How to implement and measure the success of multi-touch attribution models with Dan Mcgaw

Anna Shutko:

Welcome to another episode of the Marketing Analytics Show, the podcast that helps you get better at marketing analytics.

I’m your host, Anna Shutko. Today, our guest star is Dan McGaw, a marketing technology professional based in Oregon.

Dan is the founder and CEO at McGaw.io, a marketing technology and marketing analytics agency. He’s also the founder and CEO of UTM.io, a leader in UTM campaign link management and campaign data governance. Prior to that, he worked as a head of marketing at Kissmetrics and founded the National Association of Marketing Technology. He also wrote the book Build Cool Shit, which is your blueprint for creating a modern marketing tech stack.

So, today we’ll talk about multi-touch attribution models, and from this episode, you’ll learn:

What you should consider when starting to implement a multi-touch attribution model
How to optimize campaigns after you’ve started to implement this kind of model
What reports you should have in place to see how effective your model is
Common mistakes marketers make when it comes to implementing the multi-touch attribution model
I hope you’ll enjoy this episode.

Hello, Dan, and welcome to the show.

Dan McGaw:

Thanks for having me. It’s great to be here.

Anna Shutko:

Awesome. I’m really looking forward to this episode. So, let’s start with the first question. What is your definition of a multi-touch attribution model, and what, in your opinion, should marketers consider when they start implementing this kind of attribution model?

Dan McGaw:

Yeah, and really, really good question. So, in most cases, multi-touch attribution modeling is going to be when you’re trying to take your advertising spend and really be able to understand your return on ad spend, right?

So, as you’re spending dollars in Google AdWords, Facebook, LinkedIn, or even SEO, and a bunch of different channels, you need to be able to measure those conversions and then attribute those back to those channels. Because as a user, I’m going to come to your website, but I’m going to hit a Facebook ad, I’m going to hit a Google ad, and I’m going to hit an SEO. And we need to distribute the revenue that I make as a company across each one of those touchpoints.

So, to distill it down, really, multi-touch attribution is the measurement of each one of those touchpoints, applying distributed credit from that revenue conversion to each one of those channels. So, that way you’re not counting one conversion across all of those. And the biggest thing with multi-touch attribution is people need to understand that it’s meant for optimization. So, it’s not meant for reporting. It’s meant for optimizing advertising campaigns, and they need to make sure to have good reporting. They need to have good data coming in. So, if you don’t have good data, you don’t have a good taxonomy, it doesn’t matter what you do. You’re not going to be able to be successful.

Anna Shutko:

Right. That sounds awesome. I really love the focus on ROAS there, and also, really good points. I absolutely love that it’s meant for optimization, and good reporting definitely has to be key. Now, if we talk more about optimization, could you please tell us, how would you optimize your campaigns based on this multi-touch attribution model? Any tips you could provide to our listeners?

Dan McGaw:

Yeah, of course. I mean, I think the biggest culprit that we see now with multi-touch attribution models is that a lot of times when people are using last-touch attribution, they saw that their brand AdWords campaign was driving all of their conversions. So, they just dumped money into that. Come to find out, once you start to run multi-touch, you start to find out that that’s not as valuable as you might think. And we see a lot of companies distribute that spend that was happening on their brand key terms and put that into more of the middle of the funnel, where they actually can make more money because the last click was not actually what caused that conversion. I think that’s probably the one that we see the most common.

But that being said, once you have a multi-touch attribution model, of course, you now actually know the return on ad spend on a per-campaign basis. Depend on how crazy you get. Sometimes, you can get it down to keywords. Sometimes, you can get down to all kinds of other information, but you start to realize, “Oh, this campaign sucks. I should shut this off.” And then really, that’s what it’s about. Right? It’s seeing those campaigns that suck and turning them off and then taking that budget and putting it into the campaigns that are working.

Anna Shutko:

Right. Awesome. And now, if you could tell us a bit more about the actual optimization and maybe how to identify whether the model is actually successful for your optimization. So, could you please tell us what you have to have in place to make sure the attribution model works effectively and when you are starting to do reporting because you mentioned there has to be good-quality reporting in place what data sources would you typically use?

Dan McGaw:

Yeah. And fantastic question. I mean, I think let’s start with the data points there. I mean, when you think about all your data sources, anywhere that you’re spending money on marketing should definitely be a data source. You need to be able to collect the advertising spend, the advertising channels, all of that campaign information. You need to be able to pull that in and be able to really report on all that stuff. So, you do have to have all of those channels tracked, and even channels that maybe don’t have ad spend in them. I see a common thing. One of the things we built a custom model for, for a client recently, is they’re spending a ton of money on direct mail. A million pieces a week are going out to their target audience.

Well, that direct mail still has a cost, right? It might be offline, but we’re able to collect that information, still, and tie that to a user identity and then be able to track it. So, you really want to make sure that you get in all of that information so that way you can really have all of your channels tracked.

Now, once you have that data, of course, you have to start then, of course, processing that through your attribution models, and then, of course, being able to get your reporting. But when you think about your attribution model, the first thing you have to start thinking about first is your business, right? The first question is, well, how long is my sales cycle? How long is my buying journey? Am I a subscription business where I make more of my money from the lifetime value compared to that first purchase?

And you have these two things. You have what’s known as a look-back window, and then you also have what’s known as a look-forward window. In a look-back window, you’re really looking at, how far back am I going to take into consideration all the touchpoints that I’m going to allow credit? So, that look-back window should be based upon how long your sales cycle is. So, as an example, at my company, UTM.io, which is a SaaS business, our sales cycle is roughly 60 days, right? So, I usually use a 60-day look-back window to be able to take in that amount of period so I can really apply my models to that.

Now, with that being said, my product, UTM.io, it’s a campaign link generation tool, so it’s all about making UTM links to track your campaign. What we have to take into consideration is our first conversion, $169 a month, is not the lifetime value. So, if we spend advertising on somebody to acquire them, we’re spending upwards of $500, $1000 to acquire a customer. So, if I just looked at that first touchpoint, $169… or, excuse me, that first conversion… naturally, my ROAS is going to be upside down, right? It’s not going to make any sense. I have to take into consideration my lifetime value.

So, when you have a model like that, there are only a few tools out there and you have to build these, many times custom that have a look-forward window, which enables you to track the revenue repeatedly, because there are multiple conversions of revenue, into the future as well. And that’s where the look-forward window really comes into place. And those are two major parts of your model. Are you going to have a look-back window, and how long is it going to be? And then, of course, are you going to have a look-forward window, or if you only are going to look at that first conversion as your revenue number?

Now, with those things being said, you then apply your model on top of that, whether that’s time decay, linear, point-based, right? There’s a lot of different models. We always tell people to start with linear, right? Just do a straight linear model, which means we distribute the revenue evenly across every single one of those touchpoints, and then let’s start optimizing based upon that.

And the reason for this is because if you start out with a custom model, well, now you have somebody who’s just trying to tell one story about the data compared to actually listening to the data, and you have to listen to the data first. You can’t say, “Well. I feel this way about that thing, and this thing should be double the credit or time decay.” Just start linear, run that for three to six months, and start optimizing.

But if you’re really advanced and you’ve done this 10 times before, yeah. Maybe you can start with a time-decay model or something of that effect. But in my opinion, one of the best ways to look at this is actually channel-based. So, based upon the channel, how much additional credit should that channel get based upon time decay and some other variables, which would then start getting into really custom. And the reason why I would say this, is as an example when you have a touchpoint, right? That means that somebody got touched by your advertising or touched by something, and you have somebody who goes view-through conversions, right? With retargeting. Well, you’re not able to measure a view-through in multi-touch attribution because he never clicked. Right?

So, one of the things you might do is that when you do get clicks from your retargeting campaigns, you might actually increase the amount of attribution which is given to those channels just because you can’t see the view-through. So, the thing that I just would stress, start with linear, right? Get used to linear. Start turning off the campaigns or the keywords or the ad groups that aren’t working and then start to optimize after you’ve gotten a chance to get more familiar with how this works, or you can bring in an expert who’s going to be able to really help you.

And the last part that I’ll just add to that, which I think is really critical, is going back to the point, if you have bad data, this isn’t going to be effective. So, you have to get all the advertising data out of the advertising platforms. So, Google, Facebook, all that stuff, but you also need to make sure that that data in there is being sent to your web properly, correctly, using the right campaign tracking codes. So, I talked about UTM.io earlier. This is a campaign, a data governance product, which we built for free. So, anybody can use it. It replaces your ugly UTM spreadsheet. But if you have good UTM tracking, that’s going to make it so that as you measure those touchpoints, when they hit your website, they’re going to match up to whatever you have in that advertising campaign, because if your UTM data does not match what is in the advertising platforms, when you go to try to mash it all together, those touchpoints aren’t going to matter because you’re not going to be able to map them together. You’re not going to be able to see everything. And without that, it doesn’t matter. You can’t worry about models. You can’t worry about linear. You can’t worry about anything else. So, clean data is extremely imperative.

Anna Shutko:

I really love the focus on the cleanliness of the data. And also, you mentioned really, really good examples, and touching on the client is, of course, very important. Now it’s a great example of how the whole process works, and touching on that, especially because it has so many moving parts, could you please tell us more? What are some common mistakes marketers can run into when trying to implement this process, when they’re getting started with the first on linear, for example, models?

Dan McGaw:

Yeah. I mean, the first thing that I think I see problems run into is they treat multi-touch attribution like it’s the holy grail, right? Like it’s going to change their entire business, and they’re now going to become a legend at the company that they’re at because they rolled it out. So, the first mistake that I see companies doing is, of course, set your expectations correctly. Just like everything, it takes time. So, you’re not going to see your massive improvements in your first 30 days. You’re not going to see them in your first 60 days. It’s going to take you 90 days to six months to really understand how you’re supposed to optimize this stuff. So, I would definitely say making sure that you set your expectations correctly.

Now, there are obviously multiple different ways that you can also set up multi-touch attribution, and I think that’s one of the hardest parts out there for companies, is that they don’t think enough about what is the right way for our business to set up multi-touch attribution. And if we think about that, you, of course, can do the custom way to build it. So, I’m dumping data into a warehouse. You could also buy a product, but you have to make sure that you choose the right product or make the right model for yourself based upon your business, and we see a lot of companies basically just go out there and choose a tool and hope it’s going to work because it provides attribution and the sales rep said they’re going to be able to solve their problems. But in reality, if they would’ve done the digging to understand that product, matched it with their business, they would have found out that the data architecture of that product does not work for them.

And I’ll give a simple example, right? One of the great things about Supermetrics is that you can pull data out of that and basically send it to other places, right? So, you can pull data out and then you can go create custom models. Now, this is super helpful because you can build your own custom models inside of your warehouse or any other type of tool that will enable you. So, this super, super helpful, but now you’ve got to start thinking about, am I capable of building all these data models myself and mashing them together? So, then you might wonder to yourself, okay, maybe I use a product, right? Maybe I buy a product like Attribution App or Rockerbox or Windsor.

The problem is that a lot of companies just see one of these services, like its UI, and then goes, “I’m going to buy that one,” without ever understanding the way that that platform stores your data and then how that platform’s data storage model affects your business. And the big thing that I’ll explain there is I talked about look-forward windows. Attribution App is one of the only products that offers a look-forward window. Meanwhile, Rockerbox provides extremely good flexibility around having a custom model, and then, as well as great look-back windows, but their look-forward window is treated a little bit differently. And then you have other companies, like C3 Metrics and all these other brands, and each one of them stores data differently.

And then on top of that, the reporting layer is also different. The way that they treat timestamps, the way that they treat their date ranges, right? There’s a lot of things that you really have to take into consideration, and once again, it all gets back down to that core data and making sure that you match your business and the way that your business runs with however that attribution model service or however that custom model you’re going to build is ultimately going to work too. So, I think that’s the… Make sure your expectations are correct, that it’s not going to be an overnight solution, and then also make sure that your expectations are correct in the fact that you have to take some time to make sure you choose the right tool.

Anna Shutko:

There was a very, very, very, very good descriptive answer. Now, let’s talk more about the awesome resources you’ve created for marketers so they can succeed with attribution models. So, you wrote a book on building a marketing tech stack, and it’s called Build Cool Shit. Could you please tell us more? What’s it all about?

Dan McGaw:

Yeah, absolutely. A long time ago, we saw a huge problem is that building a tech stack has really changed over the past five to 10 years. With the proliferation of customer data platforms and modern analytics and new marketing automation tools, you really have to build a world-class tech stack to be successful in today’s marketing. So, I took all the years of experience that I have, and I’ve been doing this for over 20 years. I took all of that stuff, I tried to distill it down into a really simple book to help people understand how they need to think about the marketing stack? How do they need to integrate things? And then I also wrote a case study in there, using one of our clients, to be able to help you understand, how do you do lead scoring? How do you do personalization? How do you do any type of ongoing personalization and reporting?

If you go to McGaw.io, you’ll be able to find it on our homepage. You can get a free copy of the book. It’s called Build Cool Shit, but it really does lay the foundation for how you can build a marketing tech stack, and then with this tech stack, this is going to be the foundation for how you’re able to build your multi-touch attribution models. Now, my next book digs even more into multi-touch attribution and data taxonomy and all that stuff, but this first one is really, really awesome, super high-quality book, full-color images inside of it, and it’s great for anybody to be able to pick up and read in an afternoon because we get rid of all the fluff and we go right to the meat and potatoes for you.

Anna Shutko:

Oh, that’s amazing. I really love going into the meat, and I’m pretty sure you have a lot more amazing resources. So, if any of the listeners want to learn more about you or your company or the products that you’ve built, where can they find it all?

Dan McGaw:

Yeah. So, I would recommend two different websites. So, if you go to McGaw.io, you can definitely find out a bunch of information about what we do at our consulting business to help companies get set up with their tech stacks and all that stuff and reporting. On that site as well, there’s a lot of free tools on there. So, an AB testing calculator. There’s a keyword phrase parser. So, definitely you can find out about some of our other free products, and a lot of our educational resources are on there as well. So, anything that I do in my business, I also offer for free on our website. So, you can go check that stuff out there. And then if you visit UTM.io, you’re definitely going to love it. It’ll make it so that you can replace your UTM spreadsheets, which are a nightmare. Highly recommend checking that site out. A lot of information about how you can set up your campaign tracking correctly for multi-touch attribution is on there as well. So, if you’re looking to really step up your game and get into this stuff, go check out McGaw.io and also UTM.io to be able to further that conversation.

Anna Shutko:

Awesome, Dan. Thank you. And I have to say, it was a pleasure having you on the show today.

And that’s the end of today’s episode. Thanks for tuning in. Before you go, make sure to hit the subscribe button and leave us a review or rating on Apple Podcasts, Spotify, or wherever you’re listening. If you’d like to kickstart your marketing analytics, check out the 14-day free trial at Supermetrics.com.

See you in the next episode of the Marketing Analytics Show.