Delivering AI-First based system for OEMs and dealerships

The key goal of the Bright Box team was to create a reliable model for forecasting the probability of car owners’ visits to dealerships for service. This information is very important for dealerships, as most car owners do their first scheduled service only and never return again. For each vehicle, Bright Box used data collected up to a particular date that allowed them to assess the probability of two events: a car visiting any dealership within six weeks (particular date + 42 days) or a car visiting a particular dealership for service.

The team of Bright Box researched and evaluated segmentation drivers by their driving style and vehicle usage patterns in order to offer drivers from different categories a certain set of products and services: appropriate accessories, new car sales or an upcoming service visit.

For instance, if a customer has a record of driving aggressively at night, it may make sense to offer them new tires optimised for these particular conditions or 19” wheels. If a customer has relocated to the countryside, it might be appropriate to offer them a test-drive of a new, more suitable crossover. The Remoto AI can forecast the next service visit. It works based on the driver’s behaviour, on the ‘look-alike’ principle with very high probability. Therefore, a typical consultant at a dealership works with an average of 100 customers. Bright Box enables dealerships to provide a personal consultant to every customer in the form of an intelligent application. This way, a dealership can increase sales and service quality by 100 times or more.

The quality of provided services depends solely on the amount of collected data. Bright Box has been collecting data for over five years. Each connected car unit from Bright Box generates up to 8.7 million records (status records) - over 2.3 million car status entries on the average. As aresult of this, the company has accumulated over 1.5TB of connected car data from users from different countries.

Therefore, Bright Box have created Data Lake on the basis of Microsoft Azure, where terabytes of data gathered from cars are stored in a structured fashion. Upsell forecasts are based on the data from Telematics (mileage, GPS, speed, RPM), as well as from CRM, DMS or ERP systems (customer ID, date of the transaction, breakdowns by work/materials).

It can bring so-called ‘insights’ to light based on this data. For example, Bright Box can calculate approximately 100 indicators that company consider important: the portion of driving time spent at speeds of over 80mph, how frequently the driver steps on the gas, where they go most frequently in an urban setting or in the country, etc.

Data is constantly being accumulated and analysed from all available sources. For every user, we have a certain number of indicators that characterise only that user. It should be total duration of trips for the period, average part of trips with engine speed, part of trips in which right rear passenger door was opened/closed, part of trips with distance of the trip in range [5; 10]km, total number of trips for the period, part of trips in which trunk was opened/closed (in total number of trips), average part of trips in the mornings (6.00 - 11.00), and 120+ insights more.

We can then determine the likelihood of a response to an offer from a dealer based on every driver's indicators. If this probability reaches a certain value required to trigger a marketing or service campaign, a push notification is sent to the user's email or phone.

The car owner receives a minimally-interactive message. If the customer is interested, Remoto AI offers to send them an application, open a chat window, or schedule a return call from a dealer for further communication. The customer's reaction affects the evaluation of the probability of their responding to similar offers in the future.

AI-based algorithms detect driver behaviour changes in order to offer them new cars or accessories. Now Bright Box has offered this AI solution first for their customers. Remoto AI can be connected to any available Connected Car solution, and get the same results. Implementation of a new solution takes around a month. In 2019, the company plans to influence car manufacturers and their processes in the sense letting them know that a system or indicator they supply is not used by drivers.