Eric Dodds:

Welcome, everyone, to the latest RudderStack tech session. This is a live webinar. I will of course send the recording out afterward for anyone who missed it. We have a very exciting guest today. Welcome, Kevin. It's amazing to have you in one of the RudderStack tech sessions.

Kevin Gervais:

I'm glad to be here. It's going to be fun.

Eric Dodds:

Cool. Well, this is a topic I'm really excited about. Kevin, I want you to talk most of the time and I will interject with questions, but you reached out and just shared a project that you did using RudderStack and a number of other tools that just sort of drastically increased performance for this company. And it was just an amazing story. You reached out in our community channel, I pinged you and I said, "I have got to get the details here." So we hopped on a call. Everyone on our team just sort of said, "Wow, this is awesome." And then that's what led us here to today. So you 4X leads, you're dominating search results. So let's jump in and talk about it. We'll do some intros first, but then we can talk about the problem and then the way that you solved it.

Kevin Gervais:

Absolutely. Yeah.

Eric Dodds:

And I think everyone knows me. I work at RudderStack. But more importantly, Kevin, do you just want to actually give us a little bit of your background? I know you're founder and CTO at Touchless and we'd love to hear about that. But you've done a number of things before and you're a very successful entrepreneur.

Kevin Gervais:

Yeah. A lot of stuff that I've done started when I was seven as a coder. My parents were teachers. And so we had a Mac computer and there weren't many games. So we kind of learned how to make our own. I might have learned how to work with sprites and redo kind of icons and stuff. And then that got me into web design. So I did that for a number of years and then spent about eight years building a SaaS company. And that experience was very different because that taught me a lot about working with data, working with regulated markets. It was really shocking to see that even big companies deal with some of these data issues. When I was outside of that world, I saw kind of the output of some of these projects that they would do and I thought, "Oh, they had everything figured out." But the reality is everybody struggles with data. Yeah. As I passed the baton at my last company, what I am to now is just a new way of developing projects where data is at the center of everything. And when you do that, it's been really surprising to see what happens.

Eric Dodds:

And that's something that I definitely want to talk about. I love your approach of designing the data first before you get into the actual solution and architecture. So let's just dive in.

Kevin Gervais:

Absolutely. Yeah. The agenda we want to get through is just kind of walk through some of the challenges that this project faced at the beginning, and then some of the results, and then how we did it, but one of the reasons why we focused on this idea of data design was because that was at the center of a bunch of the problems that existed. And one of the goals, when I first got introduced to WaveDirect, was these three things. We wanted to understand the customer journey better. Where do they fall off, what causes them to be engaged, not engaged? There was a goal of enhancing communication. So that means just getting better at communicating with customers. Telling them about new offers and services or even just to be able to follow up and check in on them and just see how things are going a few times a year and do that consistently. That was a goal. But there was also some restrictions. They work in a regulated market, had a system that was on-premise and that couldn't be changed out easily. Doing so will be very impactful to the business. It would be really hard and also just a lot of cost and risk involved in doing that. So how do you do the first two if you can't change the systems out? So it's an interesting challenge.

Kevin Gervais:

A little bit about WaveDirect. First, the team is amazing. They're just this local group that cares so much about customers. When I would be on site a couple of times, there'd be clients that would call in, and just right away the staff would be like, "Oh, that's Jerry. Oh, that's [inaudible 00:04:52]." And they'd have just this really deep insight into each customer just off the top of their head. It was just really interesting, but the data didn't reflect that, but at least the people who work there cared a lot about the service. And so for many people-

Eric Dodds:

They had a culture that sort of supported-

Kevin Gervais:

Absolutely.

Eric Dodds:

Very data-driven, but all of it was sort of analog in the employees heads almost.

Kevin Gervais:

You're right. And there's a lot of the team that would be there for many years. So they know these customers intimately, but that wasn't reflected in the data either. The culture was there. And then when we did a survey to all the existing customers back in February and we saw that nine in 10 customers were willing to recommend them to their friends and family, which I've never seen before in the telecom sector. So it's just really interesting that their customers also loved them. They just wanted more consistency with the way they were communicated with. But they're a great team. And the other thing is that they're a lifeline to many people in these rural markets where they can't get service maybe through the cell tower or fiber lines because maybe their firms or they're outside the city limits. And so they really had been key during COVID to connect a lot of people to the outside world. And so helping them just felt good, too. A really cool group.

Kevin Gervais:

So these are the goals. Again, we want to understand the journey to better communicate with customers and yet there were these restrictions. So first was that the system was on-premise and made by somebody else. So we couldn't change the data model that existed there for at least some of that data. So that was just awareness. We had a way to deal with that, but we also had to deal with this legacy system allowing freeform entry of data when you added a customer or update a contact. And so if you wanted to do marketing to them, you want to better communicate which phone number do you use? What if the phone number has notes baked into it accidentally? It was someone's fault. It was just that the data ended up in this state where it was inconsistent.

Kevin Gervais:

I see Joanne was on the call, but we have people on the team, when we first met we were like, "Yeah. Let's start marketing." The first thing we noticed was, "Okay. To who? How do we know who to market to and which contact information do we use and it's not standardized?" You see here are some of the stats. Addresses were incomplete, phone numbers were not properly formatted. Maybe there was dashes or characters in them. And then sometimes the names were not standardized. So a first name might be Joe and Sarah is the first name. Is it Joe or is it Sarah? I think a lot of businesses deal with this in general. Even big companies struggled with this. So it wasn't a new concept but was holding them back. It affected decision making and they couldn't really understand their customers because of this data problem. But we can jump into the results. So this is pretty cool.

Kevin Gervais:

So unexpected first of all. We kind of came into this thesis going, okay, let's clean some data with a goal of working on retention, which we had an impact on that as well, but what was really interesting to see or the other benefits that came from this. These are just some of the other highlights, but an increase in website traffic, an increase in website to lead conversions. So when you look at the visitors that go to the site and who actually go ... who are in the market, who are in the right area and they go through the journey, we saw that going from 1.8% to 10.2%, which is pretty crazy within 45 days too. And so we pulled together the actual kind of building of the site just for some context a few weeks before the launch. We had a concept of kind of the flow we wanted, but it didn't really get kind of finalized until the two weeks before July sixth or July seventh. So just to see that immediate impact was pretty surprising.

Kevin Gervais:

WaveDirect, because of their local relationship with their community, always had a good conversion rate once people talk to them, but just again an increase over their average. I would attribute that. We'll go through some of the more steps, but just so the customers be more informed. So by the time they were contacted through this flow, they had already looked through a bunch more pages, they already had more context so they were more comfortable talking to the team. So that was just interesting to see. And then the other [inaudible 00:10:19] thing was just this average increase in market lead volume. So prior to this launch to be, a certain amount of leads would come in, but in a particular ... In this case, this is just a subset over a certain period of time. It was just a week or two weeks, but when we look at people who were in a certain area, who went through the journey that we cared about, just increasing the volume of the right type of leads maybe not the overall lead volume, although that was true, but that wasn't the goal. The goal was the right type of leads to increase that. The other benefit was just the decrease in poorly formatted data because it's possible to clean that data as it comes in.

Eric Dodds:

Sure. It's really interesting. It's a good reminder that even if you're really good on the phone, your sales team knows the customer really well. If you don't understand the customer journey, especially as it leads into the sales conversation, you're sort of in many ways treating each conversation the same way even though the intent is very different across the leads you're having conversations with. And then the other interesting thing is even though they had a good close rate, if you don't have the ability to follow up intelligently, there's just going to be a lot of fall off from the leads that you talk to who are not high intent who maybe need some sort of follow up. But if the data is wrong, or a number is entered wrong, or there isn't the intelligent sort of follow up on that end. It's just interesting that on both ends, the data can kind of be a rate limiter even for a really good sales team.

Kevin Gervais:

Yes. Good point. Because even just to enforce a contact strategy, which we learned working with the other telcos like our last company, our background was working mainly in telecom where retention is so pivotal. One of the clients has a 0.4% churn rate month over month, which is a fraction of what the others earn. And as a telecom carrier to have something like that, the secret to that was just regular communication. You can't do that without a tracking of the journey and even just that audit trail because you don't want to be contacting somebody multiple times like too soon. You're right. There's all these downstream effects which I think contribute to the fact that the conversion rate is higher, but I also think to this isn't baked in of all the some of that downstream effects. This is only 45 days in. And so I think the real test of this too will be, what happens later? Does that percentage increase if you're able to follow up with them a certain amount of times longer term?

Eric Dodds:

Yeah.

Kevin Gervais:

Yeah. Very cool.

Eric Dodds:

Very cool.

Kevin Gervais:

And so if we go like, "Yeah. So how do we do it?" I think this was one of the goals of this presentation we talked about. Okay. What were the factors? How do we think about this? And you brought up at the beginning this approach being kind of a new one. I think, for us too just to do this at an end-to-end perspective we did with this project, it's beyond normally what you would kind of do with data design, but we're able to kind of design this entire end to end process. But what's the impact of doing data design and selecting the right tooling? Is there a benefit to that?

Kevin Gervais:

So just to help those listening what the lesson is or what lesson we were trying to draw from the issues that were happening. Were some of these points the problems that we were facing in the data quality? Again, not anyone's fault because the system kind of allowed that to happen. And even systems. It's hard for them to anticipate all the things that people can do to their forms. So it's accidental, but when you're looking at the problems that existed in the data, what we knew is that the non-standardization of it became a huge blocker to the business to know where it was at and what needed to change. And we also knew we wanted to progressively identify people and not pull a lot of info about a person without their permission. So privacy was really important to this flow.

Kevin Gervais:

The other thing that we knew going into this is when we looked at the data that did flow like Google Analytics data. When you look at that, we would see a certain traffic volume, but when you actually went deeper in it, you'd see that a bunch of it was out of the market. They had some pages that really high traffic, but they were in Washington or they were in Texas. So well-ranked pages, good content, but not contributing to whether someone's going to sign up for their service because they're limited by geography. Their service is only available to Ontario and sections of Ontario too. You couldn't rely on a lot of these reports if you just go by the built-in tracker that you pop into your website. Something had to clean it a little bit and redefine what a visitor meant or what traffic meant because not all traffic is important.

Eric Dodds:

Yep.

Kevin Gervais:

Does that make sense?

Eric Dodds:

Yeah. Absolutely. This could be a whole additional webinar, but I think more and more, at least, we hear a lot that companies are sort of looking at your traditional analytics solutions like Google Analytics and realizing the sort of depth and quality of the first party data required to truly understand whether the traffic is right or not and sort of associate that with touchpoints way deeper in the funnel. You just have to have an approach where you really are ... You're strategically pulling in sort of the raw first party data and doing a number of things, which is obviously what we're going to talk about, but it is ... We hear more and more that the things that have been working for the last decade from an analytic standpoint just aren't cutting anymore.

Kevin Gervais:

And I see a lot of this in auto to where I think vendors have told folks, "Just put my pixel in. Here's another pixel to put in your thing. There's another little script to put in." And then everyone just going to trust these vendors and are going to sift through everything and really properly identify stuff. It becomes really opaque. And what we found, too, is even in this process, if you actually were to look at our Google Analytics traffic from a before and after, you would actually see Google Analytics traffic drop during this process, even though we're having all these higher results. First, because we eliminate the pages and certain bad traffic, but second, is Google Analytics will often think of two different visits to the same site from the same person is two different visits... two different users. And so just relying on them to just do it for you is dangerous because actually it could end up being inflated numbers or you might think that your numbers are worse than they actually are.

Kevin Gervais:

We knew that we were dealing with some sort of question mark when it came to Google Analytics. Even though it was a goal set up and all these things, it looked really sophisticated. No insight really. Whoops! Here we go. Just some quick things. We knew we wanted to be future-proof. We want to be able to swap the back-end and front-end out at any time with whatever vendor. And we also knew that data had to end up in a warehouse underway drift control in Canada because it's a regulated industry. So we didn't want to store that in systems like segment or whatever and store it indefinitely. It would limit what we could store if we had to use some of these outside systems.

Eric Dodds:

Which limits what you can do with the communication downstream.

Kevin Gervais:

Exactly. Why send sensitive data to a segment if you're not allowed to store that there? So then all of a sudden now it affects your marketing, et cetera. So we knew that we wanted this ability of flexibility and we knew that the data had to end up in a place that they're controlled and we also knew that core vitals was kicking in. And we didn't really know the real impact it could have in ranking, but we knew that it was kicking in. As we started fixing other areas of the data and we looked at the site and we realized the lack of visibility and the poor core vital score was enough to trigger, "Let's redesign this. Let's fix the flow just so we can get visibility and to make decisions."

Kevin Gervais:

The WaveDirect team was incredible. Especially on the executive side. Just the entrepreneurial spirit to be able to first recognize that there was a problem, but second be like, "Yes. Let's jump on it." That is so risky for companies to do and yet they made the leap. And so just a side point, because you don't normally see that. And then the other thing too is a goal is we knew we wanted to primarily service customers over text because that was what the survey said. People wanted to be communicated with through text message like actual conversations like quality conversations, not like blast, but actually like text convos. If that was the end state, there was a bunch of stuff we had to do. So the approach was, let's design the data to make that possible. We don't know what the interface will look like, but let's design the data first. Sorry. Were you going to say something?

Eric Dodds:

I love that term, data interface design, because in many ways ... and I'm not a design expert, but even if you think about things like design thinking and sort of methodologies around that where you're framing the foundational components first before you actually sort of embark on specifics in the solution. And that's the same concept here. It's really the right way to do it especially if you're sort of starting over and sort of building it from the beginning.

Kevin Gervais:

And sometimes you can't do a complete rewrite, but at least being comfortable with what the data should look like in the end in order to give you the result you're looking for is just helpful as a guide just like you do with an interface design for users. The flow was let's design the data that we want. So we cared about in market visits, we knew that we had to redefine the funnel, so we had more visibility of where breakdowns happened. And we want to be able to mock up that data flow before we ever did the actual design. We did start on UI design a little bit, but it altered the interface design once we finished the data design process. So some of the stuff you can do in parallel, but the interesting thing is, the data design drove the interface design. And then we were able to increase page speed as a result, which we'll dive into later. But what we realized is where does this end up?

Kevin Gervais:

Okay. If in the end we want clean events and a clean understanding of customer records and we want to eliminate permanently, ideally [inaudible 00:23:00] to say, but is still a work in progress. There's always polishing you have to do, but if the goal, anyway, is no duplicates, let's redefine how the data is stored. And so we thought instead of thinking about contacts, accounts, and some of these objects we've been taught to kind of think of the we realized let's think about it as persons, and addresses, and organizations, and then identities, and events. And those can have relationships with a person. And those relationships can change. Someone's phone number can change, their last name can change. And so just having this idea of relationships, really future proofs. This looks probably not super exciting to some of the business folks on the ... who are listening in, but just seeing that the data can end up in a clean way which makes your marketing targeting much more effective.

Eric Dodds:

The people listening in who are technical or looking at a recording are sort of you ... There's almost a sense of relief where you're ... Okay. This feels so much more extensible than the forced hierarchy of your traditional CRM system that relies on fields that can be overwritten or where history is very hard to manage. So you're sort of always having to do gymnastics to get around this archaic structure enforced by a CRM. And so if you kind of take that away and say, "Great. Let's decouple these components into objects that can have relationships." The amount of flexibility from there is just freeing.

Kevin Gervais:

The cool thing is we did all this in Postgres and ran Hasura on top of it. So now you've got an API.

Eric Dodds:

That was so cool.

Kevin Gervais:

It's used in a production setting to be able to use in a sort of cloud to just manage these different multiple databases. So it all together into a GraphQL API and then you can do some really cool stuff with that. So not to geek out too much. So a couple things just in this thing and I'll move on, but just to point out two things. First, you'll see that these two contacts have ... Just even the naming structure. We typically are thinking about people as first name and last name, but sometimes you have someone likes ... prefer the name Maria [inaudible 00:25:51] Rodriguez Maria Santos. And so first and last name doesn't fit that. And so if you actually think about future proofing and international, just allowing the data model to accept something like that, and not force you to jam things into these fields like first name and last name.

Kevin Gervais:

The other thing you'll see here is just the standardization of phone numbers. So just putting everything always an international format. Even though this is a local Ontario based company, that makes it instantly compatible with Twilio and a bunch of other services if it's put in that format.

Eric Dodds:

Yep. Just one note there that ... I don't have any insight into WaveDirect, but if you think about a data format like this building your company with this sort of extensible format, if you are the type of company who wanted to sell to a larger organization or sort of become acquired, this sort of infrastructure is so valuable to bring to the table in terms of its extensibility, its ability to sort of be integrated into other systems. In many ways you can think of it as, A, as certainly as a competitive advantage, but B, almost as a multiplier if acquisition is in your future.

Kevin Gervais:

Well, and even just to that point, it actually ... The data model directly impacts your ability to partner too or if you want to acquire someone too or to float somebody in. So it kind of works the other way as well. I learned this in my last company, but our original data model, we built our app as long as SaaS companies do it a certain way because that's kind of what we did. It was a reporting system. And then as the business changed, we were stuck. It wanted us to go in these different directions. Our data model would not allow us to. And it actually closed off probably millions of dollars opportunity because of the data model being incorrect. And so again, if you can think about things in this more universal way, it makes your ability to say yes, that makes it easier. It makes it easier for you to be able to do that. Okay. I just don't want to geek out too much on that.

Eric Dodds:

Let's keep it rolling. Sorry. I just had to say that.

Kevin Gervais:

I think on the business side people struggle with understanding why that matters. They look at that and like this looks like a different language. Yeah. It's overwhelming. On the technical side, it's hard for you to tell the business side why this matters and kind of make your case. So it's a challenge that everybody deals with. Let's go through some examples of what this did. So once we knew that's kind of where we want to end up, this is some examples of the visibility that we got once we implemented RudderStack in the mix. And so we said, "Okay. What's actually really important the journey is we need to have somebody do a lookup. If they don't do a lookup, they're not really a visitor we care about."

Eric Dodds:

Right. Just to make sure I understand and the audience understands, they're doing some sort of search on the website for service in an area or product or something, but that's the key customer event that shows like, "Okay. This is [inaudible 00:29:21]." That's the moment of truth in the customer journey.

Kevin Gervais:

Right. It can just apply universally to all businesses, but in auto for example, the journey could start with somebody saying, "I'm looking for trucks." As soon as they start filling in what are they looking for ... It's like a search, right?

Eric Dodds:

Yep.

Kevin Gervais:

We wanted that to be a key event that would start the journey. And while visits are important, it's not a key metric. It's the lookup. The other reason why we did that is because WaveDirect, their service has changed depending on what address someone is. Whether it's cable, fiber, or stuff that just make sense to make this the key of that.

Eric Dodds:

Yep.

Kevin Gervais:

It's the start of the funnel, then hover of plan. Did somebody kind of start to hover over certain buttons, click the plan, and then separating out leads to be in market versus out of market? And so that way, you can also use that data to plan. If you have a whole bunch of out of market leads that are in different cities, maybe you expand to that city. And then we look at page visits. And then the other thing was this idea of progressive identification. We use the RudderStack identify function, but progressively as someone fills things out. So as they fill out an address, not only are we log in at the activity, but it just logs and identify that. And the benefit to that is the next time they come back, they're recognized. And so you end up kind of with this.

Kevin Gervais:

These are really from an hour ago. This is the flow. So a person typed in address, went to the support page, quickly, then hovered on plan, clicked on a plan, became a market lead, came back, hover over plans again. Once they've become an in market lead, they get identified. And now every future visit, they have to stay identified. So that identify feature is pretty powerful to be able to continue to remember someone.

Eric Dodds:

I remember when I was working on a growth team at a company and we sort of first implemented this event-based paradigm of customer journey tracking. I look at the screen and I remember the first time I saw the same thing and thought, "Oh, my gosh. This is so powerful. We can learn so many more things. It's just incredible." There's a big password of like customer 360, and full customer journey, or whatever.

Kevin Gervais:

And these people will pay so much money for the service to kind of get to this state, but you can get here now just by implementing some of these things. We leverage cloud for a bunch of this flows. Here's what also is cool. So this is a real lead, actually a really lead came in today. We can see their visibility, but then look at the timeline. We know because of the event tracking, we know what they started from Google. They were to search for this most likely, rural internet coverage Kingsville.

Kevin Gervais:

You'll notice that we can optimize the site very highly for performance, but because the content is customized per area, it ends up getting really high ranking for local searches. And so you can see number two and three in the Google rank is WaveDirect. And so they click the first one and then they kind of started with Google then moved into typing address 10 seconds later. They didn't even read the article. Even though the article actually has a library information, they just want to start the flow. And so the blog post drove people to in 10 seconds just typing address and then 50 milliseconds later, they are now in the packages page because of Gatsby and some of the other optimizations to be done. So then we see, okay, they stayed there, then they went to why WaveDirect. Sad, they didn't just read everything, but we noticed though. Sometimes people are just looking for that mental check mark. Okay. Cool. Like these people, checks out, next. And so that's what happened.

Kevin Gervais:

Basically end to end from Google Search in two minutes and three seconds today, we have a full visibility into what happened and the lead that's in market. They're in Kingsville. And so it's just cool to see that ability to track that flow and really unpack it. And so if you want to optimize things, maybe there's more ways we can optimize the WaveDirect page. Maybe we leave it as it is. Maybe slowing down that purchase process by making this too easy to browse means that they actually ... makes the funnel follow. I don't know. But we can now run those experiments because we can track it.

Eric Dodds:

Right. You were saying they didn't read the article. And I know this is just one customer journey, but even just that type of insight is invaluable in terms of optimizing the customer journey.

Kevin Gervais:

Also scary. That is insanely scary actually. Especially because companies include WaveDirect. Spend money to write various things and time and energy to write high quality articles and yet you can see the consumer today sometimes. It's not always true, but sometimes it doesn't even ... they just want to get through the flow. Yeah. It's a very good point. It's surprising. So this is the tracking, but then in the end, it ends up in the warehouse in Postgres, which is accessible as an API through Hasura. And we also know things like this person uses Cogego, competitor, whoops, which helps the sales team have context. We know the timeline of events, we know the identities the user has typed in themselves. We haven't scraped that from somewhere without their permission, also, because that will become illegal in Canada shortly. It was all really important about privacy. How do we make sure that person is opting in to give us various information. And even by giving an email, we're not sending it to some automated email journey. We want to make sure that it's a human-powered experience. For WaveDirect it works, but for other companies, you might do it a bit differently. We also know that they were using a Windows desktop. So maybe when the installer comes, they'll know ... be ready to work with Windows.

Eric Dodds:

I love that.

Kevin Gervais:

We also know of all the other lookups that are done that are not this lead. And so that might inform where to expand next. So just having that warehouse is really important. But here's kind of the flow of how this all works. As one data source, we have this on-premise system in my sequel. It can't change. And so that gets synced into AWS. So we set up Postgres there and use AWS's data migration service, which is a pretty tricky thing to be able to push that live, within half a second or a second of it updating on-prem, it's pushed to the cloud. And now it's available as an API because Hasura is on top of it. And so we've basically made an instant API on top of their on-premise system. And then we can use that for a bunch of things. We also store in that Postgres database, the events from RudderStack and other slated clean records, too. So we showed that universal data model that ends up in that Postgres database again, with Hasura on top. We have Sanity which is your headless content lake. So that's where you store these basically content blocks.

Kevin Gervais:

So if you look on the web direct site, there's various parts of it like the title, and articles, and things, but the team can write those articles inside Sanity and that auto triggers a build process depending on what changes they made. Sometimes it takes 20 seconds, but that flow on the left-hand side is about getting data in. If it impacts the site, design, or content, we use Gatsby cloud which then pushed us to Netlify to be able to build the site. And that's how we got the performance to be where it is where, by the time someone goes to the site, their browser is not thinking about anything. They just are rendering HTML that's been kind of pre-generated for them. So that's that left side. And then RudderStack becomes pretty key.

Kevin Gervais:

Once the site is live, RudderStack is collecting events. Again, it's pushing that those events back to so Touchless. On our side, we have a product called EXO, which is this data layer that just brings these various sources together and orchestrates kind of this flow. So RudderStack, when it collects things, it kind of goes back to us and we can then kind of close the loop. So in this case, like auto generate page is based upon some of that data or update slack with a notification. Those are kind of things that we do with ... some of the stuff that RudderStack would do. But the flow is once we get to the site being built, RudderStack comes in to track events and then pushes it to various places.

Kevin Gervais:

That's how we started with RudderStack. Where we're going with it is to use RudderStack for more things that it can read from and more flows. But just to understand the web experience, this is how that flow ends up. And what's cool about it, because RudderStack is made so you can connect multiple sources and connect multiple destinations, it future proofs this whole flow. If we want to push things to HubSpot or whatever, it's not a huge lift for us to do that because we've got this kind of plain, this sheet of glass, that just makes it really easy to put other things in front of it. And then we then push from RudderStack as well into marketing systems. So we're looking at MoEngage as a potential flow. We tested that experience, but that's some of the goals for the rest of this month. And then we also push this over to Statflo which has a programmable messaging layer that frontline staff can use to talk to customers. And so that way, they end up with some of that context that's in the data layer so they can have more informed discussions. So this just helps show the pipeline that we ended up with.

Eric Dodds:

Sure. I know we want to leave some time for Q&A. So I'll be brief, but what I love here is if we go back to data interface design, what we see is all of these modular components where the data is flowing in and out, which is just incredible. You're not bound by the end destination or by the source even and the data can flow freely, which is incredible.

Kevin Gervais:

And on that point before RudderStack, we were on the web experience hitting all these various things ourselves. And so that would actually incur more costs and more time. And so if we were going to move in a different direction, we would sometimes have to say, "Sorry. I can't do it right now." The data becomes empowering. I'll go quickly, and people who want to rewatch this you might have to a couple times. I'll go through some of these slides just to kind of highlight some of the stuff that we set up before we go to questions. But so just you can see here RudderStack is collecting events and then pushing to various destinations.

Kevin Gervais:

By centralizing it this way, it dropped load time 50% lower than what we were already getting with Gatsby alone. So Gatsby got us down to one second, RudderStack took it further. And then we're cleaning data at the point of entry and then you're pushing two destinations that can change. Now, here's the impact, here's some of the cool stuff. So you can see the change in page rank position happened almost instantly, actually within a week of the site launching with this new flow, and Gatsby in the middle of it kind of this ... kind of headless workflow.

Kevin Gervais:

You can see the impact that Google's new core vitals update does to ranking because they really emphasize page experience. So the real benefit though is ... Again, you can see once it went up, it stayed up pretty consistently, but RudderStack kind of came in when we kind of hit a ... We had already launched the site. And then we started realizing that we were hitting a plateau in how good our performance could be. So just that impact of dropping the page speed. Again, you can see the impact that had on average position for various keywords. And then here's just more granular. You can see this is for a specific kind of keyword pattern like rural internet. So the starting point before this redesign, we're seeing a click-through rate of 2.1%, an average position around 20 in the Google rank, and now it's around six to seven. Some cases like you'll see here in the picture, it's number five or number four and five with a click-through rate of 17% and that's within 45 days. And RudderStack was brought in again just halfway through that journey and it allowed us to kind of move up to a higher tier.

Eric Dodds:

Amazing.

Kevin Gervais:

This is seeing the impact in core vitals. So before RudderStack, you'll see this. So previously, the site this is before even Gatsby in the Jamstack kind of experience. You see that the reason why the ranking was around 20, 30 position was because it was failing core vitals on mobile and desktop and the page experience score was very low. And so once we did the redesign, it was pretty good. But just getting that extra kind of bump, you'll notice in this graph, came once we had added RudderStack to the mix. So it's not like RudderStack alone will kind of just make this experience that much faster, but once you do add it, it just gives you that extra edge. And then you can see here the impact it had on percentage of good URLs in the page experience score in Google Search Console. Prior to the redesign, there's around 25% that were considered good. We got about 70% hit a plateau, added RudderStack and now we're at 98%. I'm still working on that last percentage point, but it just shows the impact when it can have if you do improve the performance.

Eric Dodds:

Amazing.

Kevin Gervais:

And these are just some of the comparisons to just top 10 ... in the top 10 keywords that we care about compared to some very large companies. Just the number of keywords that WaveDirect has been able to get top 10 position. In comparison before this launch and before the core vitals update, they were kind of comparable by the 30 mark in terms of the number of keywords in the top 10, but now it's I53.

Eric Dodds:

Amazing. One thing that is such a good reminder, and there's so many things here even just the site speed, even just outside of the data components and the flexibility around that is that it's easy to forget that a rigorous focus on site performance alone can increase business results. Moving from 20 to the average position of 20 to six, that's going from the second page to the first page, which is game-changing literally. We see that in the numbers, obviously. And don't get me wrong, there are a lot of components to SEO. It can be a very complex subject, but if you're having the foundation where the technical side of it is completely buttoned up, is the best foundation on which to really sort of unlock the potential.

Kevin Gervais:

Yeah. It becomes the enabler. And then you don't have an excuse as to why you can't. And with this one, we wanted to push the boundaries. What I used to do is you get to one second load time and stop there because who gets faster than one second load times? And then once we got to where it brakes 0.5 seconds or 0.4 seconds consistently. It's like, "Huh." First, I didn't know that was possible. Second, Google seems to care. And then it's also impacting click rates. Some of these is just crazy to see. Because of the performance change and the page experience score, Google highlights it more. And so you're seeing for some of these things, some of these keywords, click-through rates, 30%, 20%, 15%. It's just amazing to see. This was not the intent. So it was worth it. We knew that we wanted to get performance up, we knew that it mattered to humans, but it shows Google ... because they want to make sure they're showing relevant results, they care about it.

Kevin Gervais:

And then this was the core vital score. This always changes depending on where you are and whatever, but typically we're seeing kind of results around this like 0.3 seconds depending on when stuff loads. So Jamstack doing this flow where things get pre-generated alone with Gatsby can get around one second load time, but RudderStack first prevented us from adding extra bloat because most times you just add an extra pixel to your site.

Kevin Gervais:

The crazy thing is the slowest performing thing on the site is actually the font. It could be even faster. It was the font and it's the Google Maps API and Google Analytics. And so those are the things that actually can add half a second to a second to the load time. And so just by having it in the mix, it allowed us to get that better performance, and especially this is more for [inaudible 00:49:18] where we care about Touchless, but building stuff for people on the autism spectrum. And on the autism spectrum, what happens as we use things is that you expect a certain kind of feedback loop. And it can be really frustrating when you don't see that 50 millisecond kind of response time or 100 milliseconds. It breaks the flow and it's intense. It makes it really frustrating to use the web. And I've always hated how the web has excluded people with that. And there's so many sites that I just get angry going to. Probably most people don't have that reaction, but for me, it's, "Ah!" It's just like, "Why does it need to happen?" And so it's not just about hitting the numbers, it actually impacts people when the performance is this way.

Kevin Gervais:

And then just to cap off quick. I always say quick, but it's really cool stuff. Why we did it this way is that designing the data first enabled creating richer customer experiences. So this is an example of that. We talked about that Stat flow product which is this flexible thing for people to communicate with customers over text especially in regular markets, but it requires good data. Once you have the data, you can put that in front of the frontline staff. So in this case, these context widgets or maybe on-prem data, or things. So because of Hasura and because of RudderStack kind of communicating with Postgres, we're able to enable data to get put into the front line. And so we're progressively adding more and more of these widgets and stuff over the coming months. I can see Joanne is part of WaveDirect here. It's coming this way. But it's cool. You can enable this type of experience, but then the other piece is just the data really should be an enabler across the business. So if you design it first, yes, it slows down the process a little bit at the beginning and it's a little boring, but it can really make things more efficient.

Kevin Gervais:

And we were able to do some of these transformation work at 80% less than what a traditional transformation would cost because we're able to design the data and be intentional about it. Again, not knowing that that would be the outcome, it was interesting to see how ... now this perfect set of plumbing that exists now that really wasn't in a mature state two or three years ago, is a real enabler if your data is done properly. And it can make interfaces faster to build and it can be used to connect two frontline experiences and make better decisions. Yeah. That's kind of just a bit of an overview of what happened.

Eric Dodds:

I love it. I'll just reiterate sort of one point and then we have a couple of minutes for questions. Incredible case study. One of my big takeaways here looking at the data that was in the tools that the frontline employees are interacting with customers or potential customers, that's the dream. It's being able to get that real-time context as you're helping a potential customer. That is the dream. If you can do that, it creates happier customers, happier employees, and better business results, which is just truly incredible.

Eric Dodds:

Well, we have a couple of minutes for questions. Feel free to raise your hand. You can click the button at the bottom of the Zoom window. Type something in the Q&A. I'm also happy to unmute you if you want to raise your hand and you can speak your question as well. One question here. Starting out, Kevin, how long did it take to get the basic infrastructure operational, sort of starting, I mean getting all the plumbing setup?

Kevin Gervais:

Well, we did it in a piecemeal way. The plumbing, the pipeline that you see here, kind of all came together right at the time around the site being built within those two, three weeks.

Eric Dodds:

So just a couple of weeks?

Kevin Gervais:

Yeah. Just with an asterisk that it's going to change. I think that's also the power of RudderStack. Is that we can swap things out, we can add new transformation steps. What the schema is today, maybe we realize there's a better way to do it. I think that's okay. So the way that we approached it first starts with what they had. So we took the data as messy as some of it was. It was like a manual cleanup process to begin with. Put it into Postgres, kind of looking at it, start to understand where it was, and then that was job one. Let's see what clean data looks like and then we could have the numbers in front of us, then it was about serving customers and bringing that in. So we kind of phased it in. It was like, "Let's get an awareness of where we're at. Let's get an awareness of what's the pulse and then let that inform how we design the rest of this." And we knew that the one truth out of all of this was that we had to have a plane in the middle of it. We don't really know or care about actually which ... what was the source and what was the destination. We just knew we wanted a middleware layer.

Kevin Gervais:

And then eventually, we kind of made more informed decisions as to what are the must-haves? Okay. Using Gatsby, using Hasura, using Postgres, those sort of pieces. And then we just started to add other stuff later. MoEngage investigating that is just from the past few weeks. And maybe not everyone wants to approach it this way, but you can do it in an agile way. You don't have to have it all figured out. And it is possible, because of things like RudderStack, to just spin up these things and click a button and RudderStack is connected. There's no pipeline. RudderStack is the pipeline, right?

Eric Dodds:

Sure.

Kevin Gervais:

I don't know if that answers the question.

Eric Dodds:

Yeah. That's great. I think we have time for one more question here. How much work is involved in updating the site with new information? And I think this may be in the context of if someone searches for something new. How are you deploying new content on the site? What does that look like?

Kevin Gervais:

Yeah. Again, this is an iterative process. Depending on where you're updating the site, you might want staff to very quickly type something in hit and button and it always goes up and it's like there's not much formatting to it. There's other types of pages like say the plan page or maybe some key kind of area where you want more of ... more controls in place. So that way, people don't just push up a bunch of garbage accidentally. But the flow of this, because of using something like Sanity in the mix, is they can log in to that, make a change, hit the publish button, and in under a minute, sometimes 20 seconds later, a new page is created or changes live. So you can decouple these things.

Kevin Gervais:

Before you used to have to build a site all at once all the time, but now you can do things incrementally. Yes. It should be very easy. And then in areas where that are not connected to things like Sanity, I think the job of any integrator or person building with this is to just connect more things into places like Sanity that enable the marketing team. So one section, for example, in the WaveDirect site just because we were rushing to get it out, we made one of the things just kind of code in React knowing that we're going to replace it with Sanity later. Keeping on top of that type of stuff is good to do. You can sometimes get stuff out and solve the problem and then start to make sure that everything all is connected, and I and I think that's okay. But yeah. It's possible to make it so you don't always have to have a coder involved in managing this stuff.

Eric Dodds:

Sure. Very cool. Well, Kevin, thank you so much. What an awesome story. What an amazing set of use cases. We're overtime at the hour here, but thank you to everyone who joined. We'll send out a recording. Keep an eye out for additional tech sessions in the future. We'll have you back on, Kevin, for sure to learn about other cool stuff you're doing and other cool stacks you're building.

Kevin Gervais:

Yeah. Thanks for having me. This was fun.

Eric Dodds:

Yeah. All right. Thanks, everyone.

Kevin Gervais:

Okay. Thanks.