Keeping Garbage Out of Your Data So You Can Make Better Marketing Decisions With Dan McGaw From McGaw.io [AMP 223]
CoSchedule's Headline Analyzer Studio will score your overall headline quality and rate its ability to result in social shares, increased traffic, and SEO value.Test every headline before you publish. Try Headline Studio »
Data problems are probably lurking somewhere inside of your marketing stack. Don’t freak out, just yet. Most analytics packages and marketing software services that deal with data have some gaps or inaccuracies.
Today’s guest is Dan McGaw, CEO and founder of McGaw.io, a marketing technology and marketing analytics consulting company. Dan talks about how to make better marketing decisions—identify and fix deeper issues to avoid data disasters. He explains everything you need to know to keep your data clean and metrics moving.Please add mp3 file in field 'Link to mp3 file' on edit page!
Some of the highlights of the show include:
- Why is data cleanliness important? Analytics + Bad Data = Bad Decisions
- Directional: Data is not meant to be perfect, the goal is to grow and take action
- Data Spectrums: Everybody has unreliable data—how bad is it?
- Marketing Stacks: Different problems stem from data issues
- Taxonomy: Common problem is not having consistent or connecting names
- Be Intentional: Set up and configure marketing tech, or set yourself up for failure
- Audit: You know there’s a problem, but you don’t know what it is, where to begin
- Solution: Plan and be more proactive by understanding how data flows in
- Best Practices: Urchin tracking parameters (UTM) are culprits of bad data
- How to Build Cool Sh*t: Take it slow, take your time, don’t try to rush projects
Ben: Hey Dan, how’s it going this afternoon?
Dan: I’m doing good. Thanks so much for having me.
Ben: Absolutely. You might be doing a little bit better than I am right now with your 73 degrees down there in Orlando.
Dan: That’s cold for me. I don’t know what you’re talking about. I know it’s colder where you are right now, but I’m going to wear a jacket when I go outside later.
Ben: I too am going to be wearing a jacket. It might be a little bit heavier, but it is a jacket nonetheless. Before we get too sidetracked, would you mind taking a moment to introduce yourself to our listeners and explain what you do at McGaw.io?
Dan: I’m Dan McGaw, the CEO, founder of a company called McGaw.io. I’ve been in the marketing technology space for over 20 years. I got my start in 1998, sending mass emails before there was mass email. I just like to round that up and say I’ve seen some […]. Some people recognize me from when I was the head of marketing at Kissmetrics. I helped lead that rocket ship to do some really cool stuff. Before that, I was the head of growth at another company called codeschool.com which I helped get acquired at Pluralsight. I’ve just been around when it comes down to startups and I’ve seen all kinds of different things.
At McGaw.io, they were a marketing technology and marketing analytics consulting company. What does that mean? Naturally, we help companies get visibility into their customer journey or we help them optimize their customer journey for conversions. But we do that by leveraging marketing technology. We’re usually integrating tools, operating tools, accelerating growth, using them, and our backbone has always been analytics. I accidentally started this company when I left Kissmetrics. As you might understand, when I left Kissmetrics, I had a big portfolio of friends who needed analytics help. Naturally, our backbone has just always been analytics here.
Ben: Very cool. Your expertise ties in really well into the topic of this episode, which is data cleanliness, data trustworthiness, just broadly talking about how to keep your data clean and trustworthy so that you can actually make good decisions based on your data. At a high level, in your view, why is data cleanliness important right off the bat and why is that something that marketers should concern themselves with?
Dan: A really good question. You have to understand that if you have analytics and your analytics have bad data, that means your analytics are wrong, which means that you’re naturally going to be making bad decisions. I’m working with a company right now where they have really bad marketing attribution and they know it. They basically have stopped all their marketing spend. We’re not talking about some small 4-figure spend, we’re talking 5-6-figure spends. They’ve pulled it all down because they don’t want to lose any more money because they have bad data. It can have a massive impact on your business.
Even in my own profession, in my own career, I’ve been at companies where we had bad data and then we made a good decision. We’ve come to find out that a good decision was made on bad data. It was a great decision for the data that we had, but after we found out that it was bad data, the outcome was crap. You really do have to make sure that you have good data. I don’t think it has to be perfect. Data is not meant to be perfect. The internet is not perfect. If you thought about any place that is messed up, it’s the internet. That means that your data is not going to be perfect, but if you can at least get directional, you’re going to be in a much better spot. If you continue to clean the data, it’s going to help make sure you don’t make more mistakes.
Ben: One word that you dropped there that I’d like to expand on a little bit is directional. I know what you’re getting at there. I know that internally, even for us at CoSchedule, a lot of times we will look at data or at metrics in general and we’ll just treat that as a weather vane that gives us an idea of which way the wind is blowing rather than getting too hung up on getting a real super granular read on how things are doing. Is that what you mean when you use that term, or did you mean something else by that?
Dan: I definitely mean that. You need to know which way the wind is blowing is for sure. But there are definitely areas where the data can’t be perfect. A common one is going to be multi-touch attribution or marketing attribution. In multi-touch attribution, there’s no way for the data to be perfect. The goal is to be directional and tell you how to do that. There are definitely certain areas where you have to be directional, but at the same time, personally, from my perspective, the companies that are typically growing the fastest are the ones who are less focused on definitive and more focused on how do we get directional data that’s going to tell us which way is growth, and let’s start moving, and let’s take action.
Ben: Sure. I love that emphasis on action there. Hence, the Actionable Marketing Podcast. We are all about action here.
Speaking of the inability for these things to be exact, but in your estimation, how common do you think these kinds of problems are? How widespread do you think issues are related to marketers just having unreliable data.
Dan: I would definitely say everybody has an unreliable data situation somewhere, every company. We’re an analytics consulting firm and I can promise you this, we have unreliable data in one of our tools about something. I would definitely say that we have things that are not tracked, which probably should. We have things that are tracked wrong that should or shouldn’t be. Everybody has unreliable data somewhere.
One thing to be very conscious of is that when you think about data, you not only have data that you used for analytics, but then you also have data about the people, of them doing stuff, and those are two different spectrums of data. When you think about it like an analytics setup, we’re sending events and properties with that stuff that goes into an analytics tool and that can be stored incorrectly.
At the same time, when you have things that go into your marketing automation tool, it’s a totally different type of data. Those are identifying calls or trait calls, whatever they want to be, and those are saved in different ways. One is used for reporting. The other one is used for personalization or automated messaging. In either case, companies have problems on both sides of that spectrum and everybody has it. It’s just how bad of a problem it is. How architected is it? How complex is the problem? That’s one of the reasons why four years ago when I was looking at our agency, what are we going to do when we grow up? What is the thing that not everybody else does?
Obviously, the first thing that came to my mind not to do was PPC, or SEO, or any of the things that there were a million other people for. We looked at what was the area in marketing that is hard and that was the marketing technology stack. That was the problem that we set out to solve because the data reliability issue is both on the analytics side but also both on the marketing automation side. But if you’re an analytics specialty shop, you can’t fix the stack. If you’re an automation shop, you can’t fix the stack. We chose to be right there in the middle. It worked out well. It’s been a lot of fun because we see a lot of data issues. It’s been a lot of fun.
Ben: I will say that it’s definitely an area, not that I’ve looked particularly hard but it does seem like an area that’s underserved. In general, just agencies like your own that are focused on those types of issues which is interesting given that this is a problem literally all of us as marketers have somewhere. I think that’s super interesting.
From your perspective, as you had said, you’ve been in space for a long time. You’ve looked at a lot of data. You’ve looked at a lot of marketing technology stacks. You’ve worked with lots of different clients and different companies. I’m sure you’ve seen all manner of different problems that all stem from data issues in one form or another. What are some of the most common problems that you’ve seen that marketers face when they fail to properly manage their data?
Dan: The biggest problem that we tend to see is that people just start running without ever putting their shoes on kind of situation. In data, whether that be analytics, or the stack, or marketing automation, the first thing you have to focus on is your taxonomy, and what is your nomenclature. What are you going to call things? The most common thing that we see is the taxonomy is messed up. At companies you’ll see, to the marketing team, it’s called a sign-up, to the development team, it’s called user-created, and to the customer success team, it’s called a registration.
Getting your taxonomy in order is by far the number one issue that we see. Simple things like capitalization, of course, is the stuff that breaks all kinds of different tools and getting that—as we would call it the stack taxonomy—synonymous throughout all of the products. A common thing that people forget about is that all of these tools should be connected. Your business is the platform. All of these tools are ultimately connecting to it and all of these tools need to be connected to each other as well. If in one tool it’s called first name, another tool is called Fname, and another tool is F_name, that taxonomy gets harder and harder to map and makes it more difficult. If you first make sure that you have good taxonomy, that’s going to make sure you set up the rest of this for success.
Ninety-five percent of the time when we see these problems, that’s because nobody ever focused on taxonomy first, or when they did go build their taxonomy, they just got super lazy, and they just did the minimal viable thing and spun it up. But if you take the quality time to do taxonomy right, you see really good outcomes, and trying to make sure that taxonomy works across the stack is where you get the best outcomes as well.
Ben: A recurring theme underpinning much of this conversation is the need to be intentional when it comes to setting up and configuring marketing technology. If you rush into things without a plan, or if you treat complex systems like simple out-of-the-box solutions, then you’re setting yourself up for failure right off the bat. Even in a best-case scenario, if that’s what you’re doing, it’s unlikely you’ll be prepared to get the full value out of your marketing technology investment. Put one person in charge of overseeing how data gets managed on your team, establish some basic taxonomies and some naming conventions for tags, and files, and the like, and take the time to plan things out before you get too far along.
Those might seem like really simple things, but sometimes, they are enough to dodge really big headaches further down the road. Now, back to them.
If our listeners are concerned about the trustworthiness of their data, maybe because they know they have a problem, or because they didn’t know they had a problem. But now after listening to this conversation, they suspect they might have a problem somewhere. Where would you recommend they begin assessing their own situation? Assessing their own data in their own stack so that they can begin troubleshooting potential problems with that stack, or with whatever process, or tool, or method it is that they are investigating. Where do you start? You know there’s a problem somewhere, but you don’t know what, and you don’t know where. Where would you recommend they began maybe just doing an audit of things just to get a lay of the land, so to speak.
Dan: I definitely would always try to start with your analytics tools first because they’re typically going to have the most robust information. When we think about trying to audit a website, the best way to do it is to start out in Incognito mode, going through the website and using whatever analytics product you have and their debugger. There are many products for Chrome like Segment, Amplitude, and Mixpanel. They all have Chrome extensions that you can use to debug. They all live views. Google Tag Manager has its own debuggers. A lot of these products have that. The easiest way to start is of course to go use your own website Incognito mode, and go through the site, and start testing it.
If you want to get more advanced and you want to understand where the issues come into play, we recommend using a technology called BrowserStack. BrowserStack is helpful because you can test different devices, different browsers, different computers, and still go through that same process, and see how the data flows in. I hate to say that the best way to do most of this testing is by going and doing it. There are technologies out there that have some automated testing and stuff like that, but they’re all pretty new and pretty pricey. They’re not always the best. They’re not going to get you down to the core problem. At times, you do have to ask somebody who just knows where to look.
I love Segment. Segment is a customer data platform, one of my favorite products out there. I’ve used Segment, mParticle, and met around over a bunch of other CDPs. The intuitive way to set up a CDP is absolutely backward. The way you think you’re supposed to set it up if you do that, you’re going to screw yourself over. That’s what everybody does and then six months later, they have to redo it. The average person implements their CDP at least three times to finally figure out how to do it correctly. I hate to say it, but the best way to audit is to build good rigor around your analytics, understand how that data flows in, and use the auditing tools to be able to do that. The solution is just having more planning and being more proactive in the first place.
Ben: It sounds like you might run into situations where people treat these tools out of the box solution when they’re really not.
Dan: Yeah. They think they’re smart.
Ben: You’ve got to be more strategic, maybe about the way that those things are set up, maybe having a little bit better understanding of even what problems you’re trying to solve. Could you maybe say if that’s accurate?
Dan: I would definitely say that’s accurate. Most people think that these tools are smart and they’re not. They’re basically dumb visualization tools that have different types of visualizations. The reason why you would choose Mixpanel over Amplitude is there are a few data architecture things that are different about them which matter. Ninety-nine percent of people would never even know that those two things are different. They would choose them because one has prettier visualizations than the other. They’re both dumb analytics products that just report your data back to you. If you put dumb data in it, it’s going to give you bad data back.
There is a tool—I think it’s called taginspector.com, which is super cool, which you can load up your website and it shows you how the data flows from your website, how it flows into each tool downstream. That tool is super cool because it shows you your true data path. That way you can start to understand, maybe the data is coming in and it’s broken because it passes through this other tool to get there. There are some cool tools out there that do make this a little bit easier.
Ben: Generally speaking, what are some best practices that marketers can follow to ensure that their data stays clean and they can ensure that they aren’t filling these systems full of junk in the first place. How can they make sure that they avoid the first problem by making sure that the data that gets into these tools is not compromised in some way?
Dan: A common problem that we’ve seen for years is UTM parameters. UTM’s are the culprit of many bad data. We created a free product out there called UTM.io which makes it easy for you to basically create your UTM taxonomy and then to give an online builder to your team which is Workspace. It has rules It has templates. That way they can’t break it. When you’re working in an organization, what you want to try to do is come up with processes, tools, and systems to make it so that people just can’t break things. That’s where I would say leveraging a product like UTM.io is very easy to do because you can set your taxonomy, and then the rest of your marketing team can basically build their links. Your UTM campaign data which is some of your most valuable data comes in clean.
The same thing goes for when you think about trying to make sure that you have good accurate analytics data or marketing automation data. Try to figure out what are those products or services that you can put as, in essence, intermediaries, that protect your downstream tools. That’s where some products like Segment have protocols that will enable you to protect your taxonomy. Amplitude has its own taxonomy feature called Schema. A lot of these tools have that but when it comes down to trying to keep the data clean. Start with your taxonomy, don’t rush it. Try to take your time. If you went to McGaw.io, our website, and went to the footer, there’s a downloads and resources section. We have all kinds of webinars which were like, how do you design a taxonomy? How do you build taxonomy for personalization?
The content and education are there to be able to get this done, and it’s not that complicated. Don’t let me oversimplify it. It’s not easy, but that your best practice when it comes to getting clean data is prep. It’s the planning stage of it all. That’s what I would highly advise.
Ben: That makes a lot of sense to me. If you have multiple team members like they’re all running off doing the wrong things with these powerful tools. It’s pretty easy to see how that turns into a mess really quickly.
Dan: A lot of companies have a sales and marketing operations person that covers these things. We definitely recommend trying to have one person who—as we call them the data lord—the person who ultimately gets the help to call the shots. In a big company or even a small company, you have sales, marketing, CS products. Naturally, there are multiple people that are involved with this data, and you need somebody who does wrangle all these people together, and maybe maestro it a little bit better to make the data work for everybody.
Ben: That’s great advice and that does it for all the questions I had prepared. But this has been a great conversation. Before I let you go, if there’s anything else on this topic that you think is important for our listeners to know that you can leave them with. I realize this isn’t phrased in the form of a question, but if there are any things that you would like to leave our listeners with as a parting note, that would be fantastic.
Dan: The biggest thing that I would recommend with analytics is to take it slow. Take your time with all the data stuff. Don’t speed it up. Don’t rush. The number one problem that people try to do is they try to rush it out. I would definitely take your time. Every company who wants to expedite their contract with us from six months to three months we’re, whoa, we’ve got to slow this down. Even when they do expedite it, we still wind up doing the same six-month projects because it just takes time to do this stuff. There’s a lot of back and forth. There’s a lot of communication. Estimate a longer time frame to be able to get this stuff done. If you’re interested in getting your data in order, I would recommend checking out my free book. If you go to McGaw.io, you can request a free copy of my book on our website. It’s called How to Build Cool Shit. Maybe that’s something that will help you get this path started and give you some real case studies and tactics that you can use in your company.
February 23, 2021