Introduction
Metrics stands for the business objective and business logic that will guide your prediction model.
In the Forwrd platform this is done by defining three key metrics:
Success Criteria - the success criteria is the definition of how the desired outcome looks like in our business application.
For example, in a 'Lead Score' use case, where we try to predict whether or not a lead will be qualified by sales, the relevant business application is our CRM and when a lead is being qualified its status is changed accordingly. So in this case the success criteria might look like this: Lead Status
Equals"Qualified" (i.e. all the leads in our CRM with a status of Qualified are examples of the desired outcome).
Failure Criteria - the Failure criteria is the definition of how the un-desired outcome looks like in our business application.
Going back to our 'Lead Scoring' example, the failure criteria might look like this: Lead Status
Equals"Unqualified" (i.e. all the leads in our CRM with a status of Unqualified are examples of the un-desired outcome).
Audience - the audience is the definition of the population that we would like our model to predict.
In a 'Lead Scoring' example, the audience could be all the new created leads. The criteria will look like this: Lead Status
Equals"New" (i.e. the model will predict the outcome of all the leads in our CRM with a status of New)
Setting Your Metrics
Go to 'Metrics' and click on 'New'
Name your model metrics and choose the base that is relevant for this metrics (learn more about bases here.....)
Define each metric by clicking the 'Set' button
For each metric define an expression that represents it.
Use the search box to find the relevant attribute (field) for your expression and select it
Once selected built your expression. You can add additional fields for your expression by clicking on 'Add Condition' or 'Add Condition Group'.
Repeat these actions for each criteria and click 'Done' when finished
Your Metric is now defined and ready to be used by a new model.
Next Step: Build a New Model.