Predictive intelligence - How to deal with masked\Different data in Subprod

[KLee19]

Hi All, 
I am working in an environment where ticket data is not cloned to sub prod environments. And its not likely we could win a case to get 10,000 partial records pushed to subprod.

So how do we deal with wanting to look at PI and training models? Has anyone had these limitations?
If we do get approval to have records in sub prod it will only be for a short period of time and only and then not in all sub prods. EG dev and not UAT. Just wondering how this will impact testing in UAT etc.

[Mark Manders]

If you have the assignment to showcase PI, you will need the data to train it. It's as simple as that. Creating huge amounts of dummy data won't help you, because it's not 'real' data. 

There is a reason for not cloning the ticket data. If you could make a case to use 1 non-prod instance for just PI and lock out all other users on that instance, you may win the case to have that data on that instance. Otherwise people just have to trust that PI will do the job they hope it will do and you just implement it on PROD. That's the choice the decision makers have at this moment.