Pub. 1 2019-2020 Issue 4

Issue 4 2020 23 WVADA TransUnion to help you see industry trends and identify high-quality leads. • Determine when prospects are most likely to buy a vehicle by applying pre - dictive modeling. • Develop personalized sales scripts. Scripts are another way to apply predictive modeling because the modeling can tell you why potential customers are interested in buying a vehicle. The driving force is the data, and more particularly, the predictive modeling that results from analyzing that data. What should you know about predictive modeling? • It has multiple names that include predictive analysis, predictive analyt - ics and machine learning. Of the four terms, the one that has beenmost pop- ular on Google since 2004 is machine learning. Excluding machine learning, the general and most commonly used name is predictive modeling. • Commercial applications often use the term predictive analytics, but academic settings also favor predictive analytics. • Machine learning is technically consid - ered to be different from predictive modeling because it relies on statisti - cal techniques so computers can build predictive models. In practice, the two terms are often used interchangeably. • Machine learning is part of the devel - opment of artificial intelligence. The valuable thing about predictive modeling is that it can challenge key as - sumptions. People see what they expect to see, and sometimes they are blind to what the data tells them. Being able to consider future events and evaluate out - comes is as helpful as it is difficult. Still, predictive modeling can increase accu - racy and help a dealership maintain a competitive advantage because of those unexpected insights. Is predictive modeling perfect? No. It tells you likely outcomes, but sometimes the outcomes are anything but likely. Experts thought it might be time for a recession before the Shutdown Recession, but they didn’t expect it to result from a pandemic, and no one expected the severity. Election outcomes are also difficult or impossible to spot in advance; Nate Silver has been on both sides of that difficulty. But even an imperfect assessment can put you ahead of other dealers, and the quality of those assessments will improve over time. The data used to feed predictive models comes from a variety of sources: • Customer Relationship Management (CRM) • Customer service • Demographic information • Digital marketing and advertising • Economic information • Geographical information • Machine-generated data from telemetry or sensors • Surveys and polls • Transactions • Web traffic Of course, the trick is not just modeling data; it is modeling data into relevant information for your specific strategic business goals. Equally important is evaluating the accuracy of a model in hindsight. Youdon’t knowhowsuccessful something is until you measure that success. Those measurements can then help you to improve the system and get better results with the next model. Pay attention to how the model includes the following areas: • Benchmark analysis • Data gathering and cleansing • Goal evaluations and key performance indicators (KPIs) • Action plan development • Plan execution • Process improvement When using data to sell cars, remember that the process is part of the same process that will someday bring us autonomous vehicles. We are in the early stages of an AI revolution. There’s plenty of room for growth, but predictive modeling has al - ready provided dealerships with new tools to help themsucceed inways nobody could have anticipated a few decades ago. t We are in the early stages of an AI revolution. There’s plenty of room for growth , but predictive modeling has already provided dealerships with new tools to help them succeed in ways nobody could have anticipated a few decades ago.