# Does your model change and how would that impact your Identity Risk Score? Follow

Our models are consistently improving as both our data coverage in the Identity Network and Identity Graph improves, and as the landscape of fraud continues to evolve. We retrain the model as needed to stay abreast of the latest fraud trends, the latter is what results in potentially abrupt distribution shifts.

The distribution of our model shifts as it is trained and ingests more data (especially outcome data). Subtle shifts in the distribution occur as the model incrementally improves the AUC and continues to become smarter at predicting fraud. That being said, our models are always normalized such that a score above 400 will always represent the riskiest 20% of identities and a score below 100 will always represent the least risky 20% of identities in the Identity Network.

To understand how frequently we update our models, please review our model release approach external FAQs.

## Comments

0 comments

Please sign in to leave a comment.