Sunday, July 24, 2022

Why Shopping Algorithms Are Terrible

Generally I've had very little luck with mainstream sources for men's clothing.

(it all seems to be the usual dark color range of navy/burgundy/gray and lots of plaid - By All the Great Powers Put in Their Drinks, I despise plaid - men's fashion sucks)

I've tried.  I mean I'm an √úbergeek, so I should be able to hack the system to find what I need, right?

Did you know that Amazon has all sorts of interesting code set up to prevent you from automating searches on their websites?  

Did you know they also took away the ability to exclude keywords in searches?  It used to be that you could search for "short sleeve shirt -plaid" and you wouldn't see any plaid.  The Plaid Cartel must have got to Bezos, 'cause that's long gone.

I even tried a couple of the "Personalized Stylist" shopping sites. They're maybe a tiny bit better, but that's only because they're not as impersonal about giving bad recommendations.

Something I've taken up lately is doing an image search when I'm looking for non-plaid shirts (men's fashion sucks) using pictures of shirts that I like.

Not Plaid

This works surprisingly well at finding colors and styles I like ...

Also Not Plaid

... but that's only half the battle (or maybe one-third).  I don't like the feel of synthetics, and so far I haven't figured out how to add in keywords to exclude.  Then there's the additional problem that I'm unwilling to pay $180 for a frickin' shirt.

Today I found a shirt I liked the look of (but not for $110) on an obscure website, and then I realized it didn't matter because it was out of stock ... and so were the shirts on either side of it.

In fact almost everything they have listed in the store is out of stock.

I think the only thing they actually had in stock was this one, and even though it's not plaid I still don't want it.

But all of this is somewhat aside from the point.  The question I keep coming back to is, "Why are the shopping algorithms so terrible?" ... and the answer is "Because they're designed to be terrible."

Let's say you run a shopping website with eleventy-billion visitors each day, and you program a special search thingy to help them find exactly what they want every single time.  Everybody wins, right?

Well, no.  Your marketing people come to you and show you the numbers about how some products have a higher profit margin than others.  A lot higher.  Worse, they also show the same numbers to the board of directors or the investors or whoever.

Now you could keep things the way they are, but the company would (theoretically) earn higher profits if they showed the shoppers the higher-margin items first, and not let them exclude anything they know they don't want.  That increases the odds that anyone who doesn't have gobs of time to waste scrolling through  thousands of ugly plaid shirts will give up and buy a more profitable one that's not quite as ugly as the rest before they get to what they really wanted.

... and if you don't change things to make the extra profit you'll have to explain that decision to the board and investors.  Good luck.

This is the digital equivalent of how grocery stores put the higher-margin items on the endcaps, and the lower-margin stuff on the top and bottom shelves in the middle of the aisles.

Paid advertising and recommendations on the site make the problem even worse, and this effect also shows up in the selections offered for every streaming video service.

I'm not sure what, if anything, can be done about it all though.