Dealer feedback - finding hidden mechanical problems before they impact profitability

[Sorel]

Hi All: Looking to get input from dealers! We have been collecting data for many years and have trained an AI model that analyses the acoustic sound data of engines. Multiple applications, but the one we wanted feedback on was using it on trade appraisals at the dealership. The ultimate goal is detecting mechanical failures that OBD scanners, visual inspections, and appraisal software miss. Ultimate goal: a more robust way of finding hidden mechanical problems before they hit the bottom line.

Both OBD data and the engine sound data are incorporated into an algorithm - and gives a "FICO" score on the mechanical health of the vehicle based on these data points. Takes 3 mins once the process is learned. An analysis of nearly 10,000 of the vehicles scanned showed that the AI detected that there are engine or transmission issues on a good % of vehicles where no OBD codes are seen. A LOT!!

Thoughts on using this for trade-ins?
How it would it fit into your process...get used in your appraisals and numbers put on trades etc?
Special use cases - buying centers?
Marketing your used cars as passing a more rigorous check than your competitors? (-:

Thank you!

 

[Vochi]

Hi All: Looking to get input from dealers! We have been collecting data for many years and have trained an AI model that analyses the acoustic sound data of engines. Multiple applications, but the one we wanted feedback on was using it on trade appraisals at the dealership. The ultimate goal is detecting mechanical failures that OBD scanners, visual inspections, and appraisal software miss. Ultimate goal: a more robust way of finding hidden mechanical problems before they hit the bottom line.

Both OBD data and the engine sound data are incorporated into an algorithm - and gives a "FICO" score on the mechanical health of the vehicle based on these data points. Takes 3 mins once the process is learned. An analysis of nearly 10,000 of the vehicles scanned showed that the AI detected that there are engine or transmission issues on a good % of vehicles where no OBD codes are seen. A LOT!!

Thoughts on using this for trade-ins?
How it would it fit into your process...get used in your appraisals and numbers put on trades etc?
Special use cases - buying centers?
Marketing your used cars as passing a more rigorous check than your competitors? (-:

Thank you!

A couple quick thoughts...

1. How are you validating the AI's predictions? Of the vehicles where the system flags an engine/transmission issue but there is no OBD code, what percentage were confirmed by a technician or repair?
2. How does the output look? Is it simply a score, or does it identify the suspected system/problem and provide a confidence level?
3. How much additional time does it add to the appraisal once the process is learned? Three minutes sounds good, if that's the true end-to-end time.
4. Can the data be retained with the VIN?

Can you share real-world examples of vehicles where the OBD scan was clean, the vehicle otherwise passed a normal appraisal, the AI flagged something, and the problem was later confirmed.

 

[joe.pistell]

The ultimate goal is detecting mechanical failures that OBD scanners, visual inspections, and appraisal software miss. Ultimate goal: a more robust way of finding hidden mechanical problems before they hit the bottom line.

Genius concept, dealers will be skeptical (it's in their DNA). They won't use it as a standalone product; it must be nested within an existing appraisal tool and serve as one part of the appraisal process. If this process is statistically valid, it would be a great tool for auction condition reports.

 

[DanMorgan]

 

[Burt Lewis]

Very interesting concept. I can see this having value beyond just trade appraisals.

One use case that immediately comes to mind is giving a buyer a tool they can run right from their phone at an auction, when evaluating a trade, or when a vehicle first arrives at the dealership.

Ideally, the process would be simple: plug in a Bluetooth OBD device, have the phone collect the OBD data, record the engine/mechanical sounds for a couple of minutes, and have the AI combine both into a mechanical health/risk report.

For our buyer, something that ultimately says LOW RISK / INSPECT FURTHER / HIGH RISK, along with what it detected, could be extremely useful before committing thousands of dollars to a vehicle.

The acoustic side is what really interests me. If you're actually detecting developing engine or transmission problems that aren't throwing OBD codes yet, that's potentially very valuable.

Do you currently have an API or SDK that would allow your acoustic analysis to be incorporated into a dealer's own mobile vehicle acquisition/appraisal application?