D.R. Truth - ask him anything!

[jon.berna]

You had the axes swapped, I should've included in the original chart.

That means these were higher inventory stores with poor engagement months. For this chart we only grabbed this year. We are going to run it for the last 6 months and add in some additional data points to see other influences to engagement

 

[JayKelly]

Thanks for answering my question DR Truth, I've seen this correlation a lot, it's nice to see it in a study!

100 units ... Avg Engagement = ~225
200 units ... Avg Engagement = ~250
400 units ... Avg Engagement = ~310
800 units ... Avg Engagement = ~420


Smaller Store Takeaways:

• Marketing: Smaller stores without any inventory advantage (i.e. scarcity or Velocity) are doomed to attract grinder shoppers (who'll visit 3,4,5 stores)
• Sales: What a bi-polar world you're in. You've got loyal hyper-local shoppers (who know your store's name), and then there's everyone else. They don't know your store, they're a friggin battle.

Big Store Takeaways:

• Marketing: You've got it all. A big ad budget, a deep product selection... the perfect setup for a 1stop sale.
• Sales: Sales come easy here, even order takers will make a great living (if the GSM doesn't ax you .

To the victor goes the spoils... Big stores have sooooo many advantages. This is just the tip of the ice berg.

I would like to see these correlations be market specific. If you are a small store in a growing market vs a small store in a market with flat population growth, the stories will be totally different.

 

[joe.pistell]

You had the axes swapped, I should've included in the original chart.

I had to guess what the axis was, this seemed like a far less likely layout to me, wow! now this presents a very different story indeed!! Can you flip the axis's?

Look at the lowest qwebs, say under 50, fascinating.
Look at that amazing qweb range for the > 400 units.

Can you define what qweb is?

 

[jon.berna]

Qweb is an advanced user segmentation in GA that anyone can create. Ours is based on time on site > 400 seconds and page views > 7 and sessions > 4.

We also add two additional segments for used and new (Qweb New and Qweb Used). For these we add another condition that they view a new VDP or a used VDP based on the website provider's corresponding VDP logic.

For the next review I expanded the range from the previous one and did some different correlations. Note how Used in stock and Qweb Used have the highest correlation. This to me makes complete sense.

 

[jon.berna]

I would like to see these correlations be market specific. If you are a small store in a growing market vs a small store in a market with flat population growth, the stories will be totally different.

To figure that out population growth is available from census.gov. I have spent an easy 500 hours over the years learning about geo coding from FIPs codes to census tracts and all the weird nuances within state and local demographic data. I recall trying to mirror the way Mazda tracks service retention. It was so hard to figure out, I had to convert each of our customer's address to their census tract (AORs in dealer agreements are based on census tracts not zip codes) I used Texas A&Ms free tool http://geoservices.tamu.edu/Services/Geocode/. Mazda wouldn't tell us who was in the AOR that we had to keep vs those that weren't included in the calculation. I was able to target those in our AOR with marketing and calls and essentially beat the flaw in the system. I was pretty proud, even though not a single person knew what I did. I did try to explain to the Mazda reps....

Back on your question, although population growth is a factor to a market there is a statistical tool to determine how much of a factor it is. The process to determine a factor's impact on a variable is called a regression analysis. The dependent variable in this case is the number of users that meet an engagement condition (Qweb). The independent variables are the variables we want to test. This could be inventory count, vdp count as well as population growth rates. My gut and experience with these tell me population growth rates would be lower on the regression analysis versus other metrics that are closer to the dependent variable.

We do this at driven data to understand what truly impacts our data specifically in predictions. Here is an example:

This is a regression analysis of the effect weather has on showroom traffic (full study here). You will see we also included weekday. As you can see weekday has over double the impact on showroom traffic than any other independent variable. Regression analysis is a very simple yet powerful way to focus on the things that truly matter.

 

[joe.pistell]

Jon,
Plz explore qweb (dealer web site engagement) and sales activity.

1. Is there a relationship between qweb score (i.e. shopper engagement) and future sales activity?
2. Are there events that trigger higher correlation (i.e. a store's qweb increases past some threshold & sales follow x days later), or, do sales trends follow qweb trends week over week/ MoM?

Have you done a sales -vs- qweb study on stores who's qweb has made a significant increase/decrease?

Example:
If a store has a qscore of X and commits to engagement over lead gen, what is the likelyhood they'll they see a lift in sales?
If there is a correlation, is there a time lag (i.e. qscore rises in week one of january, sales rise X weeks later)
​

Great stuff!

 

[joe.pistell]

We all speak about the unintended consequences of a dealer's obsession to generate more and more leads, leads, leads.

Is there evidence of 'unintended consequences' of high lead gen rates? Is there a relationship between qscore and lead/session ratio? IOW, do sites with high lead/session ratios have low qscores?

 

[joe.pistell]

...Note how Used in stock and Qweb Used have the highest correlation. This to me makes complete sense.

Fascinating!
So if USED units are sticky(er) to car shoppers (than new), is there an unseen benefit to a new car store that has a strong used car presence? Do you have evidence where a dealer significantly increased its # of used units? And if so, what happened to qweb?

A few tag-along thoughts...
Is there a way to measure how much shopping happens in
-New cars only
-Used cars only
-Both new & used

Does the 'New to Used Ratio' influence qweb? Or, is there a threshold of the number of used units alone that causes a pop in qweb?


Wow Jon, this is a look into how inventory mix impacts engagement!! This is powerful stuff!!

 

[jon.berna]

Traveling next two days will try to review when i can!

 

[joe.pistell]

I've got money down that says "avg days in stock and qweb (used and new) are highly correlated".