Welcome to FNI’s tips and tricks section of our training videos. Today we’re going to talk about
leveraging the champion challenger functionality within Blueprint. This is useful if you have a new strategy or decision model that you would like to test against another model.
What we are showing today assumes you have already created and saved the challenger model. Let’s go ahead and get started starting in blueprint and with the administrator rights.
Let’s navigate over to general admin and drop into the strategy admin. Here we show three active environments, our workspace where we can build our staging area, where we can test, and finally our active, which is in production. Let’s go into the workspace where we can build and then promote to the other environments. The premise for this champion challenger is that we have a new model that we would like to test on one in every five applications: a modification to the apply workflow will allow us to do this. The apply workflow is a very simple workflow designed to control the initial intake and review of the applicant.
In this example we’re going to check first for pre-bureau declines and exit the decision process if that occurs. Second we’re going to call the bureau and then finally we will call the
decision model. Very simple. Since we want to add a challenger model, we need to introduce conditional logic into the workflow so let’s add a condition. An easy way to do this is leverage one of our random number fields with numbers created between 1 and 100 for each application and then use that number to drive the routing of the application to either the champion or challenge your workflows.
Let’s go ahead and set that up let’s say 80 of the time i want to use model A I can add a decision step here that includes the random number field and when lower or equal to 80 use model A. Now i want to look at using model B the other 20 percent of the time this is going to be my challenger model. So let’s call the second model challenger.
And now with the same idea I want to leverage the random number field we used before and conditionally call this challenger model when the number generated is over 80.
It’s as simple as that to set up random conditions to call the two models. So let’s take a look at the workflow now.
An application comes in. We do a pre-bureau decline check and
then if it passes we do the bureau pull same as before and finally we conditionally route the application eighty percent of the time to model a and twenty percent of the time to the challenger model.
Now we can promote that into a test or a production environment and run the models against the bulk of applications and we can compare the performance results between the two models.
Tune in next time where we show you how to simply change variables in the models.