Presumably these lenders find these scores accurate enough to prevent enough bad loans while not turning away too many credit worthy customers.
That said, obviously the models can be improved with more data. I worked for a lender for a while whose edge was giving people a slightly lower rate (or lending to some folks that just missed the cutoff with other lenders) because we thought our model was a bit more accurate than the credit scores / models used by other lenders.
But even we admitted to ourselves that these were just marginal improvements, by and large the credit score is a super meaningful indicator.
And this is precisely the point. If there was an error in your score, you'd have no way of knowing, no reasonable basis for any appeal, and it just sucks to be you. Lenders don't care if individual scores are accurate, they only care about whether they're making more money than they're losing.
It's analogous to doing AI gender prediction: it works fine in the aggregate, but you run into issues at the individual level, and it's the individual's rights that I care about.
I don't think that's true. I remember many years ago getting a free credit report on myself (it may have been a free trial from a monitoring service but I believe you can also just get it from the agencies itself)
You get to see the data that is factored into your score. Anything that majorly dings your credit should be obvious (eg Experian thinks you are delinquent on something that is actually paid off)
I never had to deal with that but likely the error then is with the reporting bank so you call them to fix the issue.
Is there a reason you think this isn't possible?
> Is there a reason you think this isn't possible?
The methodology for the calculation of credit scores is secret. You can't go on the internet and figure out how they do it. They don't tell you, they don't publish it, you can't know. Therefore you don't have the ability to check their work.
The more likely scenario where someone is going to troubleshoot is "I paid off all my debts and have a lot of unused credit, how come my score is too low for mortgage lenders to talk to me" - and likely a gross mismatch like that is going to be obvious.
They give you a pretty meaningful guide on what to look for in determining your score [0] that should be enough.
Note, I am not arguing that I think it's good the exact formula is not known, I just think it's not a real problem for either the lenders (who are the ones who actually base their business on this data) or the consumer.
[0] https://www.myfico.com/credit-education/whats-in-your-credit...
Don't they, though? How do you know? For a person who gets refused access to some important service or product because they're just 1 point under some arbitrary and unknown threshold, yes there's an impact.
Generally credit scores falls in wide bands that give a general credit risk.
573 vs 605 are both very high risk and you’re getting slammed regardless.
The scores are no more arbitrary than the requirement that bankruptcies roll off your score after 7 years.
Those were just random numbers I made up. With only a little bit of imagination, you can consider the significance of other number discrepancies.
Presumably you're talking about the case where somebody never checks their credit report, goes out to get a car loan or whatever, and gets a worse interest rate because there's an error on their report. My question is, what are they supposed to do? They're already required to give you a free report per year. There are various services that give you your score and report for free multiple times per year. I guess a push model would work better, but would you want all the credit bureau to send you a letter to your address on file every time your credit report changes?
Just because I haven't been repeating myself enough, the first thing I'm suggesting they could do is publish their credit scoring methodology.
Your comment is irrelevant to what I said.
What's that supposed to accomplish? At the end of the day, a credit score is ultimately supposed to tell the lender what % chance you're going to default. Suppose the data is already accurate (remember, you can already request your credit report to check whether it's accurate), what additional value does having the methodology provide? Are you doubting their ability to correctly execute the model? Or do you want to dispute the fact that their model gave you a credit score of 573 when you think it should be 605?
Do they?
If that was true, how could we verify it, beyond an initial gut feeling? There is no published rule book or methodology for credit score calculations!