Here's an interesting conversation I had with a relative. Might make sense why they suck so bad after I'm done telling the story, made sense to me anyways.
This relative of mine started a wholesale radiator company, specializing in locating radiators and selling to the dealers. He created a software program (this is in the 90's) that had a feedback loop, so the system would know how many of what product were sold, returned, warrantied, etc etc. They then would have a bunch of data on what manufacturers did well at what vehicles. For example some would do better with Audi's, others better with BMW, and maybe even some better at certain models of BMW's but not other models. He went on to tell me how much computing power this program ended up requiring and how many times it crashed. Took many years and many specialty folks to come in and help him out. They eventually got it figured out and took his radiators from a roughly 60% success rate to above 95%, which is stellar in the retail world. He ended up selling the company and parties his life away now (actually a problem).
I got to talking to him about all these parts stores like Napa and Auto Zone and how they're set up. They're basically making buyout agreements with wholesalers and trusting that their wholesalers have the product right. I asked him why the wholesalers wouldn't make the same sort of system as he did and he replied "That would take an absolutely absurd amount of data! Like an unreal amount of data!". Think of how many struggles he had getting his company going, and then instead of simply just radiators, apply it to every car manufacturer and every component that is applicable to any car in any year made. Then compare every sale of each part with every return of each part and every warranty claim of each part. Once you start multiplying it out you can see how huge the data sample becomes.
Not making excuses for them, but if one of them could get a similar system online and in place, they'd take over the market.