That means agents and brokers aren't reaching the best prospects.

In this same data, there isn't a metric that identifies customers that may not be ready for cross-selling or up-selling today but could be in the near future. That's because the current models don't have the capability to capture how existing customers have used products over years, thus creating a hole in recognizing how to serve clients as their needs change.

The next step

This is where the multistage offset model comes in. This approach works to statistically account for the shortcomings of the traditional model used to predict which customers and products at which time. In fact, it generalizes the difference model approach in a way that:

• Is scalable to multiple time frames;
• Captures the impact of both raw and change variables;
• Can revise past predictions as new data is obtained to create increasingly reliable estimates; and
• Accounts for uncaptured data without having to measure it.

Multistage modeling in action

Suppose you would like to predict how a set of customers will respond to a pitch for an additional rider such as valuable articles insurance, perhaps to cover jewelry or a wine-cellar. You might look at whether a customer has insured any new properties, the length of time the home has been insured, and whether there is insurance on any existing high value articles such as furs, art pieces or sports cars.

To make the prediction, we 'd run a logistic regression model at least twice — at the end of Q2 and again at the end of Q4 of 2018 — to identify customers who are more likely to make such purchases.

Getting the job done

If you have the people who understand the math behind these models, then the methodology of this approach is quite simple to implement and can be used to enhance the predictive power for any Generalized Linear Modeling (GLM) based modeling or tree-based algorithms.

By doing this, you'll discover there are three groups of customers:

• Those who are ready for your pitch now;
• Those who will be ready in the near future; and
• Those who simply are at a low probability of engaging in a cross-sell or up-sell opportunity.

To mine the opportunities in the middle group — the ones who will eventually be ready — the model can be run periodically to create additional data that will better predict which customer is ready and when.

So what does this mean?

You have the capability to use the data you currently have to find the right customer the right product at the right time. You also can operationalize the whole cycle to run periodically so that it is data-driven and delivers maximum return on investment.

That, after all, is what we're all striving for.

Dheeraj Pandey is assistant vice president at EXL Analytics, a provider of data analytics solutions to financial organizations including P&C Insurance firms. To reach this contributor, send email to [email protected].

This article is based on an upcoming webinar, “Application of multistage offset models in cross-selling and up-selling Insurance products” scheduled for May 3, 2018 at 12.30 p.m. EST. Here's the link to register and attend: http://bit.ly/2HS9viJ

See also:

Here's how hybrid service models give insurers an upper hand

From model ready to business ready: Connecting data to ROI