Why Doesn't Amazon's Price Sort Work?

Ex-Amazon engineer here. I worked on the smart refinements team, which shows relevant categories and refinements on the left hand side of the search and browse pages.

Broken price sort is a common complaint both inside and outside of Amazon. There have been email threads, tickets, and half working solutions to this but nothing -- so far -- has worked well enough.

First of all, no engineer at Amazon would tell you that the current state of price sort is acceptable. It is a well known problem that gets worked on from time to time (In fact, better price sort was going to be my next project before I left so, uh, sorry Internet...). The current price sort has been around for a long time and nothing has outperformed it.

Amazon can't sort by price well for several reasons.

1. What do you mean by price?

2. Low cost accessories

3. Bad item classification

4. More things that I can't get into

1. Price is hard to figure out. Do you want to sort by the base price of an item, the price plus shipping (which depends on where you live), or the price after discounts or specials? What about products that are sold both by Amazon and third party sellers, where the 3p seller charges less for the item but more for shipping than Amazon? Which price should be shown?

2. Low cost accessories dominate searches for a product. If you search for "iPod" in all-products search (APS, the default search on the front page) then you get pretty relevant results that span a few categories. To sort those you first need to select a category (we'll get to that later) and then sort by price. If you select electronics you'll see a few relevant results and a ton of accessories.

You could argue that Amazon should have better algorithms (and that's probably true), but there's only so much you can do. If you really want to sort by price then you shouldn't care about the relevancy of the results. To argue otherwise means that you won't have results that are increasing in price as you scroll down. You'll see more relevant but more expensive results at the front.

3. Both 3p and internal item-to-category classification continues to be a struggle at Amazon. I knew several people on the team that helped improve item classification while I was there and they did great work and made huge improvements to the catalog accuracy. Amazon has systems that learn the correct classification and suggest the correct categories to 3p sellers. Unfortunately, you'll still find sofas in the electronics category. This really screws up the results. Your outs are only as good as your ins.

4. There are some fundamental design choices that make solving this issue really hard. I obviously can't go into such fine detail, but try to appreciate that really smart people are working really hard so you can buy the lowest priced product as easily as possible.

In direct response to the article:

> Why would Amazon fail badly at something that seems so simple? The immediate cause is probably caching. Amazon likely has a task that periodically caches the low price for each item.

Well, caching is used extensively, but this probably isn't why the results look like they do. See #1.

> Tellingly, if you search for items without many Marketplace listings, the price sort is much more reliable.

Again, that's because if you exclude marketplace sellers the price is easier to calculate.

> Perhaps they can stop trying; why not just exclude Marketplace items from the price sort calculation altogether?

Seriously? Exclude a huge portion of all the products sold on Amazon? Crazy.

Price sort is hard. Just sort by popularity.

I think your points 2-4 all make sense, however:

    1. What do you mean by price?
    1. Price is hard to figure out. Do you want to sort by the
    base price of an item, the price plus shipping (which
    depends on where you live), or the price after discounts
    or specials? What about products that are sold both by
    Amazon and third party sellers, where the 3p seller
    charges less for the item but more for shipping than
    Amazon? Which price should be shown?

What possible reason would we have to care about the "base" price before shipping, discounts, or specials? When I say sort by price, I mean sort by how much money my bank account decreases by, and I think Amazon already knows this, because when you look at "New" or "Used", it sorts by base price+shipping, just like it should.

That being said, the other 3 reasons sound intractable, and are exactly the reasons I would have guessed.

What really gets me is when I'm on a site (Newegg does it too) and I search for, let's say, 4GB DDR3 1333. What comes back in the results? A $3000 PC that has two sticks of the RAM I'm searching for. That's not RAM. That's a product featuring that RAM. Or more to your point, search for HDMI and you get TVs and video cards that have HDMI ports on them. If I wanted to look for TVs, I would have said TV.

Sorting by most sold/popular is ok on the fist items because other ppl looked for the actual stuff and bought it.

but even that doesn't help on less popular categories

Except that shipping is not necessarily obvious. There are several options (overnight, 2 day, longer) with different price points. Also, if you're a Prime member you get free 2 day shipping, but what if I'm prime but not logged in/using a different machine?

And setting a default shipping cost is likely to further confuse customers. E.g. if you show the cheapest option always, there will be a price bump at checkout, but you don't want to show expensive option by default because it raises all perceived prices.