I wanted to see if I could get some guidance around propensity thresholds. The OOTB strategy (at least in 8.5) applies the threshold decision data using the final propensity, which is set from the adjusted thompson sampling propensity.
I don't see anywhere in the strategy that we are ensuring the model is mature already. Since Thompson sampling makes the propensity completely random when new, we could exclude newer actions in error using propensity thresholds.
I don't see any extension points to add this type of logic, is there any precedence on adding a filter for ismodelmature to the OOTB propensity threshold strategy? Also open to any thoughts/comments on the above!
*Have already taken the exams/read the academy materials on this*