• Stop being a LURKER - join our dealer community and get involved. Sign up and start a conversation.

Using AI To Purchase Wholesale Inventory At Auction

bringles

Skate Alert
Apr 17, 2013
51
15
Awards
2
First Name
Erik
Hi,
I am hearing a lot about using AI during the sales process but has anyone heard of any AI software solutions to help with acquiring inventory at auctions?

I have searched online but can't find anything. I imagine some company is already working on this.
 
  • Like
Reactions: Tallcool1
Hi,
I am hearing a lot about using AI during the sales process but has anyone heard of any AI software solutions to help with acquiring inventory at auctions?

I have searched online but can't find anything. I imagine some company is already working on this.
To a degree, that is what vAuto does with some of their product lines. I can't remember the names of their products, but someone in here surely knows.
 
We built a tool for this at CarStory years ago. That was one of the reasons the company was acquired by (starts with V rhymes with broom).

Couple free thoughts for you, so assign exactly that much value to them.
  • Market demand and market value aren't equivalent to YOUR demand and YOUR value.
    • We saw this a bunch. One of the productizations of the data was built to help dealer groups trade used units amongst their stores to maximize retail profits. The same VIN had different demand and value indicators at different lots. Everyone knows this is true regionally, no need for a cold weather package in Miami for example, but interestingly enough it was true hyper-locally too.
    • Point is simply to take the "market demand" data with a grain of salt. The tool we built took a patented similarity engine and bounced the VIN against previously sold units to determine its value to the store.
      • This was a great problem for a computer with no capacity for emotion. "How did a similar unit perform" is not a question to answer with feelings and fond memories. You need to incorporate PDOL, market scarcity, feature demand/value, color preference by consumer search, etc.... it's a data question.
  • Features drive significant demand
    • Another obvious one, but worth mentioning.
    • If I were building an AI tool for acquisition, I would want CV to confirm the listings data. Automotive data is inherently dirty. You can't trust that the features, trim, and even trans in a listing are accurate, build data doesn't account for modifications and dealer adds. CV provides a level of visual inspection to verify the listings data.
    • What I would REALLY want is CV to alert me to high demand/value options that were MISSING from the listing. Those would represent higher profit potential than a simple market demand index.
  • CV for damage
    • This is another gimme, but not easy.
    • CV for damage is getting better day over day. I'd put this in the must have category to justify a SAAS acquisition tool.
 
Here is an article on AI for disposition. Data and analytics in the driver’s seat of the used-car market

For acquisition, you can upload local and regional auction data, including pricing, sales %, etc., virtually any data point. Use a code interpreter, and it will give you a visual output. If you understand scripting, you can call this data nightly. Very powerful and FREE. For more robust outputs, you might need the paid versions of LLama or ChatGPT.
 
Hi,
I am hearing a lot about using AI during the sales process but has anyone heard of any AI software solutions to help with acquiring inventory at auctions?

I have searched online but can't find anything. I imagine some company is already working on this.
Over the past couple years I've been searching for a better solution to acquire inventory and haven't came across anything better than what already exists. I've also played around with the idea of AI in assisting the process and I've come to the conclusion that until a couple factors change, AI will not be a contributing factor.
CRs are way too inconsistent for an AI to generate a recon value based on number grade or announcements alone. This issues has been exacerbated over the past year with quality inventory becoming harder and harder to find, while sellers are pushing harder and harder for better grades.
Existing vehicle data isn't accurate enough for AI to formulate a price based on value-added options. VAauto is notoriously terrible for options not showing value vs the average. This is due to a variety of factors but my feeling is that most of them come down to improper data entry from dealers.
It all essentially comes down to garbage in, garbage out. My best guess is that AI could one day assist with this but not with our current data practices/options.
Carmax has used a machine learning algorithm for years to assist their buyers in sourcing inventory and it's quite effective. But it's effective because they use all of their own proprietary data (sales numbers, value added options, down to the trim and packages). One could build a model off of this but it would require TONS of clean data that I simply don't think exists outside of a company like Carmax.
 
Over the past couple years I've been searching for a better solution to acquire inventory and haven't came across anything better than what already exists. I've also played around with the idea of AI in assisting the process and I've come to the conclusion that until a couple factors change, AI will not be a contributing factor.
CRs are way too inconsistent for an AI to generate a recon value based on number grade or announcements alone. This issues has been exacerbated over the past year with quality inventory becoming harder and harder to find, while sellers are pushing harder and harder for better grades.
Existing vehicle data isn't accurate enough for AI to formulate a price based on value-added options. VAauto is notoriously terrible for options not showing value vs the average. This is due to a variety of factors but my feeling is that most of them come down to improper data entry from dealers.
It all essentially comes down to garbage in, garbage out. My best guess is that AI could one day assist with this but not with our current data practices/options.
Carmax has used a machine learning algorithm for years to assist their buyers in sourcing inventory and it's quite effective. But it's effective because they use all of their own proprietary data (sales numbers, value added options, down to the trim and packages). One could build a model off of this but it would require TONS of clean data that I simply don't think exists outside of a company like Carmax.
MMR is starting to include option packages, that could make a significant difference in more accurate pricing. When option packages are factored into valuations, it gives a clearer picture of the vehicle's true market value, which should help with more informed purchasing decisions.

If there is a tool or update that has been launched since October 2023, it could definitely make a substantial impact on how inventory is bought, priced, and sold. A solution that integrates MMR data with real-time market conditions, option packages, and other key metrics like vehicle condition and regional demand could provide a big advantage.

Have you come across any new tools or platforms recently that address this? If not, it would be worth exploring what’s available to see if someone has launched something innovative to improve pricing accuracy. A tool like that could be a game-changer in helping dealerships make more informed, data-driven purchases, optimizing their inventory management and profitability.
 
  • Like
Reactions: joe.pistell