon recommendation, and it's always felt almost entirely useless - "you just bought a vacuum, here's 15 more vacuums just in case you're a vacuum-obsessed lunatic"

That, and the ONLY analytics based feature I used was the "After viewing this item most people purchased" section WAS THE ABSOLUTE BEST for shoppers. It made me feel like I could look at a vacuum, but then be directed to the BEST one. It was like, I'm doing all this crazy research and stuff, but how could a billion other people's research help me here. It seems like they run experiments now where once in a blue moon you might see that section.

I know why they got rid of it - and Amazon has this problem all over the place, but some products just sit, while others, with really no difference just fly off the shelves.

This feature still shows up almost constantly for me... there are a few flavors (some of which appear contemporaneously on a single product page for me) including "after looking at this item people purchased..." and "similar items to this item..." and "people who bought this also bought..."

Wouldn't it be nice if you could just choose what information to get, unless of this dice-rolling and analytics and cookies approach?

True, I think like most market incumbents, Amazon has become lazier lately. The vacuum-obsession problem, the proliferation of counterfeit goods, and subpar third party sellers are all more recent problems.

But if you just bought 1984 for example, there's a good chance you'd also want to buy WE, Brave New World, Erewhon, etc. The algorithm made sense for the book store, but it makes less sense for the "Everything Store".

FWIW, I just got a better book suggestion from your comment than I ever got from Amazon.

I wonder if some of this is just....the very low average item quality on Amazon these days.

My recommendations right now have seven separate ten-dollar tire valve extenders, two screen protectors from the same brand (only one of which is for my phone- and that one I already bought!), a $1,000 drone i accidentally clicked an ad for last week, two more screen protectors from other brands, four Qi chargers, and five cases for my phone (I already purchased a case for my phone through Amazon).

Very persuasive!

Their book recommendations were generally relevant to your interests. Much like early Netflix, once they had enough data on what you liked they could correlate that against other customers with similar tastes.

I doubt it is. Recommendations can be great for certain kinds of purchases -- books, Amazon's original niche, being an obvious example: "people who looked at this book also tended to look at these books" can be a meaningful signal. But I don't think that's what helped Amazon to astronomic growth, particularly when we're comparing it to Sears.

The recommendations were useful to me while I was still browsing, not so much after I'd made a purchase. It made browsing a lot easier.

Yup.

Worth it

The recommendations have never been the selling point for me.

Instead, the thing that got me to use Amazon was that it provided you with access to a long tail of books—and later, other stuff.

Bookstores and record stores did do "special orders", but there was a lot of friction in the process. You'd have to find out about the book, visit the store (twice!), and if what you wanted was even available, you would wait a few weeks for your order to be bundled in with the next shipment from the publisher.

Amazon removed all of that—and the reviews even made it less risky. Want to read the next book in a series? Check out the reviews and if you want it, done deal! Need some obscure technical title? It'll be AT YOUR HOUSE next week.

In a way, I miss the serendipity of browsing bookstores. There was a time when it was a circa monthly ritual to go into Harvard Square on a weekend (I think stores were mostly open on Sunday by then) and make my way through all the new and used bookstores as well as run other errands. But, really, although I still like going into a good bookstore now and then, having this vast collection of (however imperfectly) reviewed books that can be in my hands within a few days--or immediately via digital--is pretty magical.

I do wonder how typical your experience is, though. I personally never found much by doing that (partly I was paralyzed by choice without a good algorithm for narrowing things down), and consequently stopped the random browsing in bookstores long before the bookstores themselves mostly disappeared.

When I was much younger (pre-teens and teens), I was a voracious reader and would spend a lot of time at the library. But even there, I would mostly stick to the long-running series that I knew well, or prolific authors that I knew I liked.

For me, it depends on the store.

Some places feel "well-curated": they may not have a ton of books, but it seems like the books they do have were carefully chosen. Maybe there are "theme" collections (local authors, ghost stories for grownups, Russian authors", etc) or well-read staff members making recommendations. I love browsing these places.

Other places seem to go with the "pile of books" approach. There's a lot of them, roughly organised ('fiction') and many of them are obvious tat ('Lose Weight the Dr. Oz Way' or something. I find it easy to leave these places without buying anything.

Also, they let users review the products. Not very long ago, it was a novel idea to let customers tell other customers via the store's own website that this or that product is no good.

Amazon’s “recommendations,” complication and clutter are a mess. They are not a desirable feature for me. They also often don’t have the best prices. I still use them a lot, but feel ripped off by Prime. I have an Alexa, but wish I didn’t. Walmart’s online is good, uncluttered, low priced, and if you’re pissed, you can get in your car and physically return it immediately for a refund. I use it when I can, to try to dollar vote against Amazon.

Free 2-day shipping was an absolute game changer. Customer reviews was also a huge competitive advantage that brick-and-mortar retailers lacked since it let you quickly identify which specific product was most popular for what you were looking for: much better than asking a store rep for their recommendation.

I'd really be interested in a graph that showed some growth metric like revenue or number of orders plotted against a timeline of when Amazon introduced certain features. I remember seeing one for Facebook that was really illuminating.

In the early days of Amazon their book recommendations were much better than any other online source I knew of. I would often visit their product pages to find new books to checkout at the library. At one point I bought some books I didn't really need because I felt guilty about always using the site without buying anything.

These days I find the recommendations pretty useless and have a slightly different opinion of buying products there just to help out the company.

I would wager that had a very tiny impact on the ultimate success of Amazon, more likely you do machine learning and are inflating its importance due to self-involvement with the subject.

The word for "when an idea that seems obvious in retrospect" is "hindsight bias"

That's generally not a useful bias to base beliefs off of :)

They used mechanical Turk to build recommendation lists.