Strive worked with Aible to clearly define the use case, understand the business goals and customer contact center capacity constraints, set timelines, and identify what data Strive Group would provide to train AI models. The AI would be trained on 60,000 Strive service call records collected over a 3-month period.

Aible enriched Strive’s data to gain additional insights and add value. By leveraging postal code information showing the relative locations of customers and a targeted dealership, Aible found that owners of specific vehicle brands and models were more sensitive to distance from the dealership than others when it came to booking a service call. Further features were added to represent absolute date fields as relative periods - for instance, the actual date a vehicle was registered has little impact on the models, whereas the age of the vehicle is a significant factor. Likewise, the date of the previous service is largely irrelevant but the time elapsed since the previous service is not. A custom blueprint was developed based on business assumptions and cost-benefit tradeoffs unique to Strive. This blueprint can be used by Strive for future model training and as a foundation for additional AI projects. More than 400 predictive models were trained and sequenced enabling dynamic and rapid scenario planning at the subgroup level. Over 20 unique model driver variables were identified, such as whether service is overdue, how many days until servicing is due, days since last service, days since last visit, and vehicle age. These drivers were ordered by significance at the field and value level to determine which attributes are more likely to result in a successful booking. Multiple model drivers were identified across different dealership types that have a significant impact on whether a customer will make a booking.

Strive put Aible predictions to the test. They gave the scored customer list to their customer contact team and had them use the predictions to guide their calling strategy. In just one week with Aible predictions, Strive saw a gain of 2.5% in positive bookings, yielding estimated annual savings of £200,000.Larger potential gains were identified for possible future calling strategies. For instance, Aible found that 29% of Strive’s calls were being made to just 12% of the customers, and recommended a different mix based on dealer segmentation. The test demonstrated that a negative prediction from Aible was a good indicator of customers that would need repeated calls and yet still not make a booking. Using the predictions to selectively reduce the number of calls could result in saving about 1,275 calls per week and freeing up agents for other duties.