It would be based on your individual circumstances, but some general rules - identify borrowers with identical loan types that locked on the same day (or a range of days when the pricing has been stable). Group into protected vs. non-protected classes. Compare APR's to see if protected classes had higher APR than non-protected.
If there was a difference, identify if it was caused by different LTV, DTI, Credit score/risk that was in the pricing guidelines.
If PC borrower had a higher APR for same product/same date, not explained by underwriting differences (in policy), it could be a matched pair.
Example 30 year fixed rate, 2 loans locked when base was at 4.40%. Similar applications - male borrower had 4.75 APR, female borrower had 4.90 APR. It could be an issue if not supported by documented differences in UW issues.
_________________________
Integrity. With it, nothing else matters. Without it, nothing else matters.