Leading with AI and Analytics: Build Your Data Science IQ to Drive Business Value by Eric Anderson & Florian Zettelmeyer

Author:Eric Anderson & Florian Zettelmeyer [Eric Anderson]
Language: eng
Format: epub
Publisher: McGraw-Hill
Published: 2020-11-23T00:00:00+00:00

Influencing a Future Business Outcome

In the previous section, we built a predictive model to forecast future revenue. Now, we want to focus on a different problem: determining how to influence (not just anticipate) a future business outcome by using analytics for “what-if” analysis or scenario analysis. Here, we are in the domain of causal analytics. In a previous chapter, we illustrated how one could use an A/B test or a planned quasi-experiment to uncover a causal relationship. Now, we’d like to use our existing data—not planned data—to achieve this goal.

When applying data science to your business, many of you are going to be in precisely this situation. You want to improve the business by influencing a future outcome; but you may not have the luxury of a planned experiment or quasi-experiment. Here, we give you a framework for how to get causal insights from historical data.

To illustrate these points, let’s return to the OLA example.

When CFO Juan Rios shared the success of his predictive cancellation model with the CEO and senior leadership team, there was a buzz in the room. The senior leaders had talked about using analytics before, and there had been anecdotes of success stories within OLA, but Juan’s was the first real success story to make it into the C-suite. Inspired by Juan’s success, CMO Ashwin Arora decided to tackle a related problem. If OLA now understood factors that predicted customer cancellations, was it possible to influence their behavior? In other words, what interventions might prevent or reduce cancellations?

Ashwin pulled his marketing team together for a brainstorming session with head of data analytics Julia Simpson and her team. The data scientists explained how the predictive model was able to anticipate cancellations, and then presented the top five variables associated with cancellation. These included how airfare was booked, whether the overall booking was direct or via a travel agent, price paid, size of travel party, and customer loyalty status. Ashwin then turned to his team and said, “I want you to come up with the best ideas to reduce cancellations.”

Ashwin’s team, working with the data scientists, spent the next hour brainstorming ideas to reduce cancellations. There was no shortage of ideas. One idea was to offer a $100 bonus to travel advisors if they persuaded consumers to book their airfare through OLA when consumers booked their trip. Another idea was to identify high-risk bookings and offer discounts on trip-related activities, such as excursions and special events. A final idea was to offer customers higher loyalty status on trips with high cancellation rates, which would lead to incremental benefits like private dinners and premium drinks.

This process at OLA illustrates a second value of a predictive model beyond anticipating future revenue. Ashwin’s team used the model to ideate, or come up with, new business initiatives. Without a model, Ashwin might simply have assembled his team and asked them to come up with new ideas—a classic brainstorming session. With a predictive model, the aperture of ideation changes. The factors highly associated

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.