How Volie Processes and Cleans Data

Foreign.

Welcome back to another Volley rally. Today we've got one of those guys kind of behind the camera really if, if this were a movie, one of our key central people involved in volley and, and one of my co founding partners, Peter Sperry. How you doing today, Peter?

Doing very well, Scott. Thanks for having me.

Yeah. Well, wanted you guys to meet Peter. He isn't on many calls and he's done many of the hard things that make Valley look easy every day. And Peter, why don't you tell a little bit about yourself before we get into data matching and what you do for volley.

Sure. So a good part of my career has been spent in the automotive industry, either working for software vendors or for dealers directly on the Internet side of things, on the business development side of things, and on the technology side of things in the dealership, and have spent the last 10 years working on entrepreneurial software ventures, a few of them with Scott. And that's kind of my background. I've done a lot of, a lot of work in the global e commerce space. So very familiar with working with gigantic sets of disparate data and volley. We started a, a few years back to kind of do all the things that we ever wanted to do in a software company that we couldn't do right, but now we were going to take the opportunity and do them right.

And you're married to a beautiful woman, you have three children and you literally live on a mountain in Montana.

Yep, yep. Live in the foothills of the Rocky Mountains in Montana where we are known to get four seasons in one single day. And summer in Montana last year was really beautiful. It happened on a Thursday.

Yeah. All right, well, good. So let's talk about data. And you know, like I said, Peter does a lot of the hard things with data in Valley. He's been the architect behind it. So what types of data do we load into Valley automatically and through spreadsheets? And why is it important?

Sure. So we have integrations with many of the major DMS and CRM providers within the automotive space, as well as several of the equity mining platforms. We also support the kind of the industry, I'll use this word standard loosely of STAR and adf, as well as some custom OEM formats that are sort of derivatives of those two formats. And all of that gets ingested into volley into our data management platform, which is really the section of the application that is opinionated about data. Its job is to process data in a very expedient fashion using all that we've learned over the last 10 years about automotive data how most of it's kind of crummy and hopefully trying to make what we would call the perfect customer or the best customer. And we process the dealership customer data, sales data, service data, appointment data, inventory, and we, we mash it all into one spot. And then we also have CRM integrations where we'll take in lead data, whether it is current leads that are coming in actively from a cars.com or an autotrader, or whether it's aged leads or unsold showroom follow up prospect data that we get from the CRM and we, we match it all up and try to come up with that perfect version of a customer so that our dealers and our call center partners and our marketing platform partners have the best possible data on automotive dealership customers.

Hi. So guys, you're all using Volley. It just works behind the scenes. This will probably be the first and the last time that you ever even care about this answer. But if you're on this podcast, you care about data. So we always talk about in our demos, geez, we do deep and dedupe and normalize data. What does that actually mean?

If we kind of start with the base level of a customer. So a customer is our representation of a unique human as they walk and live and breathe on this earth. And a customer has various attributes. They have a first name and a last name. Most of the time they have one or more phone numbers, they have one or more mailing addresses, they have one or more email addresses. They have a date of birth. They've got some vehicles that they've either bought at various dealerships or serviced at various dealerships. They may have submitted leads on a cars.com or an auto trader or other, you know, the dealer's website, manufacturer's website. And so this idea of a customer is, is a, it's a human person that belongs to a single dealer that has certain attributes. And so when we say that we match and dedupe data what we know from the various platforms in automotive, if we take the DMS and the CRM specifically, is they don't do an incredible job at enforcing uniqueness. There is only one Peter Sperry in the world. My wife and, and Scott sometimes tells me that the world could only handle one. So that's why we do our best job to distill all of the different records about Peter Sperry that may come from the DMS or the CRM or maybe the dealer gets a lease termination list that they want to upload into Volley. All of those get analyzed according to their attributes. And just like when you make a decision in the human world about what things are the same and what aren't the same, you have kind of an idea that, well, some attributes about a thing are more important and more valuable and more likely to be unique than others. And so we take all of the attributes from all of those different records that we get from likely a single human, and we weight them, we clean them up so that we force things to be like they should. An email address always has to be something at something dot something. A phone number in North America has to be 10 digits, not starting with 1 or 0. We know that there are certain area codes that are bad. An address in the North America has something like a street one, a street two, a city, state and postal code. Although we do have sensitivities for countries built in, fundamentally most of our data is North American that we deal with. And so because we're opinionated about what an attribute looks like vis a vis a phone number, we can immediately toss out attributes that we know are not valid. So we toss out five digit phone numbers, we toss out email addresses that are none.com, because we have a repository of known bad or known fake email addresses. And then we use all of those attributes and we say, okay, this is what we ended up with after we cleaned everything up. And now we're going to see if, if there's someone that's kind of close to that in our database and maybe there's a hundred of them. And then we go through each of those candidates and we score them based on an algorithm that's pretty straightforward if you were to think about it. But it's pretty proprietary in terms of the software space. And we say, yep, we've already got this customer. Or we say, nope, none of the attributes match or their score isn't high enough. And so then they end up getting a new customer record created. And, and by doing that, what we've been able to do is we've been able to eliminate about a third of the data that comes in as duplicates. And then we can match those transactions to the right customer or the right prospect or the right lead within our database.

How does it help our customers run better bdcs then if we've got clean data?

Well, the reality of any outbound marketing effort is you're going to make some bad communications. You're going to call a customer who you think hasn't been in for service for six months and they actually brought the car in a few weeks ago, but it was their wife that brought it in or it was their son. So there's the, there's a certain inevitability that you're going to send the wrong email to a customer, you're going to call the wrong customer, or you're going to call the right customer for the wrong reason. So we know that. But as much as we can clean up data, if there's only one customer that represents a single human within our database, it means that we can develop a profile about that customer, about what he or she has done throughout the course of their history with that store. And it allows us to execute much more accurate marketing campaigns because we've attributed all of the sales, service, appointments, maybe leads to the right person. And that way, when you build a campaign, you're looking at exactly the group of customers that meet the exact criteria. And you know that because we don't have five different versions of a customer, some with the sales transactions, some with the service transactions, and still others with appointments, we smash them all into one. And so now you have a complete picture about what each individual customer has done.

It's, it seems so easy. It's very hard. We spend an extremely large portion of our dev budget on this. You spend an extremely large amount of time making sure that the data loads fast enough because we're sitting on, across all our customers, a couple hundred million customers, ninety million vehicles, tens of millions of email addresses, and it has to come back up in about a half a second in between every phone call. So it's, it's, it's, it's vitally important. I hope everyone got a good sense of, of what we do behind the scenes and this, the stuff that just has to work. It's kind of our secret in plain sight, I think. Peter, anything else for today?

No, it, it's been a fun call. Glad to get introduced and be let out from behind the curtain for a little bit and glad to go back behind the curtain after we're done.

Thank you. Thanks for this today. Have a good day.

Peter, Thanks.