This article aims at making a bridge between the theory of optimal auditing and current procedures applied to audit files in different markets where scoring is the instrument used to implement an audit strategy. The literature has not yet developed an optimal audit policy for the scoring methodology. Our application is meant for the audit of insurance fraud but can be applied to many other activities that use the scoring approach. On the theoretical side, we show that the optimal auditing strategy takes the form of a "Red Flags Strategy" which consists in referring claims to a Special Investigation Unit (SIU) when certain fraud indicators are observed. Fraud indicators are classified based on the degree to which they reveal an increasing probability of fraud and this strategy remains optimal if the investigation policy is budget constrained. Moreover, the auditing policy acts as a deterrence device. On the empirical side, four significant results are obtained with data from a large European insurance company. First, we are able to compute a critical suspicion index for fraud, providing a threshold above which all claims must be audited. Secondly, we obtain that if the insurer applies this policy, he will save more than E22 million, which represents 43% of the current cost of fraudulent claims. Thirdly, we show that it is possible to improve these results by using information capable of isolating different groups of insureds with different morale costs of fraud. Finally, our results indicate how the deterrence effect of the audit scheme can be taken into account and how it affects the optimal auditing strategy.