By Eliron Ekstein
It is often claimed that the vehicle grading system should be overhauled due to lack of objectivity. In reality, the best way to introduce such objectivity is to use technology that removes the total reliance on human damage disclosure.
If you think selling and buying used cars is difficult for the ordinary person, consider the life of a used car salesperson. They fight fluctuating demand and supply, uncertain pricing and need to clear inventory in time otherwise their financing costs go through the roof.
Used car dealers buy hundreds and sometimes thousands of units every month.
How do they make quick decisions on which used car to buy?
The industry has come up with condition grading methods (most common in the US is called ‘AutoGrade’ which is accessed through platforms such as AutoIMS), that sum up the vehicle condition on a scale of 1 (total wreck) to 5 (mint condition). The inspector who writes up the Condition Report (“CR”) on the vehicle, is supposed to disclose all damages and issues, submit them and receive back a score that is supposed to be objective.
But the grade varies depending on the quality of the inspector and the purpose of the report. If the inspector is hired by the seller, you can expect it to be a bit higher. Many sellers (and buyers) complain that a single vehicle can be regarded as 4.2 and then get to the auction to achieve a dismal 3.8. The swing in value could mean hundreds or thousands of dollars each way - causing a lot of pain for those who live on this slim margin per car.
Many are calling for a change in the methodology. At Used Car Week, which I just returned from in Vegas, we heard players like ACV Auctions suggesting there was room for multiple grades for different parts of the car; a more nuanced grading would give additional transparency. Others would like to scrape the grade altogether, forcing buyers to take a harder look at the condition report details.
My own view, which I've expressed during the panel on damage detection, is that a grade is only one tool to assess a car’s condition. The CR should include necessary details on damage: multiple images, from various angles; data indicating mechanical and electrical health, and more.
Most importantly, the data feeding into the grading systems should be as objective as possible. With artificial intelligence, we are no longer fully relying on an individual inspector deciding what to disclose and what not.
It is already possible today to take a video of the vehicle, which is a great baseline. The tech then needs to analyse it in a scalable way to turn it into a digestible condition report. Finally, a strong data platform should support millions of condition reports to identify most common issues and pitfalls. It is only through methodic study of vehicle condition over time, will we gain the insights and trends allowing us to point to deviations and biases.
Our own internal effort at Ravin.ai led us to connect the inspection AI in our Inspect app today, with a condition grade that comes in very close to what a ‘perfect human’ would report. This means that, for the same number of damages observed on the vehicle during a routine inspection, the AI can produce the right assessment which is very difficult to dispute given the visual evidence. In the development of this algorithm we have analysed numerous condition reports performed in and outside of our system, in the real world.
Tune in as Dr. Alex Kenis, Ravin’s Chief Scientific Officer, will share more details from this research soon.
We invite you to ping us at email@example.com and learn more about this research and more.
We are on a fascinating journey to automate inspections, which will translate to a much more objective and reliable condition score.