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

Uncle Joe's Makeover Diary 2.0

If you could know something about your competitors, what would you like to know? Please post your thoughts.

Off the top of my head:
  • # of Turns
  • Fastest sellers
  • New Arrivals
  • Biggest investment (YMMs)
  • Inventory by age
  • Local Sales Velocity Rank
What would you like to know, and, will it bring business value to you?
Example: If your sales are slowing, are your competitors? If so, how so?​

Are you using a Market Intelligence Platform now?
If so, which one?
What do you like about it?​
What do you wish it had?​
 
Shadows heading off in different directions give it away. Not so obvious in this example but give it an hour of taking pictures and the last car will be dramatically different than the background. AI can be a little glitchy. There’s a big chunk of the shadow of the hood missing on the ground in front of the car. Shoppers will scrutinize more than we will.

With so many dealers whose buildings are landmarks, placed at major intersections with traffic lights. Familiar sights for their local clientele. Yet their photos are taken in generic plain white plastic booths.

Photography should be front and center in the environment your customers have already familiarized themselves with, while waiting at the traffic light to and from work.

One of the primary drivers of human nature is comfort. Familiarity is one aspect of that comfort. And your building is already a familiar sight to people who have been driving by for years. Uniformity is an another. The same photos in the same familiar place, comforts shoppers into being customers. If fake backgrounds spook your shopper, you’ll never know it, because they don’t click. Or they click away without buying.

In 2018, the senior vice president of Autotrader, stated that stock photography get some astronomical percentage of less clicks because the customer is under the impression that it’s not the actual car. The best photos taken in photo booths look like stock photos.

If the building is situated to the sun, advantageous to photography for two or three hours a day. You should take them during those hours. If not, you should build a landscaped area with maybe a large sign with the dealers logo.

Pull up Photo Ephemeris and find your building. Pick the time of day when the sun is shining on the building and shoot at that time.

Photography is recording light. You should be taking pictures when the light is optimal. Unless you’re after comforting your porters instead.

Keep up to date with that website because those times change a little everyday and before long it will look wrong if you follow the clock instead of the sun.

It’s so easy now to get good content. You really don’t have to fake it.
 
Last edited:
  • :light:
Reactions: Jeff Kershner
The order that vehicle photos are arranged in a photo carousel can make a huge difference in user experience and lead conversion.

We have a preferred "standard" of what photos to take and in which order, but with countless different photo providers and the churn that comes along with that type of entry-level position, the adherence to our preferred standard does not even come close to 100%.

That's why instead of relying on human consistency, we're developing a computer vision model that is trained on our vehicle photography and can automatically re-arrange photos into our desired order, and remove any photos that we do not want (for example: 4 separate photos of each tire tread or a photo of the VIN sticker).

Yet another great use case for AI.

1698782455370.png
 
Photography should be front and center in the environment your customers have already familiarized themselves with, while waiting at the traffic light to and from work.

I tend to agree with this. While the unnatural white background may help the vehicle pop a bit while omitting any distractions, I still prefer natural sunlight with the vehicle staged in from of our dealership.

This time of the year (fall/winter,) due to the suns position/distance and much shorter window of time, is the most difficult for catching the best natural sunlight. shorter window

Screenshot 2023-10-31 at 4.17.30 PM.png
 
@Ryan Everson We use CV (Computer Vision) for image classification too. Another use case for CV is seen in your screenshot:
1698833879742.png

Photographer & Photo Stage Scoring.​

Different stores and different photographers create a wide range of photo quality. CV struggles when it gets poor pics. The cause is usually the photographer and/or the photo lighting (i.e. the photo stage).

I've not tried this, but what if you rolled up CV's "Confidence Threshold" data into a report. Woa metadata could reveal strengths and weaknesses. By photographer, by time of day, by class (i.e. "far-cockpit"), by exterior or interior, and by VIN (i.e. needs photo reshoot?)

A classic DR leading edge contribution, ty Ryan! Good stuff!
 
  • Like
Reactions: craigh
You've gotta $82,000 Used 2023 GMC Sierra 2500HD Denali with 5k miles. It's filled with WOW-FACTOR features, options & packages.
1698849009630.png
You just posted it online, there are 18 like it for sale near you. You want to make it stand out...

You counted 66 features, options and packages on the window sticker, and you want most all of them on your VDP...
1698849496533.png

Question:
How many of these 66 key features are MISSING on
  • Your Site VDP
  • Cars.com VDP
  • CarGurus VDP
  • AutoTrader VDP
What's your guess, how many are missing? 0? 10? 20? 40?
 
Last edited:
@Ryan Everson We use CV (Computer Vision) for image classification too. Another use case for CV is seen in your screenshot:
View attachment 8207

Photographer & Photo Stage Scoring.​

Different stores and different photographers create a wide range of photo quality. CV struggles when it gets poor pics. The cause is usually the photographer and/or the photo lighting (i.e. the photo stage).

I've not tried this, but what if you rolled up CV's "Confidence Threshold" data into a report. Woa metadata could reveal strengths and weaknesses. By photographer, by time of day, by class (i.e. "far-cockpit"), by exterior or interior, and by VIN (i.e. needs photo reshoot?)

A classic DR leading edge contribution, ty Ryan! Good stuff!
Yes, exactly where I'm headed next! Will be pretty easy to grade photographers against our standards and help them get to the point where AI correction is no longer needed (until they leave and we have to start all over with a new photographer :))

And then I'm also working on something called instance segmentation, I can identify where a vehicle is at in the frame to identify and fix photos that were either taken too close or too far away.

Too far away = crop closer
Too close = use generative AI outpainting *if the conditions are right

You can also use it to detect reconditioning and detail issues like dirty seats, door dings, etc. And identify valuable hot ticket features that might have gotten missed when booking the vehicle out - ventilated seat buttons, premium Bose speakers, etc.

Consistent, quality merchandising at scale is difficult to achieve. AI and computer vision is helping us get there. Having quality photos is perhaps the single most impactful thing a dealership can focus on.

Vehicle photos have a trickle down effect that impacts the effectiveness of nearly all digital advertising campaigns - third party marketplaces, Google VLA's, Facebook AIA's, etc.

1698852881786.png