Tag: VSC

The Case for D2C

A while back, I wrote a survey of Direct to Consumer VSC Sales.  This was a “how to,” and today I am writing about the “why.”  The short version is that D2C is a large and unserved market.  Franchised dealers sell service contracts with 47% of new vehicles, which is great, but that leaves the other half unprotected. 

Add 6% to reported F&I gross, plus 4 to 5 times that amount for the backfile

Depending on which “touchpoint” you wish to pursue (see here) this market includes roughly 67 million vehicles.  That’s how many are on the road, less than six years old, with no coverage.  Dealers are the group best positioned to serve this market.  Of course, a dealer can only address his local share of the market, not the whole 67 million.  See Profit Opportunity, below.

To succeed with D2C, you must have an existing relationship with the customer.  That’s because success requires digital marketing, and anti-spam laws limit what you can do without a relationship.  For example, an OEM can email their customer a solicitation for their factory-label protection products, but a TPA cannot.

So, the dealer has the inside track.  He has the relationship, the contact info, and a service department to verify eligibility.  Plus, every additional VSC aids in service retention.  Depending on the dealer group, it may also have the other ingredients.  Here’s the parts list:

Direct to Consumer (D2C) Operation

  • An advanced CRM with the ability to run a scheduled, multichannel contact program.  Salesforce calls this a “customer journey.”
  • A source of premium finance, like SPP, Budco, or PayLink.  Dealers will already have one of these, for their instore F&I operation.
  • A call center, which could be the BDC, to participate in the selling journey and also to deal with issues around premium finance.
  • A branded website capable of presenting and selling the service contract, including Visa checkout and premium finance.
  • A service facility.  If you’re not a dealer (trying to cover all bases here) there are still things you can do with Pep Boys and mobile facilities like Pivet.

Depending on the dealer group.  Obviously, if you’re Lithia, you already have a finance arm which could (with training) handle bounced Visa charges.  They’ll need to comply with the PCI security standards.  Maybe that’s best left to PayLink. 

There is a host of such decisions, for which you will need expert assistance – but let’s get back to the “why.”  We are going to make a gross profit calculation in three steps:

Direct to Consumer (D2C) Profit Opportunity

  1. Compute the potential product gross that didn’t close with the vehicle sale.
  2. Estimate the likely D2C conversion rate.
  3. Add the backfile of customers from prior years.

Let’s look at AutoNation.  Sorry, NADA Average Dealer doesn’t provide enough detail.  Even the AutoNation data doesn’t provide much detail on product sales.  Still, we can draw some inferences using the 2019 annual report, industry norms, and these remarks from then-CEO Cheryl Miller.

AutoNation reported sales of 283,000 (new) and 246,000 (used) with F&I PVR of $1,904.  That’s the headline figure, including finance reserve.  Owing to adjustments, the figure in the annual report is a little higher, and the calculation based on Miller’s summary is a little lower.

The $1,350 (new) and $1,050 (used) are averages across all units.  We can infer that product gross was roughly $2,580 (blended) on 47% of units.  That leaves the other 280,000 vehicles unprotected, with a potential gross of $720 million, equal to 71% of AutoNation’s reported F&I gross.

You can use 75% of F&I gross as a rule of thumb.  In general, product gross is two-thirds of F&I gross, and this is derived from fewer than half of the vehicles sold (omitting ancillaries).  The ratio of D2C opportunity to instore penetration is 53/47 of the two-thirds, which makes 75%.   

This $720 million is the potential gross AutoNation left on the table in 2019.  Okay, that’s not fair.  It’s only on the table assuming every one of the D2C prospects will buy the product, which they won’t.  Most won’t, in fact.  The conversion rate is the product of three factors:

  • The number of contacts per customer, based on your touchpoints and your programmed journey.
  • The take rate, which is the percentage of people who take action by clicking the PURL link, scanning the QR code, or whatever.  The industry norm here is 2-5%
  • The close rate, which is the percentage of takers who are closed by the call center or self-close on the web site.  Expert closers can do 20-30%.  Remember, this is within the self-selected “takers.”

The conversion rate is what counts, but we break out the components for management purposes.  For instance, maybe the take rate is high but your closers are weak.  Success requires a lot of contacts, with compelling CTAs and good closers.

Let’s say we utilize all of our advantages as a dealer – an aggressive journey on all touchpoints – bringing our contacts to 10.  Multiply this by a conservative 20% close rate and 4% take rate.  This gives us a conversion rate of 8%. 

In our AutoNation example, this would mean roughly $58 million of additional gross.  All of this arithmetic generalizes, too.  Simply take reported F&I gross and multiply by 6% (8% of the 75%, above, makes 6%).  So, now I can crack the Lithia annual report with F&I gross of $580 million, and reckon that D2C could mean $35 million to them.

This incremental income recurs annually, since it’s based on one year’s volume – but we start the game with a backfile of unserved customers from prior years.  We might reasonably want to go back five years for new vehicle buyers and three years for used. 

AutoNation is a convenient example because they sell new and used in roughly equal parts, so this works out to (blended) four years’ worth of volume.  If your mix skews more toward new vehicles, then your backfile opportunity will be richer. 

In short, you can use the 6% rule to compute the annual recurring, and then add a one-time opportunity of 4 or 5 times that amount for the backfile.  This more than pays for setting up the operation.  So, is it worth the hassle to earn an extra 6% of F&I gross?  Ponder that next time you see the Car Shield ad on television.

AI-Based VSC Risk Rating

I have been working on a startup that will use artificial intelligence to rate vehicle service contracts.  For a VSC provider, increased accuracy means sharper pricing and, potentially, lower reserves.  Outsourcing this work to a specialist bureau means reduced costs, too.  Our business model is already used for risk rating consumer credit, and the technology is already used for risk rating auto insurance.

In this article, I present an example using auto insurance data.  If you would like to see how our approach works with VSC data, please get in touch.  We are currently seeking VSC providers for our pilot program.

The French MTPL dataset is often cited in the AI literature.  It gives the claims history for roughly 600,000 policies.  Of these, I used 90% for training and set aside 10% for testing.  So, the results shown here are not just “curve fitting,” but predictions against new data.

The Gini Coefficient

The challenge with insurance data is that most policies never have a claim.  This is known as the imbalanced data problem.  If you’re training an AI classifier, it can achieve 95% accuracy simply by predicting “no claim” every time.  You will want to use an objective function that heavily penalizes false negatives, and you may also want to oversample the “with claim” cases. 

The dashed line in the chart above represents cumulative actual claims, sorted in order of increasing severity.  This is called the Lorenz curve.  You can see that it’s zero all the way across and then, at the 95% mark, the claims kick in. 

The blue line is the Lorenz curve for the predictive model.  A good fit here would be a deep concavity that hugs the dotted line.  That would mean the model is estimating low where the actuals are low (zero) and then progressively steeper.

The Gini index is a measure of the Lorenz curve’s concavity.  This 0.30 is pretty good.  The team that won the Allstate Challenge did it with 0.20.  The downside to Gini is that it only tests the model’s ability to rank relative risks, not absolute ones.  I have seen models up above 0.40 that were still way off on actual dollars.

Mean Absolute Error

The key metric, to my way of thinking, is being able to predict the total claims liability.  This automatically gives you the mean, and Gini characterizes the distribution.  I like MAE because it represents actual dollars, and it’s not pulled astray by outliers (like mean squared error). 

Here, you see that the model overestimates by 1.2%.

You may be wondering why MAE is so high, when we are within $1.00 on the average claim.  That’s because all of the no-claim people were estimated at an average of $72.50, and they’re 95% of the test set.  The average estimate for the group that turned out to have claims (remember, this is out-of-sample data) was $130.70. 

Neural Networks

For claim severity, I trained a small neural net, including my own custom layers for scaling and encoding.  I really like TensorFlow for this, because it saves the trained encoders as part of the model.  You want to use a small neural net with a small dataset, because a bigger one can simply memorize the training data, and not be predictive at all.

This dataset has only nine features and, in fairness, a linear model would fit it just as well.  My code repo is now filled with neural nets, random forests, and two-stage combo models.  What this means for our startup is that we don’t have to hire a platoon of actuaries.  We can get by with a few data scientists using AI as a “force multiplier.”

Earlier this century, I played a key role in moving the industry to electronic origination.  At the time, it was clear that the API approach would liberate VSC pricing from the confines of printed rate cards and broad risk classes.  Each rate quote could be tailored to the individual vehicle.

As I said earlier, our approach is current, proved, and working elsewhere.  It’s just not being used in the VSC industry … yet.

Which Admin System Are You Running?

Thanks to the people who responded to my query on Linked-In.  I am still learning about the systems used by my customers in the vehicle service contract (VSC) market.  Most, it seems, have homegrown AS/400 systems.  I have also gotten referrals for Sirius and Stone Eagle.  Please comment here, and tell me what system you are running, whether COTS or custom, and what the platform is.  Thank you.