03 August 2004

The depth, breadth and quality of credit data in Callcredit's closed user SHARE database has enabled the credit reference agency to further improve the predictive accuracy of its CallScore credit scoring models for both SHARE and non-SHARE members.

The Fair Isaac-developed models, which rank order consumers according to their relative risk of credit default, have been empirically reweighted and revalidated using the additional volumes and types of SHARE credit data now available.

The result has been a significant uplift in the performance of the scores - in one case a 17 point increase in the Gini measurement of performance.

Fair Isaac has also aligned the scores from the revised CallScore models so that the same score from each of the models used will have the same good/bad odds. This will especially benefit organisations that use more than one level of CallScore and those which plan to move from one level to another.

Callcredit's product director Graham Lund said:

"We are delighted to make this more powerful tool available to our customers and the good news is that CallScore will keep getting better as the depth and breadth of data continues to grow.