I am thinking we first start with a dataset.

Name
Day of Week
Time of Day
Suburb
Massage Price
Duration
Extras Price Opening Offer
Extras Price Paid
Extras 1
Extras 2
Extras 3
Extras 4
Quality Rating


If we could extra the PunterPlanet dataset that could be a good start.

From here we start to get the attributes which might lead to the algorithm. I am sure you guys can think of other attributes which would assist in predicting the best punts.

It may turnout that some attributes do not influence a good outcome while others may be critical.



The reinforcement method would apply where we are training the girls aka providers. If you have a bad punt then the payment has to be lowered. If it a good punt then pay more. The reinforcement model falls down because punters will pay for poor service...which then reinforces that the poor outcome is what is being optimised and all the current problems arise.


We could start with one of those little iPads a reception, "How was your visit today?" A new punter arriving and seeing the previous 10 punters left sad faces would avoid. A punter arriving with 15 green happy faces gets ready for a hot time. PS, this is totally corruptible