I'm going after a new target segment - how do I update my scoring model?

As a Marketing Ops leader, there are many business questions we need to tackle every day. With MadKudu, we help make answering those questions easier.

Story

One of the business questions that we've seen many customers have is:

"How do I update my scoring to model to go after a new target segment?"

The trigger for this is Sales is complaining that the company's new target segment is not showing up as qualified in the prioritization model, so Sales is not seeing those leads come through.

As a Marketing Ops leader, the objective is to update my scoring model to show that this new target segment is qualified to Sales.

How to do this via the MadKudu platform?

Here's a step by step guide:

(1) Navigate to the Data Science Studio homepage. Duplicate the live Customer Fit model that you have in production.

(2) Navigate to the duplicated model > Data Science Studio > Overrides. Create an override with the target segment to boost them to at least good. The override depends on the criteria for the target segment and you can create it based on a MadKudu enrichment or custom attribute (from your own systems if the data has been pulled into MadKudu). Here's an example:

(3) Open up the live model and duplicated model in separate tabs. Navigate to the Validation tab for both models and compare side by side whether the performance has greatly decreased in terms of conversion rates for each segment (refer to the second section in the screenshot).

Note: typically new target segments don't perform well on historical data so we do expect the performance to decrease slightly. If the performance decrease did not drop too much, we are fine to move forward. Feel free to reach out if you'd like any advice.

For an additional check, multi-fit deploy the duplicated model. Once the performance page has been generated, you can compare apples to apples if the recall for each model has decreased over the last 6 months. This additional check could take up to 3 days to process the new model and performance page.