@FinStar wrote:
@Thomas_Thumb wrote:
@Anonymous wrote:
Actually the CRAs do not customize the algorithm. FICO creates the algorithm based on the unique datasets of the respective CRA's, resulting in slightly different weightings for different metrics. The result is the version is slightly different at each bureau.
The CRAs are FICO's primary customers. FICO does customize their base algorithms at the request of the CRAs. The CRAs can and do ask for certain attributes to be weighed differently such as impact of # cards reporting balances. In some cases one CRA may ask Fico not to consider a certain "secondary" attribute in scoring while the other two want to include it. This customization is not based on the dataset per se but, rather on how the CRAs want their in-house datasets analyzed for risk assessment.
The degree of tweaking allowed has been reduced with newer Fico algorithms such as Fico 9. Fico 8 allows greater customization than F9 and the older Fico 04/Fico 98 accommodate more customization than F8.
Should this be in scoring primer as a reference? I mean, glad to see TT expanded and clarified since it's very relevant information👍
@FinStar correct you are. Updated with attribution as follows:
...The algorithms are unique at each bureau, as they're generated from the CRAs respective unique datasets from particular time periods. They are then subject to varying levels of customization per CRA request (TT). Therefore, the same version's scores may vary across CRAs, since each CRA's algo is unique...
... Whereas earlier scoring versions were more customized for each specific bureau, an objective of Score 8 was to reduce disparity among the scores at the 3 bureaus. (Version 9 is supposed to have even less disparity...
{First paragraph modified. Last paragraph pre-existing.}
i'll probably also add a link on the link page.
Edited: you know me and my OCD, I had to AGAIN make a few changes and I also included the fact that the datasets are from a particular time period And the new information we have on young mature profiles.
also a big thanks as always to @Thomas_Thumb for sharing his wisdom. His wisdom has been instrumental in creating and maintaining the Primer.
