- Apr 7, 2009
- 4,865
- 2,304
- Awards
- 10
- First Name
- Joe
Who's Innovating Without Adding AI?
- Thread starter jon.berna
- Start date
- Dec 19, 2018
- 722
- 682
- Awards
- 8
- First Name
- Bill
If you could figure out a way to order dealers some used cars you would be rich, Joe.new or used?
- Apr 7, 2009
- 4,865
- 2,304
- Awards
- 10
- First Name
- Joe
I build exotic algos for stock trading, its all math based pattern hunting. Should be doable.
Setup: Create a dealer's data lake of 1st party sales history, 3rd party local demand and supply data. Have UCM create a no-go list of YMMs he'd never buy. Build an algo that can dynamically create a 'inventory demand' sheet.
Execute: WIth the 'inventory demand' sheet, go find matches. Have the bot canvas presale auction listings looking for matches... Publish the matches to a watch list page for the UCM, allow UCM to set alert criteria (i.e. set thresholds for the bot to auto-buy w/o UCM's oversight), hand off all high value VINs to UCM live personally.
I'm sure someone's building this.
Claude's bot*.
I wonder how much of this goes into GM's Recommended orders. I'd like to think it takes into account some of this, but then again they are the OEM..On the flip side, here is an AI innovation request: Let me plug in my last 6-12 months of sales and order my units for me according to sales volume. TIA
- Dec 19, 2018
- 722
- 682
- Awards
- 8
- First Name
- Bill
I was thinking what would be necessary to do this. And here's what I came up with:
Note: I'm still trying to catch up with this Ai technology.
I have been heavily using Claude the past few weeks. It doesn't shut down like ChatGPT when you talk like a sailor to it and call it every name in the book. Claude tends to do what they are now calling the bad answers as "Hallucinating", which happens quite often. It will randomly tell you it doesn't have the information but it clearly does.
Here is what I would expect every so often:
Claude:
I see from your past sales history that you should be ordering the latest Saturn because it's in high demand.
Dealer:
No, you need to review the data spec file and be systematic on the decisions.
Claude:
You're absolute right! I need to be more systematic! Let me get provide a better solution. Ah! I see the problem, I wasn't comparing your current year history. Let me pull the next 10 rows of data.
Sorry, I pulled from the previous year. Let me try again.
I see that you are Chevy dealer and the year is 2025.
I recommend ordering 10 more Toyota Tundras with the 5.7 since my research shows that most men prefer the v8 option.
But building this is not really that hard. You do a query of your data in the database and then have the Ai interpret the data for you. If you want to want to pay AI to do the work for you.
If you want to stay on topic:Ai
new cars
Just do a query to get the latest sales and then sort for the most popular, that sell the fastest, and with the higest profit margins.
for used:
Do a similar query and then merge data from your api service. If have to you could scrape a website.
These SQL queries can definitely start looking like exotic algos.
How much can AI help?
Ai can craft you AI sentences with the results.
I wonder if it can build better data results ???
It still comes down to knowing what you want to ask and then being able to know what to filter and sort for.
What is Ai currently?
If my understanding is correct, it is very advanced elastic search (document style of fuzzy searching) within a vector database (a database that is designed to store your "document" in small chunks). It then translates your sentence, "Find me all cars with a 6.4 Hemi engine" into basically the SQL query. "Selct all form database where search word is similar to "6.4 hemi engine" sort asc" this is very very simpliefied and not the exotic queries that Joe mentioned which can be hundreds of lines long cross reference all kinds of data input.
Spealing of paying.
Ai is like an API call be requires a lot more hardware. It's tied to the more powerful hardware. Think of like trying to mine bitcoin on your own hardware versus on commercial grade hardware in a huge warehouse.
What about those API calls tha dump your crm data or what not. You are paying for vin etching on your windows. I can't think of a better comparison. Yes, there is a bit of work to create it but once it's made, it works like any other webpage that shows data from a database. You are payign to be allowed to see data with out going onto that services website.
Note: I'm still trying to catch up with this Ai technology.
I have been heavily using Claude the past few weeks. It doesn't shut down like ChatGPT when you talk like a sailor to it and call it every name in the book. Claude tends to do what they are now calling the bad answers as "Hallucinating", which happens quite often. It will randomly tell you it doesn't have the information but it clearly does.
Here is what I would expect every so often:
Claude:
I see from your past sales history that you should be ordering the latest Saturn because it's in high demand.
Dealer:
No, you need to review the data spec file and be systematic on the decisions.
Claude:
You're absolute right! I need to be more systematic! Let me get provide a better solution. Ah! I see the problem, I wasn't comparing your current year history. Let me pull the next 10 rows of data.
Sorry, I pulled from the previous year. Let me try again.
I see that you are Chevy dealer and the year is 2025.
I recommend ordering 10 more Toyota Tundras with the 5.7 since my research shows that most men prefer the v8 option.
But building this is not really that hard. You do a query of your data in the database and then have the Ai interpret the data for you. If you want to want to pay AI to do the work for you.
If you want to stay on topic:Ai
new cars
Just do a query to get the latest sales and then sort for the most popular, that sell the fastest, and with the higest profit margins.
for used:
Do a similar query and then merge data from your api service. If have to you could scrape a website.
These SQL queries can definitely start looking like exotic algos.
How much can AI help?
Ai can craft you AI sentences with the results.
I wonder if it can build better data results ???
It still comes down to knowing what you want to ask and then being able to know what to filter and sort for.
What is Ai currently?
If my understanding is correct, it is very advanced elastic search (document style of fuzzy searching) within a vector database (a database that is designed to store your "document" in small chunks). It then translates your sentence, "Find me all cars with a 6.4 Hemi engine" into basically the SQL query. "Selct all form database where search word is similar to "6.4 hemi engine" sort asc" this is very very simpliefied and not the exotic queries that Joe mentioned which can be hundreds of lines long cross reference all kinds of data input.
Spealing of paying.
Ai is like an API call be requires a lot more hardware. It's tied to the more powerful hardware. Think of like trying to mine bitcoin on your own hardware versus on commercial grade hardware in a huge warehouse.
What about those API calls tha dump your crm data or what not. You are paying for vin etching on your windows. I can't think of a better comparison. Yes, there is a bit of work to create it but once it's made, it works like any other webpage that shows data from a database. You are payign to be allowed to see data with out going onto that services website.
In a perfect scenario, what data should we include for creating a new car stocking strategy. This is the tough part in my opinion.
Is the past 12 months of sales an indicator of what will sell in the future? Maybe, maybe not.
What about VDP counts on your website? maybe, maybe not.
What about past XXX amount of days of market sales in your area?
What about search demand from classified sites?
What about taking into account over or under supply of a model / trim, etc. in a market to either chase volume or fill a supply gap?
Am I overthinking it? Maybe, maybe not.. haha
- Dec 19, 2018
- 722
- 682
- Awards
- 8
- First Name
- Bill
How well do the humans doing the ordering do at considering these very good points in their ordering decisions?
- Jul 21, 2020
- 18
- 7
- Awards
- 2
- First Name
- Richard
Good idea, difficult to execute. You would need some sort of reference to the market dynamics you were dealt at that timestamp, a lot of what @jscole86 mentioned. Hurts my head thinking about it.
I learned just yesterday that companies exist just to do dealer stock ordering and my first thought was "okay... but what are they getting access to and looking at to even make that decision"? What does this service cost?
Sort of did this a while ago for some dealers running our custom order tool to help understand what trims, engines, colors being selected. But we were all in a vacuum at the time.
I learned just yesterday that companies exist just to do dealer stock ordering and my first thought was "okay... but what are they getting access to and looking at to even make that decision"? What does this service cost?
Sort of did this a while ago for some dealers running our custom order tool to help understand what trims, engines, colors being selected. But we were all in a vacuum at the time.
I fully agree! I guess it was more of a question to you. If this solution was to be created, what data do you think should be factored in to the recs?
and yes, it hurts my head thinking about this too.
