Avto.ru launched a neural network to determine the configuration and cost of a car

Context

According to Avto.ru and CM.Expert, at least one in ten used cars in Russia is sold for well below market value. One of the reasons is that used cars are put up for sale with incomplete information about the equipment.

At the same time, the difference in the cost of basic and luxury equipment within a model with the same engine and gearbox can reach 50%, the service notes. The product from Avto.ru and CM.Expert solves this problem and does it automatically.

How the service works

• User uploads photos of cars, the neural network analyzes them and determines the installed options, such as parking sensors or multi-steering.
• If there is not enough data to accurately recognize the configuration, the wizard may ask 2-3 clarifying questions.
• Based on the options found or the answers received, the equipment will be determined automatically, and then the assistant robot will offer to decipher the missing options and save all the information to the car card.
• After that, another CM.Expert algorithm will determine the exact price of the car.

“The neural network was trained on more than 1 million photos of our customers’ cars with trim levels and options already marked up and the selling price. The trained model can almost instantly recognize dozens of different options,” explained Dmitry Kuzmin, senior product manager at CM.Expert at Avto.ru, the principle of operation.

It should be noted that the service will simplify the work of dealer employees. The tool is available in the personal account of CM.Expert, and in the future all Avto.ru users will be able to evaluate it.