430: How to Get Marketing Attribution Right

[0:00:01]

Steli Efti: Hey everybody, this is Steli Efti.

[0:00:05]

Hiten Shah: And this is Hiten Shah. And today on The Startup Chat, we’re going to talk about marketing attribution.

[0:00:12]

Steli Efti: Yeah.

[0:00:13]

Hiten Shah: There’s a very specific reason why. Steli share why we’re talking about this.

[0:00:18]

Steli Efti: Yeah. So I think talking to a ton of marketing leaders recently, we’re looking to hire leadership on the marketing side of the Close. And so I’ve been reaching out to many super senior people, that work in really amazing organizations to ask for advice. And one thing that really surprised me and it didn’t, because on the one hand, I know that every company’s full of shit and all things always look better from the outside. But then again it’s good to be re reminded. But one thing that surprised me still in the moment, was that a lot of these people would share like the problem of running marketing organizations for really big SAS companies. And we were talking about the type of companies that are like 20 to 40 million in ARR, right? So much later scale. And one of the things when I asked them about like big projects over the last couple of years, things they wish they’d known and done differently when they first joined. The surprising thing is how many of them said, “Wish we’d had a data and attribution models figured out earlier.” And some of them even like they worked at analytics and metrics SAS companies, were like, “It’s really embarrassing. I would never admit this publicly, because we’re an analytics and metrics company, but we don’t have our metrics analytics shit figured out. Our stuff doesn’t work, our attribution doesn’t work and we’re still working on fixing that.” So it was surprising to me to see, I knew that early stage, a lot of companies don’t get this right, but I felt like at the 20, 30, 40 million, I just assumed that more companies would have marketing attribution figured out or prioritized. And it seems like it still was a struggle for many. So I felt like, maybe this is a topic. Maybe this is something we should bring up as advice for SAS companies. So that they don’t get to this level of scale and still struggle with it. And you’ve built a number of analytics and metrics companies, you’re pretty famous in this area. So just wanted to like quickly pick your brain and share it with the world in terms of like what should startups do to get marketing attribution, right? What do they have to do and why is it so difficult for most of them to do it?

[0:02:43]

Hiten Shah: Yeah. It gets more complicated as you scale. And you start going away from channels that are directly measurable, which already has marketing attribution issues. So directly measurable would be people visiting a website, right? And you knowing that those people converted and signed up, purchased, et cetera. So then you get into like basically channels like billboard ads, TV and things like that, where a lot of the things you’re doing, your goals are oriented around awareness. And then you have to think about how to measure those things and include that in your marketing mix and then figure out how to attribute it as well. So I have this like … Man Steli, like you couldn’t have picked a better topic that I can go so deep into, down to the how you track every single visitor appropriately so you can do the attribution appropriately, all the way up to or in the middle would be like your marketing stack. And all the way at the top would be like how do you make decisions based on this information? And there’s no right answer. I know folks, even at very large companies that are very big, ad based companies, which I’m not going to mention that struggle with this as well internally. Regardless of how like sophisticated large, multibillion the businesses are. And so this is one of those where it’s like do the best you can, at the point that you’re at right now.

[0:04:47]

Steli Efti: With what you have. Yeah.

[0:04:49]

Hiten Shah: Yeah. With what you have. Now hear the famous thing. Now that being said, and this is a big, that being said, if you know better, do better now. So I know better. I know way better in my businesses, because of my experience as well as like my conversations I’ve had with so many people across so many sized companies, that we instrument as early as we can. Here’s a few best practices. So we instrument analytics as early as we can. Everything in terms of the events. So things like page views, conversions, things like that, they always have a date of when they started being tracked, because believe it or not, that’s a very important thing. Because when you look at graphs and things look off, oftentimes you were just not tracking the thing early enough. So that’s one big piece of like the date you start tracking. Second piece of it is audit the crap out of your tracking. I would almost say audit the crap out of your tracking as often as you can possibly like handle it. So if you could handle it with every deploy that you do to your marketing site, or every update you do to your marketing site that you feel like might tweak the script, do that. And same with like your web app, because a lot of times I’ve had people update the web app, do a redesign and somebody forgot the JavaScript to track. So nothing was being tracked. And this gets worse when you’re like you’re really great at optimizing and improving things that you’re shipping a lot. Optimizing, onboarding, changing up your marketing site, improving the application. This is where the problem gets even crazier and that’s when it’s like, crap like our tracking is hose. Do you know how many times I’ve heard our tracking is hose, we have to start over. Or oh no like we missed this call, this tracking event or we forgot that this needed to be tracked and it’s not being tracked anymore, right? Like there’s so many of these things that keep coming up constantly, that I keep seeing and keep worrying about and see businesses worry about. So number one, know when an event was tracked or when you started tracking. And what I mean by that is, if you start tracking today and it’s like noon, that doesn’t count, because most analytics tools won’t show you from noon today, show me the data, right? So that means tomorrow midnight, right? Onwards is when the tracking actually you can trust it and it started, because today was only like a half day, right? Those are the nuances that people don’t get right. I know this has little to do with attribution, we’re not even into attribution yet, but like those are a couple of best practices that mess up any tracking and any attribution. I’m going to go on attribution for a second and then like I’d love to hear some thoughts from you, because like this is just the bane of some marketers existence. And they just do the best they can and that’s all you can do. And I think there’re some companies out there around helping with marketing stacks and stuff like that. So you can look at that. I’m not one for like talking too much about tools these days, even though I know people love it. But one thing I really think about a lot is, do you know the people who come to your site and convert on the first visit? Because if you do, then that’s one bucket. Then do you know the people who come to your site, who didn’t convert on the first visit, but converted on the second? Great. Those are other ones you put aside, right? So my game on attribution has a lot to do with, they got to the site because there’s a lot other attributional awareness on that, but I’m not as interested in that. But they got to the site and I know that they got to the website. And once they sign up, I know that they signed up and now I know who they are, and I know all the other times that they came to the site before they signed up. If I know that, then like my goal isn’t to get fancy with attribution. My goal is just to know and start filtering out of the pool. So the easiest filter is, they converted on the first time we saw them that they came. Okay, great. Let’s account for all those. Second one is two times. Okay, great. That’s a little more difficult, because two times mean there’s two channels. Okay, so what are all the second visits that happened that didn’t convert and what are all the channels they came from? Okay, great. Now I know those, right? And and then it’s like what’s the second one? And I just keep going. Like how many visits and what channels are what I’m looking to figure out and I just filter out the easy ones. And the easiest one is first one channel, right? First Channel, first visit, they converted. Second one is two channels. Second, first visit. I know all the channels. Then what are the second visit channels that brought people and which ones of those lead to conversion? Great. Any of the pool that didn’t convert yet? Pull those out. And I just keep going. And so this is a very tedious task. The analytics tools do not let you do this. Most people I know at some scale post 50, 100 million in revenue and ended up building the thing themselves, because they end up having to. And building something themselves means like they’re building their own JavaScript. They’re doing their own data warehousing and that’s how it works. Believe it or not, we’re 12 people in our company. We’re technically pre-revenue at FYI and we have these systems, because like we know better. And these systems took a long time, like way longer than I would admit to get built up in the company while we were building the other stuff. And the project has been called my pet project multiple times. And it’s one person working on it who happens to be the head of engineering that worked with me at KISSmetrics. So we know it and we know the issues. And I wish we could take something off the shelf. We use off the shelf analytics as well, multiple tools and they help us make decisions for now. But like ultimately we know that there’s a bunch of tracking we need to do, around how our business works and how we want to see some of the data that none of the analytics tools do today. And that’s okay, because like they do what they do, right? And more recently, and I’ll stop my rant. More recently, the analytics tools have focused on product analytics and so they’ve gotten away from actually being able to help you track your marketing properly. That’s a shame is the way I look at it.

[0:11:40]

Steli Efti: Yeah. I agree. I think that in the early days of this and KISSmetrics [inaudible] were pioneers in the space. There was a lot of, I think effort put in to build an analytics products that would make you understand where is traffic coming from, how is it converting, why is what happening? It was very marketing top of funnel focus. And I don’t feel like the products that were out there, like solved that problem. And I hit in some like elegant and universal way and now everybody’s rushing towards more product analytics, which is also super important. But that first product problem is still around. It’s crazy and there’s not a simple or single solution that lots of companies could use that fixes this for them, which is fascinating. All right, so I mean you are a genius in this space and there’s not much more that I can add to any of this. But the one thing that I’ll say, I think the most important thing I am taking out of this, which is a very universal lesson, but we have to apparently relearn it again and again in all different situations, is how we started this episode. Do what you can, wherever you are, with what you have right now, right? And so I feel like the advice that we have, is that the worst thing you can do is one of two options. One is to say, “Well, we don’t have x.” Whatever x is, an engineer developer, resources, money for marketing software, or whatever else. “We don’t have something, hence we cannot do attribution. Hence we’re completely going to ignore it.” That’s really bad. And the flip side of that coin is also equally bad, which is we need to get this perfect and right. We have never dealt with this. I mean, you are in a very different spot of saying, “I’m going to get distribution right from the get go.” Because you have an amazing amount of experience. But for most companies and most founders saying, “I’m going to try to make this my number one priority and get it perfect and get it right.” Means probably either procrastinating way too long on doing anything, because it’s like, this is going to be this overwhelming massive project. Let’s push it down the road or it’s going to be, you start off and you run out of steam or by the time you’re done, you realize you spend too much time, too much money, too much effort, and it’s still not right. And so those two strategies are probably really bad ideas. The alternative that I think we want to share here is, think about it as improvement versus perfection, right? Think about like what do we understand about attribution today? How can we understand a little bit more about it tomorrow? How can we continuously add to our understanding and add to it? And would you mention what’s just like, here’s a simple way to segment. Let’s just even look at first time visitors that convert. That’s one group, that’s a really important group. Let’s try to understand there’s just one that one group. And once we got that, let’s try it on and send the next most important group. That’s a beautiful way of going about it. I remember one time talking to the marketing genius and going, “Wow, we don’t …” This is many years ago, four years ago or something. I was like, “Wow, we don’t have really the attribution model to really understand what marketing that we do has brought people to the trial sign up.” And that friend was just like, “Well you call people, right?” And I was like, “Sure.” He’s like, “Why don’t you just ask them?” And I was like, oh my God, face palm moment. I’m like I’m here screaming at people. You should call people just like you should talk to people and learn everything you need from your customers. And here’s me at the same time complaining that I don’t know where these trial signup people came from, although it’s easy for us to figure out. And then we did. That very day, we called a bunch of them and for the next couple of days and this was not perfect, but we got a lot of insights by just for a couple of days calling. 100 or so trial sign ups and asking them and having conversations and seeing patterns, right? Which needed no software and no analytics whatsoever. Just needed us having a bunch of conversations. So this is a really important part of how you build your business. Don’t get overwhelmed, don’t push this problem down the line. Just continuously chip away at it. Try to get better and better at understanding what in your marketing works and why and how. And don’t wait till you’re a multibillion dollar company to then hire a whole department and give them three years to work on this. Hopefully you’ll get there much, much sooner.

[0:16:20]

Hiten Shah: Yeah. And just to be clear, like we wouldn’t have worked on the analytics. We didn’t start the project until I had it on the list, but we didn’t start the project until we had a product market fit. And we knew what this business was going to be and we were going to like pursue it like seriously. And the reason for that, is all the measurement that we’re going to do, and we are a very high velocity business. The business we have with FYI is a freemium business. And so the analytics matter that much more. We would have tried our best to get away with like tools in the market, whether it’s mixed panel, amplitude heap analytics, or the number of other ones there if we could have. But what we realized is that there’re certain things that they just don’t track very well, specifically around this kind of stuff, marketing attribution and things like that, that we’re just going to build it ourselves. But again, we know what we’re doing. So it’s a little bit different than most people who would do it, but we still waited even though we didn’t know what we’re doing and we knew we did it. We waited till we knew this was a idea worth pursuing before we spent the time to build the custom stuff.

[0:17:26]

Steli Efti: Beautiful. All right. This is it from us for this episode. If you enjoyed it and you’ve not done this before, please do us a favor. Go to iTunes, give us five stars, give us a nice little review it’ll make us more visible. More people will listen and the Czech community will grow until next time. This is it from us. We’ll hear you very soon.

[0:17:45]

Hiten Shah: So yeah.

[0:17:45]