Can a used car price be predict using machine learning?
Buyers, car fleet managers, used car dealers and car insurers have to evaluate the ‘right’ price for a used car. More than the right price, they thrive on identifying the car depreciation because it impacts all subsequent decisions: should the owner sell the car? What is the replacement value of the car for the insurer? Should the car be sold by the car fleet manager?
Actual used car price is a key topic for these stakeholders. However, today’s online calculators mostly estimate the price of a used car based on age and on retail price of the vehicle. Other depreciation calculators use straight line depreciation (the simplest method). Even though this approach might be interesting to calculate the average price of a cohort of similar vehicles, it does not take into account the differences among similar cars. Another approach consists in consulting an expert for his/her appraisal of the vehicle, which is accurate but costly. And last but not least, it is possible for the savvy buyer/seller to check the listings of similar vehicles both online and in physical stores. However, this strategy is very time consuming. Using predictive models based on available data can help these different stakeholders better estimate their selling and buying prices for used cars in an affordable and quick way.
Car fleet managers, insurance brokers, used car buyers or sellers, can all use predictive models to estimate the best resell price for used cars. The idea is not to get an absolute selling price which would not be negotiable. Quite the contrary, the point by using a precise used car estimator is to get for used cars, the equivalent of a listing price for new cars: a non-disputable starting point for discussions. After this step the negotiations can begin !
However, stakeholders are usually not data scientists and often don’t have the right skills in machine learning nor in software coding to build predictive models.
However data from previous car sales by organizations, or online are available. These data often include such elements as: car age, retail price, mileage, brand, model, fuel type, color, location, transmission type. Professionals have often gathered information about a few hundred (at best a few thousands) used car sales. It is considered a Small Data configuration that traditional Machine Learning tools don’t handle well. They need millions of Data (aka Big Data) for training rather than a few thousand.
In this context, a Small Data-designed machine learning tool such as TADA helps car fleet managers, car insurers, used car dealers build predictive models based on their dataset. It empowers them with a tool to estimate, very quickly, the equivalent of a listing price for used cars. No data science background is necessary to use TADA, MyDataModels solution for Small Data. These professionals can use their own data as they are. There is no need for complex data preprocessing. They can create predictive models in a few minutes and a couple of clicks on a standard laptop.
MyDataModels brings a self-service solution for the professionals who have Small Data.
In 2018, in the UK, 2.7 million new cars were sold when 7.95 million used car changed hands. This is a one to four ratio. The market for used cars is massive.
“Predictive model can optimally estimate a used car price”
Current approaches are basic and calculate a rough average which does not account for the specificities of a particular used car. Instead of these outdated solutions in order to get a listing price for a used car, stakeholders can use a Small Data predictive solution such as TADA.
TADA predictive models allow car fleet managers, car insurers, used car dealers to optimally estimate and price their used cars.