Multi-touch Attribution Models with Dan McGaw

Keith Perhac:

Hello, this is Keith Perhac again with the Data Beats Opinion. Thank you so much for joining us. Today, I had Dan McGaw joining me. He is the award-winning marketer and entrepreneur, the founder and CEO of McGaw.io, which is an analytics and marketing technology agency. He’s also the founder of UTM.io, which I find is one of probably the best tools for people who are looking to get their marketing under control. Because number one question I always get is, “I have 850 UTMs. How do I manage this? Why do they all say F book or Facebook or FB? Or why is there no standardization?” So a tool that I think is super valuable for everyone in this industry. You work with tools like Shopify. You’ve worked with Shopify, Unilever, Intuit. I know you were head of marketing at Kissmetrics for a long time as well. You’ve had an amazing run, so thanks for joining us.

Dan McGaw:

Absolutely, thanks for having me. So it’s a lot of fun to be here, and it’s great to chat with you again.

Keith Perhac:

Yeah, definitely. We met number of years, four or five years ago, I think.

Dan McGaw:

Yeah, I think it was five or six years ago is when we met. Yeah.

Keith Perhac:

Yeah, it’s been awhile. We were discussing this a little bit before we got started, but you had one of my favorite company names ever. And then as you rebranded to McGaw.io, I always miss seeing the Effin Amazing emails come in every month or so, or every week or so.

Dan McGaw:

Well, I appreciate that. We miss the name as well. Effin Amazing was a great name to have. We still own the domains. Don’t get me wrong. I don’t think we’ll ever let that go. I can’t even take the full credit for the name. My wife named the company at the time. I just liked the word amazing. She saw a show called Effin’ Science and said, “Why don’t you call it Effin Amazing?” And then we were off to the races. But four years later, we decided we needed to grow up, and we changed the name to McGaw.io.

Keith Perhac:

Yeah. We had a similar thing back when we were running the agency. We were called Delfi-Net for a long time, and it worked fine when we were abroad. But when we came to the US, everyone thought we were an IT service company, because it sounds very IT techy. So we had to remarket ourselves and rebrand ourselves to a marketing agency because that’s what people expected. Naming things is hard, man. It’s a… Yeah.

Dan McGaw:

Oh, it’s not easy at all. Yeah. It took me two years to change the name from Effin Amazing to something else. So it definitely took a long time.

Keith Perhac:

Yeah, excellent. Well, thanks again for joining us or joining me rather, and I’m looking forward to talk a little bit about multi-touch attribution, because this is something that I think especially as we get in towards Black Friday is going to be a rather important. I think a lot of people don’t understand it. If they do understand it, they don’t understand it well enough. They know it’s something that they need and something they need to watch out for, and that’s about the extent of what they’re looking at.

Dan McGaw:

Yeah. No, for sure. And it’s hard. It’s something that’s been the Holy grail of marketing for years, and everybody thinks it’s Holy water. And then when you get involved with it, sometimes you find out that it’s just water. It’s not very Holy anyways, so it can definitely be really hard. But there’s definitely a way to make it successful. There’s ways that people try to use it when it shouldn’t be. So it’s really, really complicated stuff at the end of the day, so it’s not necessarily easy to just figure out.

Keith Perhac:

So what do you… At a simple base level when someone’s saying, “Oh, I really need to focus on multi-touch attribution,” what’s the first hurdle that you try to cross with them, or what’s the first no, no, pump the brakes, this is what you need to think about?

Dan McGaw:

Yeah, and really, really good question. With multi-touch attribution, one, people don’t truly understand what it’s supposed to be done for. A lot of people want to use it for all kinds of different use cases. But the real fundamental thing that you have to get them past is, one, if you don’t have good data, multi-touch attribution isn’t going to matter, and that’s what’s most critical. So even going back to talking about UTM.io, the reason why we built UTM.io was because we wanted to just do marketing attribution, let alone multi-touch attribution. But if you have bad data coming in, you’re never going to be able to do multi-touch. It’s just not going to be possible because you need to be able to measure that. So bad data is of course the biggest roadblock that you see. But I think the bigger problem that we see is that people get lost.

Dan McGaw:

They can’t see the trees between the woods kind of situation once they get it. Multi-touch attribution should be used to optimize your ad spend. That is one of the primary things that it was built for, is that, within 90 days of looking at your multi-touch attribution, you can look back and say, this campaign is not successful. Shut it off, spend the money over there. So it’s really supposed to be used for ad spend optimization and ad spend efficiency. So I spend $10, I get two conversions. If I can increase my efficiency by 20%, I spend $10, but now I’m able to get almost nearly an additional conversion on top of that. So it’s not… Everybody’s trying to use it in all these different ways to understand the customer journey and to do this, but really, the simplest way you should be using it is to optimize your ad spend. That’s what it should be used for.

Keith Perhac:

That’s interesting, because I don’t hear that come up much when people talk about multi-touch attribution. So, which… And let me make sure that I’m clarifying this. What you’re saying is that, it’s not necessarily, Oh, people did touch these four different ads, it’s the fact that they touched four ads in the first place? And how can we lower that to not be four ads, but to be two ads or even one ad?

Dan McGaw:

Well, you can definitely optimize in that manner. But a lot of what multi-touch great for is when we do regular marketing attribution today. I have one conversion and it gets attributed equally to four different channels. So Facebook, I’m spending. Google, I’m spending. Bing, I’m spending. LinkedIn, I’m spending. I’m giving a whole attribution or a whole conversion attributed to each one of those channels. So your return on ad spend number is completely wrong in that case, right?

Keith Perhac:

Right.

Dan McGaw:

So now if you distribute that, of course you now have better return on ad spend figures. So now I have… Let’s say that I spend $10 to acquire a customer on each channel and they’re worth $10. Well, it means I lost $30 overall, but I’m tracking it incorrectly. Once you see that you have that mismatch of return on ad spend, you’re able to say, Oh crap, Facebook’s really not working, because I’m able to correctly attribute my spend also with revenue back to it. So you’re no longer misplacing revenues. So your ROAS numbers are able to then show you what is and what is not working effectively.

Dan McGaw:

And that’s why multi-touch is so important, is because you can actually say, well, there was $10 of revenue. I need to split that equally between these four channels. So Facebook, you get $2.50. LinkedIn, you get $2.50. But if you spent $10 to acquire that customer for $2.50 on LinkedIn, well, you know that you’re upside down. So you should probably shut off one of those channels, or start to look at inside of those campaigns, how can I optimize so I can shut off those bad things. That’s where multi-touch is most focused, is optimizing of course your conversions for what you’re actually getting in return for revenue.

Keith Perhac:

Right. And then, I think that there’s a whole different can of worms that we can look at, which is not just the multi-touch, but also multi-device, and the views, and the things that are much more difficult for us to track, because Facebook is a black box. Google ads is a black box. To a degree, what do you see when people are like, well, I’m getting view conversions from Facebook? Or they’re clicking on a mobile device, and then they’re coming in on a desktop to convert, how do you solve something like that? How do you look at multi-touch attribution over devices or view styles?

Dan McGaw:

Yeah. We call it breakage. There’s always going to be breakage that we can’t track. Luckily, you’re still going to have that cost data. If you’re using a product, whether that be attribution apps, C3 Metrics, whether you’re using something like a Rockerbox, that’s going to pull in your cost data into a platform and enable you to still attribute that cost to whatever conversion you have. So luckily, that costs won’t be lost completely, but you’re always going to have breakage where you can’t track a user. This is a lot of the problems that companies will partner with us to fix at our consulting company, is how do we be able to attribute more of those conversions? There are vendors such as C3 Metrics which are able to get you the view-through data.

Dan McGaw:

So they are able to tell you that this user viewed this thing on this channel. Products like LiveRamp will allow you to get a little bit more closer with device ID and stuff like that. So there are ways to piece it together. There are ways to get closer, but you’re never going to be perfect. That’s something we always try to remind people. It’s the internet. Trust me, it’s not perfect. I don’t know if you’ve seen your Facebook feed with what’s happened with the election. We’re not perfect, and it never is going to be. So the best thing you can do with multi-touch is understand that it’s more directional. It’s definitely signal. You have to understand that there’s going to be noise, but you need to focus on where you have the best signal.

Keith Perhac:

Yeah. It’s interesting. You talk to marketers or people who don’t understand the technology behind the internet and they’re like, “Oh, everything should just work and be perfect.” You talk with anyone who’s ever done any networking or anything that involves the hardware of the internet, and they’re like, “It is a miracle anything ever works.” The fact that we can go to a website and see it is just a miracle. There are so many things that can go wrong. That’s one of the things that I always had to talk to with my clients from when we were doing conversion rate optimization.

Keith Perhac:

They were like, “Why is the number in, let’s say, Google optimize the revenue number not matching what’s in the bank?” I’m like, “Because there’s a billion things that could go wrong. It’s not a one for one comparison.” You have to use it as, like you were saying, signaling. This is the trend that’s going on. This is what’s generally happening, and understand that at high levels, that’s all going to normalize out to a trend that you understand.

Dan McGaw:

Yeah, for sure. I think it’s important. We’ve been very fortunate to work on a lot of multi-touch attribution models. Companies like Looker, which are a very popular business intelligence tool and Segment, they’ve actually hired us to have us build multi-touch attribution models for their products. We’ve been very fortunate to be hired by some big companies to build multi-touch attribution models for them to be able to optimize their stuff. And then we’ve also had to look at all the multi-touch vendors out there. I’ve reviewed every single one of the vendors in a research study for a client, and it was fascinating. It’s really, really eye-opening. To be honest with you, even if you choose a multi-touch vendor, the model is still inherently flawed, because it’s based upon time windows. They have look backs, they have look forwards. They have all these things.

Dan McGaw:

So you’re a unique Snowflake, so is their model. There’s a lot of things that go into it. We just published a post today on Funnel.io site about multi-touch attribution, and how we recommend you do it using Funnel and as well as Looker. Even that, it’s like even though we built a custom, it’s still not perfect. It’s not meant to be. I think that’s one of the biggest reasons why I feel that I’ve been fairly successful in my career is that I haven’t looked for perfection in everything. I’ve wanted to make sure the things that we could perfect are important, but the things that perfection doesn’t matter, we have a trend. Let’s follow that trend. Let’s run there. If we get there and we’re wrong, that’s okay, nobody’s going to die.

Keith Perhac:

Right. I think you’re exactly right there, which is, because you’re never going to be perfect, you need to understand the limitations of whatever you’re using, whatever model, whatever data source you’re using, however you’ve set that up. Understand where that variance can be and make decisions based on that, because you’re never going to get 100% accuracy at it because it’s just not possible. So then-

Dan McGaw:

Yeah, that’s the internet.

Keith Perhac:

Yeah, exactly. So what do you see as… I don’t know. Are there different situations where people want to use different attribution models? Because there’s of course the Linear. There’s equal spread. There’s weighted. There’s time-lapse. There’s 101 different models even within the larger framework. Do you find that some work better for different situations, or how do you… What’s your first step when someone says, Oh, we want to have better multi-touch attribution?

Dan McGaw:

Yeah. The short answer is I always would advise companies to start with a Linear model to start and that just means that we’re equally distributing whatever revenue across whatever channels that it came from. Linear is the easiest way to get started. We’re all used to first-touch. We’re all used to last-touch. That’s regular marketing attribution. Linear is going to make that a little bit easier, because you’re now going to start to see that. I typically advise companies to use that for the first 90 days to six months. If they really start to try to figure out whether they can be more advanced with it, they can of course attack it.

Dan McGaw:

Now, even with just a Linear model, you have to still look back at, what is my lookback window, and then how am I going to adjust that? I think that’s where the complexities really start to come in. Because if you’re a small company that has a really transactional products where you’re doing e-commerce, maybe your conversion window is 15 days. But if you’re somebody like myself, my conversion window is six months. So you have to really start opening up those lookback windows, and then you have to create what’s also a concept is we’re rolling window on how those conversions are going to work. So the Linear model just makes the basis of that much easier to start with.

Dan McGaw:

And then the way that we try to help companies take the next step of that is if you were to try to give any type of priority to certain channels. So are there certain channels which you know are going to have a greater impact more of the time than other ones? Or there channels that you know that we’re not able to track half of the time so we may give additional credit to that one. An easy culprit of that would be something like direct mail. Direct mail is not able to be easily tracked. But if we’re able to get an individual conversion from a piece of direct mail, a touchpoint from direct mail, we do know how much money we spend on some of these. We may give more credit to those types of things if we do get a true touch point from that.

Dan McGaw:

So there’s a lot of ways that you can customize it, but I never try to advise anybody other than starting out with just, Hey, do a Linear Based model. Time Decay you can look at, but just do Linear Based. Look at the 30-day look back window, or a 60-day look back window and get used to that. The reason I say that is because, if you get into the part where you start to make the sausage, and you’re also the person selling the sausage, and then you’re also the person judging it and doing all the things, the marketer can make up whatever they want and make any story that they want out of the data. That means the data is now biased. We’ve seen that happen before where… We worked with a very, very large automobile and RV sale company.

Dan McGaw:

We were building out a multi-touch attribution model with them, and it was going pretty good, but the VP of marketing that we were working with was like, “This does not match my data.” And I was like, “Well, unfortunately, I’m trying to tell you is that the data you’re looking at, that you have is wrong. This is what it looks if you use multi-touch.” “Well, I know my data’s correct.” I’m like, “I’m not trying to tell you that you’re not correct. I’m just trying to tell you that your data is wrong in regards to what you’re looking at. You’re not attributing the conversions.” We wound up not moving forward with the project.

Dan McGaw:

It was one of those things that came to the end, the project failed. That VP of marketing is no longer at that company anymore. I’m no longer working with a company, so it was a sad deal. But at the end of the day, it was because they didn’t want to believe the math and the real numbers. They wanted to create this custom thing and that custom thing. It was like, no, no, no, just stay Linear right now. You need to shut off those campaigns, because they’re losing you money. So stick Linear, try to keep it easy until you can really get advanced.

Keith Perhac:

Talking about that VP and just those types of marketers [inaudible 00:16:11], do you see a lot of other companies or people looking at the data and then, not massaging the data, but coming up with theories and pet theories from misguided data that doesn’t match reality, almost looking-

Dan McGaw:

All the time.

Keith Perhac:

… too far back?

Dan McGaw:

Yeah, all the time. I don’t just see that in multi-touch, I probably see that a little less than multi-touch. I see that everywhere where people are coming up with stuff. I’ve been lucky to have been working in the data space for a considerable time. I’ve been working in the analytics space for at least eight years now. Before that, I’ve been in the martech space for 20 years. So I’ve seen this… The easiest way I summarize my career is I’ve just seen some shit. That would summarize the way that I see people manipulate data and make bias out of it compared to trying to disprove themselves. I think that’s one thing that I am always challenging myself.

Dan McGaw:

Naturally, as a human, we have confirmation bias. So when we see data that confirms a bias we already have, we get all excited. One of the things I had to learn a long time ago, and I really learned this a lot from Mixpanel many years ago was, how do you prove yourself wrong? Well, you need to change the situation to everything is about proving yourself wrong and being paranoid that you’re incorrect, and then using the data to prove yourself wrong. Because I’ve seen more than my fair share of people create whatever story they want out of the data, because they can, not because they should.

Keith Perhac:

That I think is I think very valuable, always to trust, but verify almost in a way. It’s like, yes, the data is correct, but let’s try to prove ourselves wrong to see, are we getting the right information out of this? Maybe the data is 100% correct, but we have to understand, is our assumption about what the data is telling us correct or are we misguided? Are we bringing in those prejudices or those those kinds of ideas or those goals that we want it to say?

Dan McGaw:

Yeah. I think instinct and gut still plays a huge weight in it. So I do believe in the idea of data is helpful, but you need to be data informed, not necessarily always data-driven. So I think it’s a tough coin to go on both sides, because in either case, you could be wrong. Whether you follow the data or not, in either case, you could be wrong.

Keith Perhac:

Yeah, exactly. So when you’re talking about this data analysis and stuff, especially with click data and things that, like we were talking about before, Facebook is a black box, Google is a black box as far as they don’t let you see exactly who each click was, blah, blah, blah.

Dan McGaw:

Yeah.

Keith Perhac:

When you’re talking about attribution that are 30, 60, 90 days later, we’re talking about building essentially a data lake or a data warehouse there. What are some of the issues that you see in creating that? As people are getting started, they’re like, okay, I want to be able to monitor this 90 days out. How do you see… What are some of the ways that people go about that, and what are some of the ways that people start to screw that up I guess in a way?

Dan McGaw:

Yeah. No. So, going back to a comment that I had talked about earlier is one of the things that, you can’t be perfect in everything, but there are certain things that you can definitely try to be perfect in. For me, one of the things that we try to get really, really good at with any kind of project that we’re working on, it’s really about getting the taxonomy and the data schema correct in the beginning and really laying that out. Now, is that data going to come to us perfectly? No, it’s not going to come to us perfectly.

Dan McGaw:

But the more proactive we can do with it and the more that we can plan around it and make sure that we have the right data taxonomy to be able to have nomenclature around the actions, nomenclature around the properties and the context, around those things, and then nomenclature around the attributes of the individual person, the more successful we can be later. I think that’s where we see a lot of companies really just throw shit at the wall and hope that their data taxonomy is going to work. They don’t take it seriously enough. And then, they just let it go into the system. In the modern stack when you’re using a customer data platform or artificial intelligence or anything like that, what people don’t understand is when you have bad inputs, it’s a bad output. But if you have a bad input and then you throw it into automation, you have 10 times worse outputs. So you really have-

Keith Perhac:

[crosstalk 00:20:37] poisoned it?

Dan McGaw:

Yeah.

Keith Perhac:

It not just bad output, but you’ve poisoned the whole engine?

Dan McGaw:

Everything about it, and it exasperates the issue even more. It’s like in comparison, the flu sucks. But COVID is way worse, because it can spread more. Automation is like that. It’s spreads the problem around more. So if you get that taxonomy part correct in the beginning and you really build out a data dictionary, take the time to understand what you’re going to call things. If you were to go to my site, McGaw.io and scroll to the bottom, there’s a resources and download section. There’s five webinars that I have that talk about, how do you set up taxonomy, and how do you do this. Everything I know, I teach for free on our site. So if you get that right, that’s going to make it so the data… You don’t even have to throw it into a data lake.

Dan McGaw:

But if you throw it into an Amplitude or Mixpanel or something like that, your data is going to be much, much cleaner. You’ll be able to get some more insight out of it. But if you’re leveraging a data lake, it’s going to get it in there cleaner. If you have a good data engineer, they’re going to love you, because they don’t have to do any kind of ETL transformations. We’ll just be able to tie it all together. So it is really critical to get that taxonomy correct. Going back to something you had talked about with UTMs, it was the whole reason why we created UTM.io was, we saw this huge data governance issue that we were being faced with as analysts that we couldn’t report on data because the social media team’s using one thing, content team’s using another thing.

Dan McGaw:

The paid advertising team’s using something else. You got the PR agency using some crazy codes. We don’t even know what’s going on. So when we’d have to run a multi-touch attribution model, you’d have 64 campaigns that you’d never even knew. You’re like, why are they using keyword in social media? They don’t have… But it was, people didn’t have any data governance. So that clean data coming in is super, super critical to success.

Keith Perhac:

Yeah. Do you find that that has to be something that everyone is on board with and everyone has to have input on, or is that much better as a top-down like, Hey, we’re following this structure?

Dan McGaw:

Yeah. I hate to say top-down, because I’m so… Even in our company, I’m like, we only have four hierarchies. There’s only four layers. Let’s keep it as flat as we can. But really, when it comes down to a lot of the data infrastructure, the people who are super dependent upon it are the analysts at the end of the day, so they really should be able to force a good amount of that down. I do think there is a partnership between the analyst, marketing, sales, product and all that stuff, but you really do need to have, as we call them, data lords inside of a company who will come in and be the person that helps out and gets that accomplished, because data does need to be pushed down a little bit. We see this every day with UTM.io. It’s the VP of marketing and the data engineering team that’s creating the taxonomy, and our product just allows them to say, okay, these groups of users can only use these things. When they do X, make it do Y. Those users don’t care anyways. They just want to do their job.

Keith Perhac:

Right. They don’t care about the naming. It’s more of a mistake or like, I don’t remember what I used last time. Was it FB or F book or Facebook or what was it?

Dan McGaw:

Yeah, they don’t care. Excuse me, they just want to do their job. So that’s where I think that the data taxonomy is going to be the most critical part. Because if you don’t have clean data, it doesn’t matter.

Keith Perhac:

Right. It is interesting, because as I mentioned in the intro, this was or is still a problem that a number of our customers come to us with, because they have different people. They’re usually smaller companies, so they have a number of consultants come in. They all have different ways of doing it. So they have 30 different UTMs or UTM terms that should all mean the same thing. And they’re like, “Well, how do we clean this up?” So we build a bucket and it’s like, okay, any of these words count as Facebook. But it just keeps growing and growing and growing, because they have no taxonomy. They have no governance around it. They have no naming structure. And so that’s been a huge challenge with us both with UTMs and tags and even product names, where sometimes people will reverse. It’s community virtual something, and then it’s virtual community something. Naming is hard. It’s [crosstalk 00:24:48].

Dan McGaw:

Oh, it’s hard. The amount of time… So we had talked very briefly in regards to sending data to a data lake and stuff like that. We talked about the automation being a huge part of making that even worse. We leverage customer data platforms all the time, so products like Segment, MetaRouter, mParticle. They basically become a data pipeline for all of your data. The hard part about leveraging a tool like that, even when you’re thinking about multi-touch attribution is that you’ve now made it so that you’re not talking to a tool one-on-one. You’re now making it so that you talk to one specific tool that now translates that into 80 different languages to work with 80 different tools, and each one of those tools have a different purpose.

Dan McGaw:

So it’s very similar to speaking English, and then going to Japan and expecting everybody in Japan to understand what you’re saying, and then talking through a translator who is native to Japan. That translator is not going to be… Not all the words match. Not everything is going to completely line up. If you don’t have that good translation layer, or you don’t know how to tell the translator what you’re trying to say, that translator is going to make something wrong, because it’s only going to hit the end destination incorrectly, or they may not have a word for that. They may not have a way to do that. So whenever we have to work with a customer data platform, which is nearly all the time, we have to send data into that customer data platform, let’s say Segment.

Dan McGaw:

That same bit of data has to be used by Salesforce, Marketo, Google Analytics, Amplitude, Facebook, a myriad of different tools. So the way that we structure that taxonomy, the event, the property name, the property value, the identified call, the traits that come along with that, it really matters. It has a lot of matter to it. A lot of companies are just like, it doesn’t matter. And then they get there and they’re like, I can’t figure out why I’m not creating magic. And it’s like, well, it’s because you don’t know how to make magic. So you can’t create magic if you don’t know how to make it. So getting that data taxonomy is going to be very, very critical for multi-touch attribution or any reporting you have in general.

Keith Perhac:

Yeah. And being able, exactly, exactly, to push that data and to cover all your bases like you’re saying right off the bat is really difficult, because each of those systems expects it in different way. They report on it different ways. You’re trying to get different information out which has different properties on it, but you want to be able to cover all your basis. You have to decide what you’re going to keep, what you’re going to focus on to solve which problems.

Dan McGaw:

Yeah. No, absolutely. It’s actually something. In my book, Build Cool Sh*t, if you were to check out my book, If you go to my website, just McGaw.io, it’s on all the pages of the site. We talk a lot about, what are the true outcomes that you want to create? What are the things you want to ultimately do with all of your data? We talk about analytics. We talk about how you’re going to do automation, how you’re going to do lead scoring and stuff like that. But the background of it is how you connect all these tools together to create that.

Dan McGaw:

Earlier, you were talking about, how do you get the ad networks to talk basically with black boxes? They’re definitely black boxes, but there’s ways to get data out of them. There’s ways to get that data merged in with what you have in your CRM. That really comes down to how you build your architecture. And in my book, Build Cool Sh*t, we talk a lot about how you ultimately need to think about building that data pipeline to keep your business successful and then be able to measure on those business metrics.

Keith Perhac:

Yeah, definitely. So when someone comes to work with you or with your company, and they’re like “Hey, we have this attribution problem. We don’t know where anything’s coming from, or we’re completely wrong,” where do you start with them? What’s that first step of looking at what they’re doing and saying, okay, this is the road we need to go on to improve what you’re doing?

Dan McGaw:

Yeah. Naturally, the first place that we want to investigate is, where are the conversions happening? What are the touch points happening? And then what are the channels that we’re getting that data from? And then, are we going to be able to correctly get the data from all of these different tools to be able to build the model? We don’t even worry about what is going to be the attribution model, the look back windows or any of that stuff until we do an assessment of, can we effectively track conversions in a digital and automated way? Can we actually see all the different touch points? And then where are we spending money? Because you can give a dollar amount to anything, and you can figure out how to give attachment to different user IDs pretty easily.

Dan McGaw:

Once we have an assessment of all that, we need to figure out, how do we relate that all back to an individual identity? Are we locating that? Are we attaching that to a person like a user ID, or are we attaching that to a company which would be like an account ID? And then we have to figure out, how do we make sure that all of those touch points can map back to whatever that final record is. Because what you have to remember is that every one of these touch points is its own independent piece of data, and they’re just associated by a user ID or an account ID. So we have to be able to see how we correlate all of those things back together. That’s the first step is trying to understand, how do we correlate spend to revenue, to touch points, all back to that user ID?

Keith Perhac:

And building out not that user ID, but also, what’s the goal? Because if you look at something like a B2C where you’re looking to sell, okay, I’m looking to sell to a person, that’s very different than I’m looking to sell to an account where that person may or may not be connected to that account directly. So you’re seeing someone coming from an ad or through some sort of organic source, but you’re not able to then connect it to the company, because they came in through a Gmail address. And then the company came in through the corporate address. And how do you pull all this stuff together? It becomes this big muddle depending on how you want to track. I think that’s exactly what you’re saying, which is, it’s so important to understand what the end goal is, and if it can be tracked through that whole process.

Dan McGaw:

Yeah. I think the interesting thing, of course you’re attach it to the goal. For me, I think there’s only one goal. When I think about multi-touch attribution, it’s to optimize the advertising spend in most cases. Some people want to do a content attribution model, which to me is completely different. Multi-touch content models are a totally different bag, because it’s more focused on the content or the customer journey and what content is applicable at what stages. I think that’s definitely a big part of that. But the goal is almost always the same, is how do we efficiently convert this user, and how do we maximize that efficiency? So it’s relatively a similar goal. I think there’s just different inputs and different outputs depending upon the vertical and everything like that.

Dan McGaw:

That’s where it gets… For us where it’s like, the problem is always very similar, it’s because it doesn’t matter. You could meet there’s so many different verticals right out there, and they all have their own unique snowflake. But the thing that’s always going to be consistent is I have a record that I need to have the customer, and then I have spend, I have touchpoints, and I have conversions. I need to be able to change those things or see all those things. So the goal at the end of the day doesn’t matter. I have to first figure out whether I can do those four things, because if I can’t, if I don’t have a user record that I associate with, if I don’t have revenue that I can attribute it to, if I don’t have the touch points, and if I don’t have the spend data that I can access, then I can’t really build a model inside of multi-touch.

Keith Perhac:

Right.

Dan McGaw:

I guess those are prerequisites to even building [crosstalk 00:32:04].

Keith Perhac:

Don’t even get me started. Yeah. We have had people come to us and they’re like, “Hey, I need to know how my ads are performing.” It’s like, “Okay, what are they converting as?” “I don’t know.” It’s like, “Where’s the sale coming in?” It’s like, “Oh, we call them and they write us a check” And it’s like-

Dan McGaw:

Oh my God.

Keith Perhac:

… wow, this is going to be very difficult, man.

Dan McGaw:

The crazy thing is, we track multi-touch attribution just like that. So they write you a check. We’re able to get a QuickBooks feed. We’re able to get the client ID from QuickBooks. We’re able to map that client ID to a table. So we’ve been able to… One of our clients unfortunately which got decimated due to COVID, we were able to map their direct mail. They’re sending a million pieces of direct mail a week to small businesses.

Dan McGaw:

We were able to map that direct mail into their multi-touch attribution model combined with Google, combined with Facebook, combined with LinkedIn, get all that stuff put into a warehouse and then still map it. It was through fuzzy logic mapping, a dozen Bradstreet numbers that were fuzzy matched back to the company name, which were fuzzy matched back to the Salesforce ID, and all of this crazy glue. But it worked. Those customers got… They converted in Salesforce, and there was a check that got delivered through another service. We only cared if the check got delivered, right?

Keith Perhac:

Right.

Dan McGaw:

So there’s a way to do it.

Keith Perhac:

That’s amazing.

Dan McGaw:

There’s a way to do it. It all just depends on how much money you want to pay me to do it and how big the problem is.

Keith Perhac:

So lumpy mail-

Dan McGaw:

[crosstalk 00:33:42] a lot of fun.

Keith Perhac:

Lumpy mail has always been something that I’ve been incredibly interested in as far as tracking. Because I started advertising and doing all this work through the digital side. I’d worked at a TV radio advertising agency way back in the day before Google ads even existed. Now that we have all this tracking available, it’s almost difficult for me to even imagine what it was like back then. So doing this lumpy mail and tracking that into how it interacts with the multi-touch, with ad spend, and through Salesforce and everything, it’s super fascinating to me. So you said you were doing it just on the company names and a record of who you had sent it to, so there was no extra tracking in the mail or anything like that?

Dan McGaw:

Yeah. So there was promo codes that were added to the flyers that were going out and stuff like that, but they weren’t unique individual promo code, just because at that point, you’re dealing with too many promo codes.

Keith Perhac:

Mailings, yeah.

Dan McGaw:

They’ll start… Users don’t even use it. So you want to use it to be able to hit a vanity URL, go directly back to the website and then be identified. That vanity URL had a UTM. So when we hit the website, we knew they came from direct mail, but we didn’t exactly care that they were from this exact thing. What would happened is that whenever you… This was going to businesses, and you can do the same thing with users is whenever we were sending to a business, it was based upon a DUNS and Bradstreet number. Everybody has a DUNS number, every business out there at least. That DUNS number of course was also where we got that business’s address. So we were able to send that mail to them. There was a DUNS number.

Dan McGaw:

There was a company name to it. If that company came back to the site, what we’re able to do is of course track that person as they came into the system. We then tried to map back to that DUNS number of course later. So that way, we could then track from the mail house, okay, this DUNS number has received six pieces of mail over these dates. That information would be sent into the CDP. That CDP would send it into our warehouse, would also send it in tools like Amplitude. We were then able to then take all that data, match it into Salesforce, Marketo and all those things, still be able to keep it all relational. So if that opportunity eventually closed and the check was sent, we would get another API call, send that into the system. And then we would have all of that tracking data, not to mention also their ad spend, Facebook, all that stuff. Once your user gets-

Keith Perhac:

Because you can track in the website then.

Dan McGaw:

… to the website, there’s a user ID now at that point. That user ID gets associated. The website was still a touch point no matter what, so that made it a lot easier. But there were still major complexities, because we had to pull the ad spend data out of AdWords, Bing and all that stuff. So we used Funnel.io. Funnel helped us pull that data down, put it into a warehouse, and we can now run our own numbers on top of that.

Dan McGaw:

It was able to be done. So [inaudible 00:36:30] it took six months and probably two months of just… Two and a half months of strict planning. That was one of the companies that paid me to do an attribution study. We looked at all of the vendors to figure out who should they choose. We came back to them and we said, “You should choose C3, because they have professional services. Or you should build it yourself because you’re crazy.” And then they said, “We’ll pay you. We’re crazy. Let’s build it ourselves.” And then we did. Their data lord was awesome to work with. So that’s-

Keith Perhac:

That’s amazing.

Dan McGaw:

… where the term data lord came from, was she was so much fun.

Keith Perhac:

That’s awesome, that’s awesome. Changing track a little bit, I want to talk a little bit about your book. So you had [inaudible 00:37:15] a second ago, you mentioned… How recent is this? This is within the last couple months or?

Dan McGaw:

It’s been a year, so my book’s been out for a year.

Keith Perhac:

Wow, it’s been a year already.

Dan McGaw:

Yeah. So the book came out. So we put it out right at the end of 2019. So we just did a reprint of it with some small changes to it. But yeah, it was a ton of fun to write. It’s got beautiful colored pictures in it. So we made it specifically for the CEO and CMO in mind. So a practitioner marketer could read it and really be able to pick up the tools of the trade. But a senior level executive also isn’t going to be having to raise 600 pages of fluff to get there. I don’t know if you know this, but most C-suite and VPs, they like color books. They like coloring books as well. They don’t like to read long novels a lot of times as we’ve come to find out

Keith Perhac:

Interesting. It’s been one of the challenges with me and business books is that you get to that, usually about a third, maybe a half way through it, and then you notice that it starts repeating itself and repeating itself and repeating itself. Actually, I talked to an author friend of mine and they said, “Oh yeah, what happens is you have enough content for 60, 70 pages. And the publisher was like, no, no, it has to be at least 160 to publish. So now you got to retread everything you’ve said over and over and over again.” I always just found that such a waste. I would much rather read a 60 or 100 page book that’s just right in there, right exactly the information I need to know and get value out of it. I don’t consider myself a C level. But as far as what I’m doing, it’s similar. I’m running a business. I want to get the information and get out. I want to start implementing.

Dan McGaw:

No, I totally agree. I think one of my favorite books out there is a book called Thinking, Fast and Slow, a great book. But I listen to audible books a lot and I run a lot and listen to my books. So I run for about an hour a day, five days a week, so I get a pretty good amount of listening time in. The book is 17 hours long. But by the time you get to the 10th hour, you’re like, I feel like I’ve heard this over and over and over again. It’s an amazing book, so I’m happy I finished it. But there’s definitely some books that, and also Principles by Ray Dalio. It’s like, you could have done this in half the space. But hey, that’s them. I tried to make my book even shorter. My next book which I start writing in the next couple of months here is going to probably be double the size, just because we’re going to go a step deeper to try to make sure the companies understand some of the practices of Google Tag Manager a little bit more.

Keith Perhac:

So what’s the goal of the next book?

Dan McGaw:

Yeah. So my last book, Build Cool Sh*t is the blueprint for, how do you build a marketing stack? So it was pretty high level. How do you integrate the tools from a high level? What are some of the outcomes you can create with a stack? So like lead scoring, personalization, things like that, and really good analytics. One of the things that we noticed when we launched the book was of course, we got a lot of people really thinking about the modern analytics and modern data stack, which is super, super helpful. But the part that’s missing from the book is really talking a lot about the integration of it, and how easy it is to integrate these things. I think people get a little intimidated by technology.

Dan McGaw:

But when I created my online course… So I have an online course. If you went to CXL Institute, they have my online course on how to build a stack. One of the things that we discovered was that people don’t know how to do any of the integration. You don’t actually have to know how to do the integration, you just need to know how the integrations work. Because there’s really only three types of integration. You either have a platform side integration, which is like Marketo connects to Salesforce, and there’s a mapping table between the two. The platform side is really, really easy. Another platform example would be Zapier.

Dan McGaw:

That’s a platform side, also known as citizen coding and things like that. You then have a server side implementation. That’s when developers get together and have servers talk to servers. And then you have a client side implementation, which is most times just done with Tag Manager. We’ve learned that people don’t understand taxonomy. They don’t understand that data component. I really want to humanize that and make it so that any marketer can figure that out. I think in my last book, we didn’t talk enough about that. And then we’ll connect that data taxonomy to personalization and then how your personalization works through data taxonomy. So the next book will be focused more on that stuff. It will be a little nerdier, but hopefully we’ll make it so that regular folks as well can understand some of that nerdy stuff.

Keith Perhac:

That’s going to be gold though, because that’s honestly one of the challenges. One of the many challenges we run into is that people don’t know how to refer to what they’re saying. And so everyone’s using the same words to mean different things and doesn’t understand how the integrations connect to everything. It’s just that low level understanding of how it all fits together. The words they can use to describe it makes things so much easier when we’re having those discussions. I’ll get on a call for a demo or something and they’ll be like, “Yeah, I’m a martech guy.” I’m like, awesome. Now I can use the real words to describe exactly what’s going on instead of having to fluff everything and to make sure that it’s understandable at a higher level for someone who doesn’t know what’s going on under the covers or under the surface.

Dan McGaw:

Yeah. No, and it’s hard. I’m very lucky that I’ve been doing this for a very long time. I worked at companies Kissmetrics. I was Head of Growth at codeschool.com, so I was surrounded by developers all day long. So I was lucky to be able to pick it up. I wish somebody would have made it easier for me, because I had to learn trial by fire. But I’m hoping that I’ll be able to explain, just like I explained in my last book. I hope I’ll be able to break it down so people can really easily understand.

Keith Perhac:

Awesome. Well, we’ll link to the book and your course in the show notes as well as Dan, McGaw.io and UTM.io.

Dan McGaw:

Yeah.

Keith Perhac:

Dan, thanks so much for joining us. This has been super fun.

Dan McGaw:

Awesome. It’s been great to be here, Keith. I appreciate it.

Keith Perhac:

Definitely. Besides your website and your book, where should people find you online?

Dan McGaw:

Yeah, the easiest one is of course McGaw.io, so M-C-G-A-W.I-O. I’m on LinkedIn for the most part. I don’t really play a lot on Twitter. So if you check out on LinkedIn, just Dan McGaw, you’ll be able to find me there as well.

Keith Perhac:

You’re a lucky man for staying out of Twitter. I need to learn that lesson. All right, Dan, thank you so much for joining us.

Dan McGaw:

Yeah, thank you.