Predictive Analytics Part 2: Apply Models to Your Service Processes
To effectively optimize your decisions, you first need to define your customer service processes and ensure that your agents are following them. Next, you need to add a decision step in your service processes that allow the agent to offer a cross-sell. Then you must carefully measure the success of each cross-sell with individual customers and correlate their relative success with the data that you have gathered about these particular individuals. With this information, you’re better equipped to recognize new situations where a cross-sell will work and those where it won’t.
All this may sound difficult, but mathematical models do exist that make this data collection and analysis easy. They’re part of a scientific area of study known as symbolic regression. Symbolic regression allows you to build models based on lots of data and lots of variables whose correlation to the decision may be unknown initially. For more info about building models, check out Evolved Analytics.
The way I envision this entire process working is as follows:
- Design customer service business processes with a decision node in place (like offering a cross-sell)
- Collect a couple hundred outcomes as well as the customer/persona information of each customer who accepts or rejects the offer
- Build a mathematical model to explain the patterns of outcomes
- Deploy the analytics model at the decision point in the process
Those decisions that are measured and analyzed in this manner are going to be more successful than those that aren’t.
