9 Clouds Automotive Marketing Blog
Where automotive meets digital marketing.
This is the second in a series of three posts that explore the concept of predictive analytics and the three ingredients that will help you drastically increase your effectiveness online. Read the first post and then subscribe to our blog to ensure you get the next post delivered to your inbox!
Once we’ve attracted users to our site, we can start to use predictive analytics.
You’ve heard of analytics — the information that comes from analyzing data from people on your site. Predictive analytics takes basic analytics and goes a step further by using information about your customers to predict their future behavior. While you'll never be able to know exactly what will happen, you can determine what’s very likely to happen and you can give your customers what they need even before they have to come looking for it.
And you already have this data in your Customer Relationship Management system, or CRM. You don't need to start from scratch. You've been collecting customer data every day since you opened your store. Even if it's not digital, you know someone's mailing address, you know someone's ZIP Code, you know how much they've paid for their vehicle, previous times they've serviced their vehicle, etc. All of this is housed in your CRM, in your DMS and different places.
Predictive analytics brings all of this information together to create a holistic understanding your customers. You have the power. If we can harness this power, it can become your marketing superpower. But, as Spider-Man says, “With great power comes great responsibility.” We need to use this information wisely.
Some of the data points that we have include days since purchase, mileage, number of repair orders — but we don't want to write someone and simply say, “I saw that you have eight repair orders. Do you want to buy a car from us?” Instead, we need to use that information to inform us on who to talk to and how to talk to them.
So how do we get started with predictive analytics? We're going to start with your CRM. We want to export the information we described above onto a spreadsheet. We want it on a spreadsheet because it allows us to do some fun math to figure out who we should talk to and then use that information for sales and marketing. Once we export that information, we're going to look at the key factors and then those key factors will guide us in how to contact those individuals. So these three steps are all we need to use predictive analytics.
Export CRM Data
Let’s get started with the first step, exporting CRM data. If you use a CRM like DealerSocket, you'll see a button called Query.
You can click on that and then create a spreadsheet.
If you're using something like ADP, you can then view all your leads. We usually enter an email address that contains the @ sign, so you only have digital leads not just offline leads.
Then you can print your leads and export to CSV.
Or maybe you're on ReyRey. You can look at the contact types under prospect for example.
You can choose the fields and again we like email containing the @ sign and maybe people in the last couple of months or years.
You can then export that data.
So whatever system you're using whether it's a CRM maybe even a DMS, you can export that information to a spreadsheet.
View Key Factors: Time | Event | Forecast
Now we take that information we’ve gathered and mix it all together to create the perfect marketing recipe, which we will go into detail about in the next step. But first we need to look at a few key factors in our exported data. There are three of these key factors: time-based factors, event-based factors and forecast based factors.
We’ll discuss time-based factors first. If we want to know how we’re doing at retaining our customers, such as getting them to come back and service their vehicle, we can export our customers. We can then view the purchase and the service date while hiding everything else. Next all you need to do is create a formula to subtract the purchase date from the service date and you'll know the average number of days between their purchase and service dates.
If you do this for all your customers, you'll get a really interesting look at the success of your dealership. You'll maybe be able to know that, on average, a customer comes back 200 days after they buy their car to service their vehicle.
Now you have a benchmark you can use with your sales and marketing team. You can ask them to aim to lower that from 200 to 180 by the end of the year. It's a very quantifiable, easy to measure metric that will let you know if you are getting better or worse at getting people back in.
With all this data of course, we can get as targeted as we would like. Right now we might be talking about all of our customers, but we could drill in and only look at truck customers, hybrid customers, used or new customers, etc. to get a more accurate picture of the days before service or any other data points.
A pull ahead lease is another example of a time-based benchmark, which is getting people to re-lease, come back and get a new lease. To do this, we look at the number of days or months left in their lease when they lease a second vehicle.
These would only be the customers who are leasing a second vehicle from us. We can subtract the first lease date from the new sales date and that would tell us that maybe there were 12 months left when they re-leased a new vehicle. Again, we can take this average and tell our team that, on average, people re-lease at 12 months.
Then, instead of sending an email to everyone about moving up in their lease, our team can send it to people who are 15 months away from their lease ending. This will be more effective because if they're going to re-lease in the next three months on average, that's a perfect time to talk to them. So what we're doing here is we're finding the time that our customers will be ready to buy.
Again, we know all of this based on our historic data. Our customers do not have to fill out a form or tell us. We know it on average by looking at all of our customers.
We can also do this with event-based information. It's not necessarily a date but instead when our customers do something at our dealership. For example, we can filter our customers with sales dates. So we should filter out people who have purchased a vehicle so we can look at their mileage at trade-in. If they traded in a vehicle when they purchased, we would know how much mileage was on their old vehicle.
That allows us again to create a really interesting benchmark such as what the average number of miles someone has when they decide to trade in. Maybe we look at an average of 80,000 miles and again our sales and marketing team knows to market to or call anyone with 70,000 to 80,000 miles. This allows us to avoid wasting our time with someone with 20,000 miles because it's less likely that they're ready to buy.
That's all information we can get by looking at the current single data point. We can do the same thing looking ahead. We could, for example, look at vehicle equity, which tells us what the average purchase price and the buyback price are.
As a dealership, if we're trying to increase equity on the vehicles that we're buying, we can look at this information and decide when people will buy based on their purchase price and the buyback price we could give them. We can also do the same thing with the lease forecasting by taking a bar graph of all the customers we have in each month of the lease and then predicting the pull-ahead leases.
If we know people lease again 12 months before their lease is up, we can look at our bar graph and say, “Okay, the last few months should be some pretty big months because we have a lot of people that are up for leases.” If you don't have many people up, that's going to tell your sales and marketing team that they need to work harder in the next couple of months.
This all helps you forecast the health of your store. All of this information is obtainable with no real skilled math abilities. You just export the data, allow Excel to do the math for you and look at a single point of data like we did in the examples above. So we're looking at just mileage or just date since purchase, etc.
Contact Top Prospects and Create Marketing/Sales Actions
You’ll feel like a pro with the final step — combining the information into one. Amazon and Target are kind of infamous for this. Target makes the most money from mothers so their goal is to get someone into Target before they become a mother. They know that if someone comes in and buys prenatal vitamins, diapers and a stroller, they're going to keep coming in for the rest of their lives. Because of this, they've spent a ton of time and money trying to create a system to identify when someone is going to be a mother.
It kind of came back and bit them in the butt in 2012. They actually had a family in Eden Prairie, Minnesota, that they sent a catalog to like they do to all women that they identify as potential mothers. So they sent out this expecting mother catalog with coupons for all the things they might need – a stroller, the diapers, the vitamins.
The problem was that this young woman was only 16 years old.
When her father saw this catalog come to their house, he freaked out. He actually went to their nearby headquarters in person to chew the ear off of the people that were creating and sending these catalogs. When Target called back two days later to apologize to this father, he sheepishly told them that he had something that he had to admit — that there was something going on in the house that he didn't know about and his daughter was actually pregnant. Read more.
By using this plethora of data points, Target was actually able to identify that this woman was pregnant even before her father did. That's the power of big data. We can look at multiple data points and understand what's going on before the customer themselves maybe even know what they want to buy.
So how does this apply to automotive marketing? This requires a little bit of math but I want to walk you through how we can use these multiple data points to create a target list of customers. Think of this as the “what you might also like” feature in Amazon. There's actually a great stat that if you go to Amazon the month before Halloween, they have about a 60% chance of guessing what Halloween costume you're going to buy based on your previous purchases. So this data can be pretty powerful!
An Automotive Example
Let's use an example of how this might work. We looked at a single-point Ford dealership and we looked at about 5,000 cars that were purchased by customers. These would be customers that bought two vehicles, because we want to look from one purchase to the next and what happened that made them purchase again. Out of all the customers, it was, on average, about 238 days from when they purchased to when they serviced their vehicle. There were only 346 days between the first purchase to the second.
That means they're obviously buying vehicles maybe for someone else, maybe they're farmers buying vehicles for their farm, parents for their children, or whatever it might be. The average mileage at trade-in when they did trade in, however, was almost 90,000. These again are key points for our marketing team to know. If we know that days before the first service is 238, we better be talking to people between 200 and 230 to get them to come in.
We first looked at this information visually. So here's how the dealership's days since purchase looked as a graph. It gives you a little sense of the distribution of customers.
Here's the number of repair orders that this dealership had. Most people come in at less than five, but you would think that the people between 5 and 10 probably need a call to see what's going on.
We can also look at the previous purchase price. This gives us an idea of the price they traded in from.
And here is a look at the future history of the dealership. We can see the days until someone's lease is done and the negative would be people who have already paid off their lease. So if I'm looking at the health of the dealership, I can see that there's not a lot of people with about 200 days or less on their lease, which means I better be talking to people at 300, 400, or 500 days when that starts to scale up on that graph.
What we wanted to do is not look at a single data point, however, but combine these things like Target or Amazon. So we set up a weight of evidence model, which basically means we took all the customers and put them into ten different buckets and then identified which factors were present when they purchased their second vehicle. Some of them had zero impact.
For example, the average number of website pages they looked at had zero influence on whether they bought or not. Some of them had a little bit of influence. For example, the emails they clicked before they bought or the days since they clicked an email, or the previous vehicle price with two stars, and all of those. They had a little impact, but not enough for us to consider them as worth adding into our final algorithm.
The factors that had the biggest impact, however, were the number of days since they previously purchased, the number of repair orders that they had, and the days until they paid off their lease. Those three things were the factors that identified who and predicted who would purchase.
Here you can see the standard deviations that we took a look at and ultimately we use these three factors to export a list of potential customers.
According to this algorithm, that first person has a 96% chance of purchasing a vehicle. So if I'm working with sales and marketing, I can tell them not to just wait in the lobby for someone to walk in or cold call people. Instead, they should look at this list of potential buyers, give them a call and say, “We're calling our most important clients and asking them if they're interested in some of these specials that we have.”
If we look at this list of qualified warm leads, we're going to have a higher close rate and we're going to be talking to people who actually want to hear from us. So we're not spamming but we're talking to the right audience.
Get The Final Post
And that ties right in with our final installment of this three post series….personalized sales. Subscribe to our blog via the form that pops up to make sure you get the trend delivered right to your inbox!
Photo credit: Mark Skipper