WithClutch & RudderStack: Driving Auto Loans with Customer Data

Kristen Glass (00:10)

Hello everyone and welcome to today's webinar. We're very excited to be joined for a special session with our customer WithClutch, a super fast-growth startup. I'm Kristen Glass and I work in product marketing here. Before we dive into the webinar content, I want to cover just a couple of housekeeping items. We want today's session to be as interesting for you as possible. So make sure to submit your questions and your comments for us into the Q&A and chat. And we'll try to answer them live as well as during a Q&A session at the end if we end up having some additional time. Now I'm going to transition to the next slide and we'll go over a quick agenda for today, as well as introductions. So for today, we're going to hop into WithClutch's story and how they heard about RudderStack, how they're using us to help fuel their growth, and what their plans are with RudderStack in the future. Then after that, I'll give a brief overview of RudderStack and how you can get involved with us and we'll hop into the Q&A.

Kristen Glass (01:27)

So for introductions, I'll go with mine first. It's short and sweet. I am the Director of Product Marketing here at RudderStack. I previously worked at Cisco Meraki as well as a couple of different startups. And now I'm going to hand it over to our two special guests for the day, Chris and Rafael. So feel free to introduce yourselves.

Chris Coleman (01:47)

Sure. Yeah. Thanks, Kristen. I'm Chris. I'm one of the co-founders of WithClutch. Some previous background of me, I've been in the car space a while at this point, little different on the software side now, but we had started this business called Carlypso, which sold cars online. We sold that business to Carvana where we stayed for three years, was incredible story, had a great time working there and then in last year, started this business in July of 2020. And that's, I'll get more into that company and the background in a bit.

Rafael Magrin (02:24)

So I'm Rafael Magrin, I'm the Head of Engineering at WithClutch. I've been working with tech for more than 10 years. I've been at Microsoft. I did a little bit of consulting with big companies, small companies, different areas like credit unions, airlines. Then I worked with a little bit of a messaging platform for a couple of years. After that I joined a new bank, a FinTech in Brazil, stayed there a while, helping most on the mobile front end. And now I joined WithClutch to help get the first version live, we got it live on January. And after that, I'm starting to build a team and growing very fast.

Kristen Glass (03:10)

Well, thank you both so much for joining us today. I cannot wait to dive into your story. So first, do you want to give us a little bit more background on WithClutch and how you started the company?

Chris Coleman (03:23)

Sure. Let me start a little background around just what we do in general and who we help and why. And so one of the observations that we had while we were working at Carvana and at Carlypso as well, is often when consumers end up buying a car and this isn't true, just a car, but it's true everything, consumers end up wanting to buy a car and so they don't think too much about the financing itself.

Chris Coleman (03:46)

And so you can actually exert a lot of control over what financing terms and consumer actually goes ahead and accept or not because convenience tends to outweigh a lot of what actually happens on the rate side. And so we found consumers who bought from dealerships, they often didn't get the best rate because they got the rate that they worked with the finance company, that the deal who paid the dealer the most, and then two, there was actually if you look at capital sources, there actually is an advantage to a lot of credits and community banks who often have not for profit status, who can lend at lower rates than competitive full profit lenders, but often the disadvantage because they don't have these huge tech teams, they tend to be smaller regional.

Chris Coleman (04:30)

And so they've got a competitive balance sheet, but not a competitive tech platform. And so this is an activity we went through, but if you go through and you actually look at loan categories and what penetration credit unions have amongst the customers that are already theirs, you can see that even for things like auto loans at most, they have about one-fifth of auto loans of their own customer base. And it gets even worse for things like unsecured loans, a great example, where you have these new FinTech players like Upstart or Upgrade, where credit union rates often are literally half of what those competitors offer sometimes and so we wanted to build a technology product that really enabled these consumers to more easily, not necessarily get new debt, but switch into better debt in a really easy, fast, transparent, seamless way.

Kristen Glass (05:20)

Wow, what an incredible story. Well, I wanted to dive into next a little bit more about your product and let us know some specifics about it.

Chris Coleman (05:32)

Sure. So I'll talk about the product. We can play the video here if that's helpful, or I can voice over here. But one of the things to remember about our product is, we are not the brand and so we're building white-label software for banks because banks have a lot of distrust and banks are actually the ones trying to recapture their own customer subset or get new members. And so what we do is just try to make the process as easy as possible for both the member or the applicant and the person actually doing the loan processing, because it's not as simple as you submit a loan and the loan exists. There's actually a lot of work that goes along around in the background of what's actually happening. And so what we do is we create this white-label platform that surfaces a super-easy way to get a loan and see every loan that you might be approved for, all in one [inaudible 00:06:21] and then simultaneously, it identifies loans that you already have where the credit union products would actually beat the products that you already have on your balance sheet.

Chris Coleman (06:30)

So if you have an auto loan with Capital One we'll surface, "Hey, we notice you have a Capital One auto loan. We could save you four percent in interest each year, click this button and sign a few e-signature documents" and it should be easy as that and then in the back end, what the loan officer actually has a number of other documents that are connected by API to the lending origination system so that they can actually fund that loan as easy as possible. So what we're trying to do is we're trying to make these credit unions look like superheroes to their customers. And so that's the key for us and so, in some ways we're not building the brand, we're just leveraging the brands that we exist and that actually creates some really unique data challenges for us that we'll talk about a little later.

Kristen Glass (07:19)

Wow! Well, what an awesome story. I'm sure the credit unions just love having you guys too. So for the next piece, I wanted to have you guys talk a little bit about what's important for your white label lending.

Chris Coleman (07:40)

Yeah. Good question. So there are a couple of things too. Every customer's different in some way, every customer is different rate sheets and that means like the rates that they'll actually give to members, they serve different memberships for different purposes. They have different brand awareness. They have different ways of working loans too. So their operations are all set up differently. And so one of the things that we try and sort of understanding deeply is just, how are each of our customers working differently? And then B how can I take the best of best from each other in some ways, because in many ways, creating unions are rarely hyper-competitive with each other. They're more hyper-competitive with large banks. And so if you talk to the biggest credit union, who do they really want to beat, they want to beat the Wells Fargo's and the chase's in the world of consumer banking.

Chris Coleman (08:29)

And so for us, a lot of what we need to do is understand, why is someone starting a loan application? What are the credit profiles look like? What are all the loans on the balance sheet they have? Why I did someone go ahead with an offer or not? What is the marginal willingness to self-select into another loan? Often what we find is that $75 ends up being a key threshold of if you can save a member more than $75, that's the point in which they actually switch their finances. If it's $25 or on the lower end, the elasticity to actually do anything about it goes far lower. And then the other challenge is just, a lot of this is very sensitive, personal information, so we do collect a lot of PII. And so we need to make sure that these data systems that we have very well walled off controls on what goes where and why.

Chris Coleman (09:21)

And so in choosing any of our vendors, that tends to be very important. And then three, our vendors, to give you an idea of how the size varies, the smallest partner we work with has about 7,000 members, the largest has close to a million. And so the institution with a million members, they're actually fairly sophisticated. They've got a development team, they understand that they need to be doing quite a lot in technology, they've got, they're doing, they're connecting a lot of the data points, their own internal marketing tools and communication tools versus, your smaller teams when you only have 7,000 members, there's a limited, you have much more limited budget, so sort of what you do and where you connect to and do you have these analysts, they're actually relying on us to do a lot of that work seamlessly. So in many cases, we use our own off-the-shelf tools. And then in other cases, we're simply a conduit to the other systems within these larger institutions. And so we need to be able to do both, based on the size of the partner that we work with.

Kristen Glass (10:30)

Well, that's incredible. So you're handling a lot of data, a lot of different types of partnerships and you have to figure out how to keep that all secure and how to connect with all of the partners and customers.

Chris Coleman (10:43)

That's part of the challenge. Yes.

Kristen Glass (10:47)

Big job. So how did you guys decide given all of those challenges, especially the challenges with your data, and make sure that you're connecting to customers in the best ways possible. How did you end up deciding to use RudderStack?

Chris Coleman (11:06)

That's a good question, actually. So I was introduced to RudderStack from a struggling musician called [inaudible 00:11:15] he's a friend from business school and I think he had been an early investor in RudderStack and effectively one of the problems I was describing while we were still at Carvana is Carvana in 2017 was in a very different place than where it is now. And so it's much more mature in terms of data infrastructure, but back then, you often had multiple data pipelines, you had different users, consuming different information from different sources for different reasons.

Chris Coleman (11:44)

And so in trying to unify all of these customer interactions in a way where both marketing and product analytics and compliance would all consume these same things for different reasons, it just ended up being a very difficult problem and so I had begun this exercise off, what are other people using? What works, what doesn't work. One of our own internal teams had even piloted a segment at one point, which was a very powerful product. And so we got to talking with [inaudible 00:12:17] saying, "Hey look like these are trends that we're seeing in terms of customer data, pipeline management, that are..." We're making sense. And this is probably in the 2019 timeframe now. And so when it came time to leave Carvana and start again, we knew we needed something and the question is, what's the best something. And so that's how we came across RudderStack and came across the ability to, we knew it was something that would scale in a way that segment necessarily didn't and it would just afford us quite a lot more flexibility too. And so that is what led us to you guys.

Kristen Glass (12:56)

Well, thank you for sharing your story. And now I want to dive into a little bit more details about how you're using RudderStack today.

Rafael Magrin (13:08)

So one of the things that was important for us right now is that Chris mentioned, that was a scalable solution, but at the same time was something simple that would not create a very big overhead to us. We grew a lot of teams, but when we started using RudderStack, we had a very small team. So being able to focus on what was our main business, was more important for us, and have a big tool and a lot of complexity. So we actually started with the use case using customer IO. So Chris, can you explain really quickly our use case for customer IO in This scenario?

Chris Coleman (13:48)

Yeah. Great question. So effectively, what we're trying to do in lending is basically recapture cart abandon. And so often what happens is, we said, somebody will start going through the loan process, they'll actually get through the process of a credit poll and then abandon. But once you see the credit poll, oftentimes, you know that this person actually could have saved substantially. And so this, someone who, for whatever reason, was on their phone starts a loan app and you see an upstart loan, that's a 29% and the credit union is willing to offer them a 14% loan and that's a 100% real example by the way. And so you know this person has substantial money that they could be saving each month, but for whatever reason, they've abandoned.

Chris Coleman (14:27)

And so we want to be able to try and recapture that member who uses this saying, "Hey, I know you started, you didn't get the chance to see this, but when there's a more convenient time consider coming back and initiating those savings." And so we came up with these very simple, just four trigger points of, here are points at which we wish we could literally just say something to the end-user in order to motivate them and we don't try and annoy people who couldn't have saved or who couldn't have done these other things. So it's all intentional in trying to be proactive and positive outcomes for that end user. But there were four circumstances where we wanted to connect via email or via SMS to try and re-get this member, reuse it to reconsider finishing that application.

Rafael Magrin (15:20)

So when we start with this case, we could have connected directly to customer IO, but we are discussing a little bit, we're going to start creating some events. That means something on the business side or for what happens to the customers. So we can actually get to customer IO and take some actions. And we are also going to have some PII involved with it, like Chris mentioned if you're going to send an SMS or, or an email, we're going to need that information to send the customer. So we decide to go through RudderStack So send events to RudderStack and then go to the customer IO, would be a little bit more complex in the beginning, but then it opens up for other things that we want to do later on.

Rafael Magrin (16:01)

So then we started using it also with amplitude. But then, it comes up interesting part because, for amplitude, we don't want to have any PII either, we don't need that. We just need the information about the vendor, what happened. And with RudderStack, we use the same events where we could generate to send data to customer IO, but we just use the transformation to clean up the PII. And Amplitude events live and we started being able to analyze the information we went there.

Rafael Magrin (16:37)

Another case is that we use Datadog for monitoring the application and having those business events that help us create alerts or something going wrong on the business side. Same thing here, we need the information there. We don't need the PII, we just need the main data. [inaudible 00:16:53] one day, field task, it was working. Now I was trying to start to send also to Redshift on AWS, so in a start, having those events start long term, and we can actually do some more acquiring a little bit more to get a little bit deeper understanding of a few things.

Rafael Magrin (17:16)

That's a little bit different of the analysis we do on amplitude where we already know what kind of information I'm looking for. So RedShift's going to be a little bit more exploratory and then recently we had another use case that was quite interesting because it's more like on the operational side, not really on the business side, that because of compliance reasons, we need to understand some of the things that are happening to, for example, what are the changes application is having, who is changing something requesting GitHub and things like that. So if a field is more operational in RudderStack, we're able to get those events from GitHub, send them to data doc, and have all the alerts that we needed around comply with us without much complexity. If we're doing the connection directly from GitHub to Datadog or some different way, would be way more complex. So we think in a very interactive approach of how we use rather than stack and start to connect new pieces that help us get where we want to be.

Kristen Glass (18:30)

I love this. And how long did it take you to set up all of this?

Rafael Magrin (18:37)

So it's hard to say how long because as I mentioned, it was an interactive approach, right? We started a very simple use case and started expanding, but in general, the RudderStack part was the most simple part that was just like in a day I could set up any of those connections even less than a day. And the part where we spend most of the time is actually thinking about events and what we would send to RudderStack what are the business meaning of those events. That's where most of the time it’s sent.

Kristen Glass (19:11)

Awesome. Well, I'm super excited to see this stack. And I wanted to ask you a little bit more about some of the use cases that you have on the horizon. Cause I know you guys said that you've been starting with RudderStack, but you've also been thinking about things you're planning on doing in the future.

Chris Coleman (19:28)

Yeah. I'll jump into a couple of these, but this is where I mentioned, the way small, for this cart abandons is a great example. For these small [inaudible 00:19:38] setting up a limited standard set of automated communications triggered through our instances of customer IO make complete sense for somebody who's a huge institution. They actually want to control communications and they want to be able to have interactions where someone can jump into those streams in a manual way too. And so we actually, there are use cases down the road where we may want to inject a portion of these events actually into their marketing systems as well.

Chris Coleman (20:06)

So this makes it easy for us to say, for partner A, who has Salesforce marketing cloud, that's different than a customer, we want to send these events. And that varies by the technical aptitude of the partner that we're dealing with and what subsystems they have, some want to consume all these event streams and all these business streams and others don't. And so being able to do this in a way that makes sense for our largest partner is another way. This just makes our work easier.

Kristen Glass (20:36)

Awesome. And I love how you guys have chosen RudderStack and you're really growing with it over time. So why did choose, as a startup, to turn to something like RudderStack? I know some companies try to turn to tools like ours a little bit later. I want to learn a little bit more about why you chose to take a look so early on.

Chris Coleman (21:02)

I think I'll leave that to Rafael because I think one obvious question that we had our first use case was customer IO, integration with customer IO. And so at first, the question was why don't we stream events just directly to customer IO? Why do we need this thing in between in some ways? And so I'll leave it to Rafael to comment on why that made sense and why it made sense at the time and why it continues to make sense.

Rafael Magrin (21:30)

Yeah. So usually in the stage we are, we always think, is this over-engineering or not? And in that case, given my past experience with setting up analytics and things like that, if you don't get a few things right from the beginning, it takes a lot of effort and work to actually put that in later on. So we started playing a little bit with RudderStack to see, hey, let's see if it works for us. If not, it doesn't add too much complex to us. If it doesn't, there's no point of not using it right now, get the benefit and get the benefits later. That's what exactly happened. We started using it, it was simple enough for us not creating on an overhead. It allowed us our first use case. It was customer yield, but the other ones that were having [inaudible 00:22:26] we knew that would be easy to set them up if we had the pipe correctly in the beginning. So that was the main reasoning of, let's start with RudderStack and not like connect things directly.

Kristen Glass (22:39)

Awesome. I'm going to take a quick question that someone sent in really quickly. So they're interested. Do you have any advice for someone who is looking at developing something like you guys did or advice for someone starting with RudderStack or similar tools?

Rafael Magrin (23:01)

My main suggestion is to figure out what [inaudible 00:23:20] first use case they're going to have and build something for those use cases and not try to create something for the use case they're going to have in a year. And one of the key things is trying to figure out what are exactly the events you need to build and what they mean in the business sense, not just in a technical sense of something happen. Here's where I see most companies struggle figuring out what exactly events you need to generate.

Kristen Glass (23:35)

Awesome. Well, thank you so much for giving me your background. I know we have 30 minutes today and we're getting closer to the closing. So I want to go over a few quick details about how our viewers can learn a little bit more about RudderStack. So there are three main ways that I would recommend that you can learn more about RudderStack.

Kristen Glass (23:58)

For starters, you can start using RudderStack for free and I've put the link to where you can sign up for free on RudderStack below. We also have a fantastic blog and a team of technical writers that are writing posts on our blog about our product and about the space. So if you want to learn more about the space in general, feel free to read articles on our blog, our websites, and other good resources as well. And then finally, if you're interested in getting a demo or learning more about us from the sales teams, you can contact our sales teams. We have a contact form on our website at this link, and I would recommend that as the next place to go to.

Kristen Glass (24:43)

So now I'm going to hop into the Q&A portion since we have a few minutes for Q&A. So the first question that came up is you mentioned flexibility in your choice of rudder stack over the segment. Anything specifically that stands out there now that you're using the tool?

Chris Coleman (25:10)

Do you want to comment Rafael?

Rafael Magrin (25:17)

In general, honestly, I don't have much experience with Segment, but for RudderStack was very easy for us to use. It works very well for a use case and the support we're having is very good, so in general, that was the main decision.

Kristen Glass (25:40)

All right. And I know we have another question coming in since I see that someone's just about to submit it, but the next thing that it looks like people are asking is how was this process for set up? Was there anything that was unexpected in it? Was it pretty seamless?

Rafael Magrin (26:06)

It was seamless. I don't remember anything as partial that happened. We played around a little bit, tried a few things. Most of what we did, we didn't face any issues, the documentation was clear on what we needed to do. And that was it.

Chris Coleman (26:27)

I think we had our first viable business case actually fully running within a week. And so between when we started and when we actually had what we wanted to get out, which was just like simple a bit for abandoned card events, I think we actually had everything running end to end on our pipeline within one week.

Kristen Glass (26:49)

Wow! That's impressive. Well, it looks like we have time for one more question. So I'm going to pick one more of the questions that came in. Do you have any other pointers on the things to get right at the beginning of architecting a stack for an early-stage company?

Rafael Magrin (27:07)

Yeah. I would say, try to understand the business and what are events you need to track. I know I mentioned that before, but that's the hardest part. Once you get at and you gather information to RudderStack, it's easy to play around with different tools, connect to them and see which one makes sense for you.

Kristen Glass (27:31)

Well, Rafael and Chris, thank you both so much for taking the time out of your busy day to be on this webinar. We really appreciate it for our viewers. If you want more information about RudderStack or about WithClutch, feel free to visit our website, and thank you so much for joining us for today's webinar.

Chris Coleman (27:55)

Cool. Thank you.

Rafael Magrin (27:56)

Thank you.