Side project: used-cars fair value index --- thought?

[WickedWizzzard]

Hi all,

This started as a spare-time experiment combining two things I enjoy: data analysis and cars. I have been working on something I’m calling The Motor Index. Basically a model that estimates the fair value of used cars in the UK. It aggregates price signals across multiple marketplaces, normalises for age/mileage/spec, and tries to cut through the noise of asking prices.

The idea is not to replace CAP (Solera)/Glass’s, but to explore a gap that feels quite obvious: independent, transparent, real-time pricing. This could have both B2C and B2B applications.

I have put a lightweight interface online, mainly for fun: themotorindex.com. I would really appreciate PH feedback:

• does this kind of transparency feel useful?
• anything it gets obviously wrong?
• what features or car segments would you expect from something like this?

This version is just a POC and it's not scalable. To take this further, I would need to align with the data owners, either marketplaces (asking price) or finance/insurance (sale value). If anyone here works at (or knows someone at):

• AutoTrader, Motors, Carwow, Cinch, Motorway, etc., or
• a lender / finance broker / underwriting team

…I would be very interested in having an informal chat.

Not selling anything, just exploring whether this could become genuinely useful with the right data partnerships. Happy to take any constructive feedback (and the usual PH honesty).

[fat80b]

WickedWizzzard said:

Hi all,

This started as a spare-time experiment combining two things I enjoy: data analysis and cars. I have been working on something I m calling The Motor Index. Basically a model that estimates the fair value of used cars in the UK. It aggregates price signals across multiple marketplaces, normalises for age/mileage/spec, and tries to cut through the noise of asking prices.

The idea is not to replace CAP (Solera)/Glass s, but to explore a gap that feels quite obvious: independent, transparent, real-time pricing. This could have both B2C and B2B applications.

I have put a lightweight interface online, mainly for fun: themotorindex.com. I would really appreciate PH feedback:

• does this kind of transparency feel useful?
• anything it gets obviously wrong?
• what features or car segments would you expect from something like this?

This version is just a POC and it's not scalable. To take this further, I would need to align with the data owners, either marketplaces (asking price) or finance/insurance (sale value). If anyone here works at (or knows someone at):

• AutoTrader, Motors, Carwow, Cinch, Motorway, etc., or
• a lender / finance broker / underwriting team

I would be very interested in having an informal chat.

Not selling anything, just exploring whether this could become genuinely useful with the right data partnerships. Happy to take any constructive feedback (and the usual PH honesty).

I like it. (but I like the data and stats).

I selected Porsche 911 997 as that is a market that I have looked at multiple times over the years (right at the bottom end price wise).

I was a little surprised at the outcome. i.e. the average price was quite a bit higher than I expected. I think this is because it covers too wide a range in multiple dimensions - i.e. year and mileage (and presumably coupe vs convertible).

I suspect that to be actually useful, it needs to have finer search granularity - i.e. e.g. If I could select a 2-3 year age and a tighter mileage, it would probably give me the number I might expect (~12-13K in this case).

Presumably the fair market value calculation that Autotrader already do is roughly doing the same thing as this ?

[WickedWizzzard]

fat80b said:

I like it. (but I like the data and stats).

I selected Porsche 911 997 as that is a market that I have looked at multiple times over the years (right at the bottom end price wise).

I was a little surprised at the outcome. i.e. the average price was quite a bit higher than I expected. I think this is because it covers too wide a range in multiple dimensions - i.e. year and mileage (and presumably coupe vs convertible).

I suspect that to be actually useful, it needs to have finer search granularity - i.e. e.g. If I could select a 2-3 year age and a tighter mileage, it would probably give me the number I might expect (~12-13K in this case).

Presumably the fair market value calculation that Autotrader already do is roughly doing the same thing as this ?

Thanks for the feedback. Not surprised you went for the 997 first, such a popular market (I have one myself).

You’ve put your finger on the key issue: if you narrow the spec (age range, mileage, body, engine, etc.), the valuation lines up much more with what you would expect. That’s essentially what Autotrader’s (and others) reg-based valuations do.

Where my index differs is that it tries to track the overall market signal, not the “value of a specific car”. So for something like the 997, the “average” includes a wild mix: early 3.6s, later 3.8s, manuals, Tips, cabrios, high-mile vs low-mile, etc. That pushes the centre of gravity up.

There’s a trade-off here between:

• granularity (tight filters)
• noise (too few data points)
• coverage (keeping the history consistent)

Long term, I would indeed like to let users dial in narrower filters (e.g. age bands, mileage bands, body style) and generate a “micro-index” on the fly, but at scale that becomes computationally heavy, so for now I’m tracking the broader market. Wondering if there are uses for such broader indices too. Or whether the demand (use cases), either B2B or B2C, indeed call for more specificity.

[StevieBee]

I like this too!

I have a friend who's attempting to launch a car finance business catering for sub-prime credit customers. Their model includes assessment of cars that their customers wish to buy and I can see this being of use to them, both as part of the application assessment and in terms of asset tracking. They're stuck on the capital financing bit at the moment so not something they'd look at right now but I've pointed him in the direction of your post.

I could also see this benefiting the Classic Car market - asset value tracking and the like.

How do you propose to monetise the concept to a level that makes it worthwhile?

[WickedWizzzard]

StevieBee said:

I like this too!

I have a friend who's attempting to launch a car finance business catering for sub-prime credit customers....

Thank you, I appreciate the interest and feedback. You have picked up on the direction of travel very well. I see a few possible monetisation paths.

On the consumer side (B2C), there is an opportunity to help buyers and sellers make better decisions. Some people want a better understanding of what drives values, which segments are moving, what is a fair price, and how their own car performs. There is a niche but real audience among more informed buyers. A consumer product also creates indirect opportunities by partnering with adjacent businesses such as insurance or finance providers (advertising, referrals).

On the business side (B2B), lenders and insurers have clear use cases. Better visibility of fair value helps with underwriting, residuals management, and portfolio monitoring. Your example of a finance product for sub prime customers is exactly the type of scenario where a consistent valuation framework can add value.

There is also an interesting middle ground (B2B2C). Collectors, funds and syndicates (such as TheCarCrowd). For that audience, the ability to monitor the evolution of an asset-backed portfolio would be valuable.

So far I have focused on the more liquid end of the market (more science than art). I would like to expand into classics and rarer cars, although that requires a different methodology because of data availability (more art than science). The area I find most interesting is future classics. Still relatively common, but already showing signs of appreciation, and they sit somewhere between liquid market and collector markets. The methodology there is also between art and science.

In short, there are several possible directions, and the right one will depend on aligning with data holders and industry partners. That is the next step for this to grow into something meaningful and worthwhile. If you have relevant industry contacts or would like to discuss this directly, I am always happy to chat.

[MustangGT]

Do you mean something like what Autotrader does for trade sales, an indication of value at the price?

[48k]

WickedWizzzard said:

Hi all,

This started as a spare-time experiment combining two things I enjoy: data analysis and cars. I have been working on something I m calling The Motor Index. Basically a model that estimates the fair value of used cars in the UK. It aggregates price signals across multiple marketplaces, normalises for age/mileage/spec, and tries to cut through the noise of asking prices.

The idea is not to replace CAP (Solera)/Glass s, but to explore a gap that feels quite obvious: independent, transparent, real-time pricing. This could have both B2C and B2B applications.

I have put a lightweight interface online, mainly for fun: themotorindex.com. I would really appreciate PH feedback:

• does this kind of transparency feel useful?
• anything it gets obviously wrong?
• what features or car segments would you expect from something like this?

This version is just a POC and it's not scalable. To take this further, I would need to align with the data owners, either marketplaces (asking price) or finance/insurance (sale value). If anyone here works at (or knows someone at):

• AutoTrader, Motors, Carwow, Cinch, Motorway, etc., or
• a lender / finance broker / underwriting team

I would be very interested in having an informal chat.

Not selling anything, just exploring whether this could become genuinely useful with the right data partnerships. Happy to take any constructive feedback (and the usual PH honesty).

Nice idea. I assume you're *cough* scraping data for the moment wheras if/when you want to make this a bona fide thing you'd have to have proper commercial agreements in place to use data officially. At which point... how do you monetise this or at least cover your costs?

Also who is your target audience? For me as a car buyer I'd need more granularity on the data. It's not enough just to know the trend at a model generation level, I'd want to be seeing the trend for the model year level or even specific registration year and mileage. Basically the Autotrader fair value calculator as an app.

[WickedWizzzard]

48k said:

Nice idea. I assume you're *cough* scraping data for the moment wheras if/when you want to make this a bona fide thing you'd have to have proper commercial agreements in place to use data officially. At which point... how do you monetise this or at least cover your costs?

Also who is your target audience? For me as a car buyer I'd need more granularity on the data. It's not enough just to know the trend at a model generation level, I'd want to be seeing the trend for the model year level or even specific registration year and mileage. Basically the Autotrader fair value calculator as an app.

See above for monetization ideas/paths. Indeed, requires agreement with the marketplaces (for comprehensive offer values), lenders, and/or insurers (since they have sold values) for the data.

The challenge with creating these indices for narrower specs (year and mileage) would require backfilling on the fly, which I'm not sure is feasible.

What about having the indices with history for broader markets (make and model), and assuming that's a good proxy for most vehicles in that bucket, then using the more traditional calculators to calculate the value for a specific vehicle? Would that be a sensible compromise for a consumer product?

[Truckosaurus]

There's almost too much data there to be any use to the casual user.

If I was looking up my car I'd just want the basic info on 'what its worth' and then the bit about what the average car for sale is like.

I suspect the biggest competitor for this site is something like WBAC where you can just type in your number plate and get a value that is going to be in the ballpark of what they will actually pay rather than just comparing it to other adverts.

[48k]

WickedWizzzard said:

48k said:

Nice idea. I assume you're *cough* scraping data for the moment wheras if/when you want to make this a bona fide thing you'd have to have proper commercial agreements in place to use data officially. At which point... how do you monetise this or at least cover your costs?

Also who is your target audience? For me as a car buyer I'd need more granularity on the data. It's not enough just to know the trend at a model generation level, I'd want to be seeing the trend for the model year level or even specific registration year and mileage. Basically the Autotrader fair value calculator as an app.

See above for monetization ideas/paths. Indeed, requires agreement with the marketplaces (for comprehensive offer values), lenders, and/or insurers (since they have sold values) for the data.

The challenge with creating these indices for narrower specs (year and mileage) would require backfilling on the fly, which I'm not sure is feasible.

What about having the indices with history for broader markets (make and model), and assuming that's a good proxy for most vehicles in that bucket, then using the more traditional calculators to calculate the value for a specific vehicle? Would that be a sensible compromise for a consumer product?

You might be able to do the mileage based on simple "below average" / "average" / "above average" bands rather than try and get more granular increments?

I can't help but feel your average car buyer (if like me) would want more specificity on the make and model and year than saying "Ford Kuka Mark 2".

Having worked on the software for the finance company of a large German manufacturer I know just how much data we're talking about though! It's a lot to crunch and that's just one manufacturer.

[WickedWizzzard]

Good points. Thanks for the feedback.

[JimmyConwayNW]

As a dealer when I buy I generally look at the values of Cap - Trade and retail. So Cap clean / average etc.

I also use Autotrader values. Trade and retail.

What would this do better for me?