Lots of marketers talk about "big data" and its promise for improving the way marketing decisions are made, but few have truly explored the full potential that data holds.
This week on the Inbound Success podcast, MDT Director of Digital Marketing Don Seaberry talks about how he is using R programming language, SQL, and Python to create advanced models for extracting insights from large volumes of marketing data.
Don is a self-taught data scientist, and explains how he learned to code, and why and how any marketer can do the same.
Check out the episode to get the details on how Don is using data models, hear examples of the ways that it has changed how he approaches marketing, and what it takes to level up your game as a data scientist.
Kathleen (00:00): Welcome back to the Inbound Success Podcast. I'm your host, Kathleen Booth. And this week, my guest is Don Seaberry, who is the director of digital marketing for MDT marketing. Welcome Don.
Don (00:22): Thank you. Thank you very much. Glad to be here.
Kathleen (00:24): Yeah. Welcome to the podcast. I would love it if you could tell my listeners a little bit about yourself, who are you, how did you wind up doing what you do, and what does MDG marketing do?
Don (00:37): MDT marketing is a marketing agency primarily focused on digital, but we also do print marketing. Based in South Florida. We've been in business, this is our 25th year anniversary. And we work primarily with higher education clients. Colleges, universities career schools, that type of thing. I've been with them somewhere between four to five years. So, so you know, I'm getting a little long in the tooth with them, but I actually have one of the shorter changes 10 years, cause we probably average 10 to 11 years as far as just tenure with the company. People come and they stay.
Kathleen (01:18): Wow, that's impressive.
Don (01:20): Yeah. I've been in marketing about 15 years. I actually started, I was in sales and a law firm hired me to do inbound business development. Had no idea that also meant manage, they had a raging out of control, $40,000 a month Google ads campaign. I didn't even know what PPC was. So I learned on the job. I had a knack for it. I liked it. And it was a really good way to do something online. And you get out of the rat race that was sales. Because I did that for a long time before this. So I basically fell into it, which I'm sure most people were at that time probably were falling into it.
Kathleen (02:00): Yeah. And what do you do at MDT today?
Don (02:05): So I basically manage digital strategy for our clients. You could say all digital with the exception of social. I kind of stay out of that world a little bit. We have a specific name for that, but if you're talking Google ads, Yahoo Search, Bing ads, Quora ads, your Reddit, anywhere else, you can do Hulu, Pandora, anywhere that you can, you can do any type of digital marketing, I pretty much drive strategy for that.
Kathleen (02:33): Great. And one of the reasons I was really excited to talk with you is that you are highly analytical at what you do. And that has really helped you to be able to leverage data more, I would say than the average person, certainly more than I do in my day to day. I'm decent at analytics, but I'm certainly not on your level.
And specifically, like, doing some custom programming to extract data and really use that in decision making. So maybe you could talk a little bit about what you're doing and what you're working on.
Don (03:12): Sure. obviously probably like most, I came from Excel work, and Excel is a great free tool. It has this one limitation with stuff. And one is one limitation is the ability to handle large data sets and large amounts of data. One of the things that I wanted to do when I initially got to MDT (just as a senior analyst before my current role) was just to be able to go back and look at years of data and determine correlations between different pieces of data.
Is there really a correlation between, you know, cost per conversion and when I run in this area of the country at this time of day? But I wanted to be able to see that across all the clients, kind of get a snapshot of it, then be able to break it down and just, you know, determine trends. So we started, I had a little bit of SQL background. I took the SQL courses in college.
So I started in SQL, a little bit clunkier because you have that database piece of it where you have to pull the data in first and then you can write queries to do the analysis that you want. So I had a, a genius of a mentor named Travis Sari, who his masters I think is actually in statistics or data science.
Kathleen (04:28): Oh my God. I feel like that's my worst nightmare. Those are the classes that were the hardest for me.
Don (04:35): I hated them. Fortunately, I had a friend who was a professor of mathematics at Spelman university who helped me survive statistics class. One on one tutoring was awesome with Travis. You know, he kind of turned me on to Python obviously. And I was a little bit more comfortable with the programming language just because I just felt like it's a little bit more intuitive. So we're using that a lot to really just kind of get through large data sets right now and do analysis on them.
And with the eventual goal, there are machine learning applications, artificial intelligence applications.
Is there a way that we can take all of the data that our sources do give us, as we know the Googles, the Bings, they only so much data, but there's a mountain of data that is there. So is there a way that we can take those and start to do predictive analytics to kind of return if we do this, what can we expect as a result of that? So that's the direction that we're heading right now.
We're really more focused on using it for just data analysis, looking at trending and using it to inform, you know, not only day to day how we manage but now, we're starting to look at it to inform how we actually build things from the ground up.
Kathleen (05:47): Now I am sort of blown away by this. And I guess my first question is really, you know, given that these platforms do provide us with so much data, right? You mentioned Google and Yahoo and Bing. And they have some built-in like native tools to look at that data and, and extract insights. And then of course, as you mentioned, there's Excel, there are plenty of third-party platforms that are built to help you crunch these numbers and figure things out. What, why is it that those platforms are, do, won't do what you need to have done? And instead you've gone the custom programming route.
Don (06:29): They have their limitations, right? You know, I actually like Google Data Studio. There are things that I had built in Google Data Studio. Like if someone comes to us and they say, we want to do a PPC audit, we have a built in PPC audit that as long as they give us access to the Google ad words, account.
Kathleen (06:51): Just press a button and spin it up, right?
Don (06:54): Just connect the data and boom. And instead of it taking, you know, it used to take hours, you know, sometimes days to do that same you know, to do the analysis required to come up with the results or to give them feedback. You know, now we do that in like five to 10 minutes. So it was a definite time saver, but it has its limitations. I'm a huge proponent of granularity. I wouldn't get as granular as I can't get as deep into it as I can. And when you start talking about the type of data sets that we're looking at, the potential, you know, millions and millions of rows of data you know, you want a program that's robust enough that can look at everything and just spit it back out. And you don't want to start your application and step away and come back to it two days later because it'll take us that long to finish the work, if it finishes at all without just throwing up all over itself. So I like the granularity. I also can like to understand how things work. So it's one thing for you to spit the data out for me. It's another thing for me to understand the code that was behind. I feel like I understand a little bit better when I understand what actually caused that analysis to take place, if that makes sense.
Kathleen (08:07): Hmm. That does. So give me an example of like something that you would want to know that you would need to create a custom coded solution to figure out the answer to that question.
Don (08:19): Well, I'll tell you one thing, one of the things that we're working on, you know, a big time suck for any agency is campaign set up. So if either you bring in a new client very often, you have to build out campaigns for them. So one of the things that we're looking at, especially with Google, for example, Google ads, PPC, is we went back and we're going program by program. So, you know, a lot of our schools have the same type of programs where there's surgical techs and medical sonography, that type thing. So we're going in, we recently went in, we looked at six years of data at program level and things like we're trying to develop what we call it. So are there keywords that have proven to be successful year over year, over year, over year, as far as the cost per conversion and number of conversions, what ad copy, what combinations of ad copy have proven to be successful?
Don (09:19): When you talk about a Google where you can do a responsive search ad with, you know, 15 headlines and four descriptions, and then you multiply that over different programs, over different schools over time, that's a mountain of data to get through, right? That's not something you want to pull into Data Studio, or that you want to pull into Excel and spend hours and days trying to figure out. So what we did is we, we downloaded it at the program level across the entire MCC, six years of data, let's say for keyboards. And then we went in, we trimmed that too, by conversion type, we sorted it back in version, count, trimmed that by year and match type, top 20 performing keywords by match type by year. And then we merged all of those into one data set, spit it out to a CSV. Now we have a keyword bank that we can use.
Don (10:14): For example, if we need to build a surgical tech campaign. And we also sped out, we gave ourselves an idea of you know, what the average cost per click was just to kind of get a sense of maybe where we need to start as far as bidding, or when we set it up. We did the same thing with ad copy. The ad copy was really interesting. Cause there was just thousands on top of thousands on top of thousands of pieces of data. Once we got the custom coding written and R spit it out, like nothing.
Kathleen (10:43): Like what did you learn about ad copy? Give me some examples of takeaways.
Don (10:47): The big thing that we're looking for, really a couple of things. Number one, what combination of ad copy tends to work the best and what resonates with students the most? So one of the things that we found is I would have expected things like online.
Don (11:07): You know, you, you start to get a sense of the things that now I can address in my landing page content based on what they interact with. We started to find that people, students wanted to know what was in it for them. Can you help me with job placement? That was really successful. Can you help me with financial aid? They wanted it to be easy. They wanted it to be simple. They wanted to know that they weren't going to have to go into the poor house. Well, you know, right off the bat, that there'd be a way for us to actually be able to go to school. And we started to see that combination of those types of things. And students really love the hands on piece of it. So we started to see what they reacted to. And it was typically always coupled with, is the school accredited? Cool. Is my degree going to mean something or my certification going to mean something? And they want it.
Don (11:56): They, in most cases they had some sense of what the brand was and that helped as well. So we started to get a sense of what's actually important to the students. So we know going forward, I need to continue to leverage that, but I need to also make sure that I take that in the other content, not just ad copy, but landing pages. If we're doing direct mail to them, which we often do now, we make sure we bake that type of information into direct mail because it's important to them. And the end result is, you know, we see greater engagement on the front end and it helps us to improve return on investment on the backend once they actually engage with the school and they're trying to get them to start.
Kathleen (12:36): So that sounds amazing. And it sounds like information that any marketer would want to have. Can you speak to, how hard is it to set this up? How long does it take? Like is, is the juice worth the squeeze, I guess, is what I'm getting at?
Don (12:53): I'll answer that this way. Yes. The juice is definitely worth the squeeze. One of the things that we have actually used it for, we don't always get to connect to the client's CRM and see the final disposition of a lead, you know, does that person become a student? Do they graduate? You know, we don't always get to see that, but we do have a couple of clients that we get pretty good visibility where we can be, you know, they'll typically maybe once a week or once a month, something to that effect, they'll actually send us back enrollment data. So because of how we tag things on the front end, we actually know what channel that came through. So we've been able to use R to determine, okay, is this a channel that actually drives return on investment? You know, we market to return on investment first.
Don (13:45): Then we back into efficiency, things like cost per lead and then cost per click. That type of thing. The most important thing is did it make the client money, period. You know, return on investment. So because we have the ability to dump it into a script and us, they send us that information back. We have a script that we can dump it into. It will match it up based on the source that it came from. So we have, let's say, for example, Google ads, then we have a lead management tool that goes into first and it had this lead ID in this information. So we have a script that takes that lead management tool data, and the data that the client sends us. It matches it up. And then we start to see, okay, is this something that we need to continue to run for this client?
Don (14:28): Do we need to reduce budget or just cut it out? We actually had one client. Well, we stopped the channel that was actually driving conversions on the front end, but over a year, the return on investment just didn't make sense. So in that case if I'm talking about, you know, kind of being Hippocratic in my approach, you know, do no harm to my client, then the juice is definitely worth the squeeze. The biggest piece of it really is just the patience. You have to have the patience to do the work, to get comfortable with the scripting. But honestly, you know, R programming, Python, any of that. There's so many places you can go online that are giving you a jumpstart and it's not, you know, it won't be an issue of, I have no, I don't know where to start. I don't know where to begin. If you go on YouTube and you put R for data science, you're gonna find a mountain of information on that. And a lot of, a lot of times, same thing with just internet searches, you will find the information that you need. And a lot of times what you're looking for, you'll find that someone else has already done it and put it out there.
Kathleen (15:30): So, okay. Real talk for a minute. Like, I know I can find these things on YouTube, but for somebody like me, who's intimidated by statistics, who, you know, I mean, I get basic analytics, but the notion of learning to code scares me. And you know, is it realistic to think that somebody like me could go online and really learn this? And, and how long does it take to learn something like this? Is this like a week, a month, a year?
Don (15:59): Okay. I'll tell you what I did. If you have, if you have some comfortability with, you know, custom functions in Excel, for example, you're gonna probably be a little bit more comfortable when you get into the SQLs and the Rs of the world, because you just kind of understand the way the architecture, if you will, works. I understand how a database is set up. I understand what I'm trying to query.
That kind of carries over to R. You understand a little bit more of what you're trying to query and the information that you're trying to get at? I think any reasonably intelligent person can do it. You have to have one, really the will to do it. And just the time. You might spend a week, if you just kind of pick and poke around and find the pieces that you need, you might spend a week figuring that out.
Don (16:53): But the beauty of it is, kind of, once I figured it out, a lot of times once you've done it, like these two scripts, we've written, the one where we dealt with return on investment, I can just pop the information in now. Boom, once it's done, it's done. The asset database, it probably took me, I'd say a day, right? Each one of those scripts, one for the ads, one for the keywords. But now that I've got them, all I have to do is drop in whatever data I'd want. There might be a couple of changes that I make to it. That might take me 15 minutes and boom, it does the work. So once you get past that piece of it, you go.
Kathleen (17:31): So why don't you think there are more off the shelf tools that can do this for you? Cause it seems like from what you've said, it seems like the insights you're able to extract are pretty darn useful in terms of making decisions about your campaigns and optimizing your ROI and even optimizing other campaigns that have nothing to do with PPC. I'm sort of surprised that there isn't a commercialized like, software product out there that that makes it easy for people like me.
Don (18:01): You know, there probably is. Because we have a genius of a guy in our marketing tech department who develops all of our like reporting products. And he primarily Power BI to do that. Genius. I mean the stuff he can drill down to, absolutely genius. Obviously there is a learning curve about a power BI, or you're talking about a Tableau or something like that. I kind of take the approach, again, I like granularity. So the more granular I can get, the better. The more cause you know, the more, you know, any solution or any analysis is only as good as the data that you put into it. So the more points that we can pull together and we can give that algorithm time to kind of work through, the better. So you probably, you probably have some, I'm just kind of, I kind of nerd out on that stuff. I like that stuff. And again, I kind of like to understand how it works and I have an idea in my brain most of the time of really where I'm trying to go. And I'm just, I'd rather just have control of where the levers are a little bit more than necessarily on the box, but I'm sure there are some that will do it. But I'm sure there's, there's probably ways.
Kathleen (19:17): So who do you think should, should think, can seriously consider doing something like this or learning how to do this? Because I have to imagine that it's not really right for every marketer, you know, you have to be dealing with, I would guess, certain volumes of data. And, and you obviously need to have the ability to like put what you learn into some form of action otherwise, who cares, you know? So like talk me through, if you were talking to other marketers, which you sort of are through this podcast, like who should really think about something like this?
Don (19:55): Definitely a person, like if you live in the numbers every day and by living the number, I mean, you're actually pulling the levers on campaigns every day. A lot of platforms are pretty much forcing us more and more towards automation. So, you know, some of those day in day out practices that we would have really been engaged in you know, keyword bid management, you know, average position, Google and tools like it can sunset. So you can't even, you know, that that's not even a metric you can get to anymore. But a lot of those, those day to day things that we used to do, those things are going to go away, what it won't do. And actually it will actually write ad copy for you if you want it to the places you can do that.
Don (20:48): This is going to give you the ability to focus on what's important. Ad copy and creative and your keyword less than blah, blah, blah, and all and all of this stuff. So if you, if you're living in the weeds every day you definitely should be learning some of these data science tricks. Absolutely. No question about it. You should get comfortable with it because if we're doing less day to day management, that means we have more time for analysis and we can start to dig deeper and do a bit. And I would say if you're director level, you know, kind of the level I'm at, I still think there's value in it because one of the things that you want to start to consider is, okay, is there any way that I can use you know, data science to understand what are the things we're talking about?
Don (21:39): Is, is there a way to use it to figure out what combination of channels works the best together? So if the person comes in through Google ads and doesn't convert, do you know, do I see conversion rates go up if I send them email or if I do direct mail or not just through Google ads, remarketing, are, am I seeing better numbers of ads, Google ads, YouTube then being remarketing, you know, is there a way to determine what channels work better using data science? Those are some of the things and that one's a little trickier because you have to be able to track that lead all the way through so that, you know, there's something to think about there. But I would say anyone probably at least from director level and down that lives in this world, especially in agency world, you should be learning this.
Kathleen (22:28): So earlier you mentioned that you're now using this for some predictive things. Can you tell me a little bit more about what you mean by that and how you're using it?
Don (22:39): Right. So, so one of the things that we're trying to figure out and, and the coding is kind of still out there. We have to, we have to figure this piece of that out. But let's say for example, we're in Google ads, we're in Google ads. And, you know, if you have, you know, 20, 25 accounts you live in every day, you're constantly making changes and you're constantly trying to track those changes. One of the things that we're trying to determine is, is there a way that we can start to pull change history and to, to, you know, master spreadsheets. Pop it into R, use Python and then say, okay, here's what I'm going to determine. This type of program. When I made this type of change across the MCC, historically, what have we there's as far as maybe if I change the bid strategy, and if I switch from this particular bid strategy that, that particular bid strategy, what did I, what have I seen historically? Is there a trend that I can see across accounts that says switching from this to this, I tend to get better results. So kind of trying to predict future performance based on what we've seen in the past, just on that body of data.
Kathleen (23:56): And it sounds like really trying to extract some best practices.
Don (24:01): Absolutely. Yeah.
Kathleen (24:03): And as you mentioned in an agency setting, I can really see where that would be useful because having owned an agency myself for about 11 years, you know, one of the big challenges is always, how do you disseminate information about what works and what doesn't across different people in teams. And so I could see where that would be really valuable.
Don (24:22): Definitely
Kathleen (24:22): So interesting I'm, I'm like intimidated and intrigued all the same time as I imagine, many people are.
Don (24:32): But you have to start somewhere.
Kathleen (24:33): So, yeah, and I would say the other thing I was going to say is if this is something that interests you, if you're listening, I actually shared this with Don, a great community for conversations around analytics is Christopher Penn's Analytics for Marketers Slack group. You can probably Google it and find it, but I will put the link in the show notes to it.
Kathleen (24:53): And Christopher is also another amazingly accomplished brilliant data scientist. And he started this community just for other people, interested in analytics and you don't have to be as experienced as Don. I'm in the community. And I am a complete amateur. But it's great. It's a great resource. And you can ask questions and learn a lot there. Shifting gears, Don, I have two questions I always ask all my guests and I'd love to hear what you have to say. The first of which is really we talk a lot about inbound marketing on this podcast. Is there a particular company or individual that you think is really kind of setting the standard for what it means to be a great inbound marketer these days?
Don (25:37): There's, there's, there's a couple of people in digital world. One person that immediately comes to mind and I think, you know, him is Cory Henke. Especially because video is so important. And if there is a video Yoda, there is probably Corey Henke. Incredibly, incredibly intelligent guy. There's another gentleman that I follow quite a bit. I actually met him at Pubcon. He presented and he presents quite often. His name is Joe Martinez. He really lives in digital world, especially on the PPC. And he has just, you know, stuck his toe in everything where there's Google ads being as Apple ads, Quora. He's a Quora evangelist of sorts at this point because that platform, he really believes in it and they have definitely improved that product. I can tell you that. So I follow Joe. His company has a, they have a blog that they typically put out.
Don (26:45): So if you connect to Joe, he'll, and he'll normally like you know, let you know when a new blog comes out and this year, another young lady that works with him as well, I believe her name is Michelle Martinez, is also really, really smart. My, my hidden little gem, just really more so for the way that they think, is Fernando Machado and the team at Restaurant Brands International. They own like, Popeye's. They own Burger King. There's all these chains that they own. And they're just, it's just genius to just, you know, stalk their LinkedIn feeds and just see how they think and the difference in the way that they think, especially as it relates to driving social engagement. And you know, if you think about the craze over the Popeye's chicken sandwich and the way that went bananas, and then the way Popeye's leveraged that, that was Fernando's team behind that. And they've done the same thing with some things I've seen. They put out the moldy Whopper, where they were showing that that is organic, that it will mold over time. That it wasn't like the McDonald's cheeseburger that three years later looked the same. Fernando Machado was behind that. I just loved the way that they thought.
Kathleen (28:06): That's cool. I had not heard of them. So I'm going to have to check them out. And I love learning about new people to follow. So that's great. And also now I have a craving for a Popeye's chicken sandwich, which I have not had yet. I keep hearing about how great it is. So I'll have to add that into my rotation this week. Second question is that the biggest pain point I hear from marketers is always that it's so much to keep up with so much changes, and it's really hard to stay educated on best practices. So how do you personally stay up to date and educated?
Don (28:41): I'd definitely start since, I mean, we all have to be on, it was an 800 pound gorilla in the room, right. Especially as it relates to digital. So obviously, subscribe to the Think with Google blog if for no other reason than to just try to stay out in front of what Google is doing. I'm a huge fan obviously of Search Engine Watch. Well, most people probably think of that just in the context of PPC, but it's actually a tremendous amount of information. You can actually go on there and find articles on what we've talked about today. R programming, machine learning. There's even an article that actually shows you how to write a machine learning algorithm for your campaigns based on weather patterns.
Kathleen (29:30): Wow.
Don (29:31): So there is so much information on those three sites. Love those sites for just pure data science and learning. There's a channel that I particularly like on YouTube. The gentleman's name is Ken Yee. And he just, he just breaks it down to basics, and there's a lot of good how to's and information that you can learn from him. And then tracking and tagging, conversion tracking, it's all about Measure School for me. Measure School YouTube channel.
Kathleen (30:06): Oh, I will definitely put links to this in the show notes for all of that. Very cool. Well, if somebody wants to learn more about you or reach out and ask you a question or connect with you online, what's the best way for them to do that?
Don (30:18): Probably LinkedIn you know, I'm on LinkedIn. I'm very active on LinkedIn. That's how we connected was on LinkedIn. You know, just look me up. Don Seaberry. I'm there. I'll probably accept the invitation. I connect with as many people as I can. So feel free to reach out
Kathleen (30:37): True story. That is how we connected and how Don wound up on the podcast. So check him out on LinkedIn, Don Seaberry. And if you're listening and you liked this episode, or you learned something new, please head to Apple Podcasts and leave the podcast a five star review. That is how we get found. I would love it if you would do that. And of course, if you know somebody else doing kick ass inbound marketing work the best way to get them on this podcast as the next guest is to tweet me at @workmommywork, because I do find my guests through word of mouth. So reach out to me and let me know who you think I should interview. That is it for this week. Thank you so much, Don. It was great talking to you.
Don (31:15): Thank you for having me. I thoroughly enjoyed it.
Want to stay updated when the podcast is released?
Drop us your name and email address below and we’ll send you the show notes every Monday!
Here Are Some Related Articles You May Find Interesting