Predicting Customer Intent: It’s All About Impact

Published On: January 15th, 20203 min read

Using predictive algorithms in customer care poses a challenge you might not anticipate: Accurate predictions don’t automatically generate better business outcomes.

When you’re automating elements of your customer care system, there’s a strong tendency to prize accuracy over everything else. After all, inaccurate predictions defeat the purpose of deploying artificial intelligence and machine learning.

New call-to-action

But zeroing in on predictive accuracy is not the whole story — predictions have to be helping you improve the customer-care experience. Thus, you have to prioritize impact over accuracy.

Let’s take a look at how this works.

Predictive Modeling in AI: A Quick Overview

Machine learning algorithms comb through billions of data points, using pattern matching to distinguish between the outcomes you want and the ones you hope to avoid. With a large enough set of training data, the algorithms learn to create more accurate predictions over time.

All of this is built around complex mathematical formulas that analyze previous patterns of behavior and predict the likelihood of these things happening again. The more accurate these calculations are, the better your predictive capability.

Every time you accurately anticipate a customer’s behavior and help improve their satisfaction with an engagement or transaction, you’re producing better business outcomes. Thus, accuracy matters — as long as it’s generating the impact you want.

Deploying Predictive Models in Customer Care

In the call center, you want customers’ questions answered and complaints resolved as quickly as possible. With predictive modeling, you correlate data from multiple sources to understand the context of customer calls, answering questions such as:

  •  Are they on a landline, cell phone or web interface?
  • What have they purchased recently?
  • How much have they purchased, on average?
  • Where are they?
  • Does local weather affect their buying experience?
  • Have they made social media posts that are relevant to their call?
  • Do you know their age, gender, income and other demographic details?
  • What are the most intuitive traffic patterns through your customer-care interface?

You also have to track customers’ behavior throughout the customer-care journey. Do you force them to scroll through a long menu of choices, or do you use automation to anticipate their needs and reduce their menu options to one or two?

You also can create AI-enabled chatbots to assess exactly what the customer wants. If they want something that the bot can do easily with little risk of failure, then you keep the user within the automated voice interface.

If, however, the customer has a more complex challenge that requires human intervention, then your bot should forward them to the right person. Your customer reps should have all the data they need about the product and the caller to produce a happy outcome.

Why Impact is More Important than Accuracy
With a large enough dataset, a robust statistical model and a well-designed algorithm, you can accurately predict how people will answer the questions listed above. Indeed, you can expend considerable resources producing accurate predictions of their replies.

But you have only so many people, only so much budget, and only so much time to spend on automating your customer care system. Moreover, algorithms can provide uncannily accurate predictions that have no measurable influence on your business.

Thus, you have to think first about impact. Which predictions get customers what they want sooner? Which ones streamline your operations? Those are the kinds of questions your algorithms should answer.

The Smart Way to Automate Your Customer Care Journey

At DMI, we urge our clients to start with the goal of using AI to predict a single useful outcome. You don’t need a moonshot that automates your entire customer care system. You just need a solid foundation to build upon.

Learning to prize impact over accuracy is central to making that happen.

— Niraj Patel, director artificial intelligence