Tobold's Blog
Wednesday, November 27, 2024
 
The future of search and Amazon

I had a pretty bad search experience on Amazon recently. I was trying to buy a PC monitor. I had studied various websites and videos with recommendations, and knew exactly the serial number of the monitor I wanted to buy. But when I entered even just the serial number, with no additional generic search term, Amazon showed me PC monitors of other brands. The one I wanted didn't even make it on the first page.

How does that happen? Well, companies can pay Amazon for specific search terms. If you are a company selling PC monitors, and you know that a lot of websites recommend a competitor's product, you can pay Amazon to redirect searches for that product towards your product. Amazon doesn't care whether the search term you paid them for actually describes your product, they only care how much money you give them. Theoretically Pepsi could pay Amazon so much money, that if somebody searches for Coca Cola, he would be shown Pepsi instead.

Paying companies like Amazon or Google for search terms makes them a lot of money, but of course destroys the quality of the search results even more than traditional search engine optimization does. People add "Reddit" to their Google searches, because without that, they'll only see paid-for results which don't actually correspond to what they were searching for.

The interesting question now is whether AI will make the situation even worse, or whether it will improve internet searches. The general trend of enshittification and need to recoup billions of investment in AI suggests that things could get even worse. But anybody who has ever used ChatGPT knows that an AI chatbot has a secret weapon that regular searches on Google or elsewhere don't have: Refining, which is asking a follow-up question that builds on the previous one. You ask a question, get an answer, and if there is something you don't like about the answer, you can add further specification. Like asking ChatGPT to write you a paragraph, and then telling it to rewrite it in a simpler style.

I would love something like this on Amazon. Amazon has a range of filters you can apply to search results, but those are far from covering everything. If you search Amazon for a "box", you can't for example limit the results by the size of the box. It is easy to see how AI could solve this problem. The question is whether Amazon or anybody else *wants* to solve the problem and provide you with better search results. On the one side, you could imagine an AI company that provides better searches than Google, or makes it easier to find a product than Amazon, to be able to compete with Google or Amazon. But if you give the customer what they want, instead of stuffing the option that is most profitable for you down their throat, how do you make money?

The quality of search results on Amazon is bad because of Amazon FBA, "fulfillment by Amazon", which enables anyone to use Amazon for dropshipping. You buy some cheap product in China, then sell it for some markup on Amazon, having slapped some added invented "brand name" onto it. Because this has been widely touted on the internet as a get rich quick scheme, for any given product there can be dozens of those invented brands, all selling exactly the same item. And you can get the exact same item even cheaper if you leave Amazon, and buy it directly from China via Temu. This is all very bad for the customer, but also bad for the competing dropshippers, who have to lower their prices constantly, and pay Amazon for search terms in order to be the one of many identical products that finally gets bought. Amazon makes more money by selling the search terms than they would make by cleaning up their store, removing all the dropshippers, and selling the item themselves. Amazon sometimes even uses a commingled inventory system, where all the identical items sold by different dropship "brands" are piled up together, and the customer gets a random one from the pile, regardless from which dropshipper he ordered. Which gets very weird if somebody adds a fake scam product to the commingled inventory, and somebody else's dropship operation gets bad ratings for the scam, because even Amazon can't tell anymore whose product they shipped. The quality degradation of a search on Amazon isn't intended, but clearly Amazon is willing to accept this disadvantage in order to keep selling search terms to the dropshippers.

Thus it could be said that the monetary value of a search is in the price difference between the actually cheapest option for a product and the inflated price of the product that the search result shows. While search *could* be better, regardless of whether that is by use of AI or by other improvements, it is not in the interest of the people that offer the search engines to make it better, as their profit increases the worse the search result is for the customer.

Comments:
When I read stuff like this I sometimes feel like I must live in a different timeline. We use Amazon so much in our house we get deliveries pretty much daily all year round. We don't have any significant difficulty finding the exact items we want and when the items come, they're the right ones and they've come from the source we ordered from. If there are errors, which of course there are occasionally, usually because of the delivery agents rather than the shippers, they're corrected immediately as soon as we report them.

I've been using Amazon virtually since it became available in the UK. It is true that Amazon search is marginally less intuitive than it was a few years ago but I find it very simple to navigate still and very rarely does it need more than a small alteration or amendment to find exactly what I'm looking for. Maybe it depends on the kind of things you're buying? Or maybe there are regional variations?

One thing I do notice has deteriorated markedly, though, are the reviews. Either people are starting to write like AIs (By no means unlikely - if it hasn't happened yet, it will. because that's how language works.) or there are a lot of false reviews these days. I don't find it hard to pick out the genuine ones but it does waste time sorting them.
 
I pretty much stopped using google search, because perplexity is so much more useful. Not sure what there business model is, but so far it works for me.
 
I actually search for Amazon items on google as often I can find the exact item I want that way rather then using the search on Amazon's website.

@Bhagpuss I think it really depends on what you are trying to buy. I recently had to buy a laptop charger for an old laptop I have and trying to search for the specific manufacturer and model number directly on Amazon was fruitless as it just kept recommending generic chargers with a billion words in the title for Amazon SEO. I was able to find the amazon listing for the exact charger I wanted on the first page of google results.

For general purpose household items it's generally okay though.
 
I was fortunate enough to use other countries' alternatives to amazon, some of which make amazon look outright unusable in comparison. So what really annoys me that better implementation is not just possible in theory, but exist in practice, used by people elsewhere right now.

First, some of them display results as items, rather than as listings. So if you search for a specific model, you don't get 200 results with a same model, you get one result with 200 offers inside (sorted by price, delivery time etc). Which means that when you search for a range of models, you can choose between actual different items, rather than drowning in the copies of the same item, cloned dozens of times.

Second, some of them have 50+ criteria for filtering, which cover most real use cases and help tremendously.

I honestly have no idea how to shop with Amazon for anything more involved than socks.
 
Post a Comment

<< Home
‹Older

  Powered by Blogger   Free Page Rank Tool