Roxanne and Matt recently visited the drop-dead gorgeous The Penny Hoarder offices to interview Stephen McDermott, their IT director.
Stephen was recently presented with The Penny Hoarder’s ‘Kind of a Big Deal’ peer-to-peer recognition award on the basis that he embodies the company’s values. “He’s the go-to when you have a tricky problem you’re trying to solve. Steve is always generous with sharing tools, resources and knowledge to help his colleagues. In fact, he recently started coding classes for The Penny Hoarder employees.”
We couldn’t agree more, and we’re super happy for Stephen!
Roxanne Williams: So, you seem to have had a very linear career progression. I stalked your LinkedIn. You started in dev, then you became lead, manager, CTO, and Director of IT. So, being at the C-Suite and Director level now, do you ever miss simply just getting to code?
Stephen McDermott: Luckily, I get to code all the time.
Stephen: Yeah. I didn’t do this at Nielsen, I was more of a manager manager (I had 25 or 30 people – it was sort of different) but it was very compelling to be here because I was allowed to be technical, yet manage at the same time. I love to code, I love to build, I love to create, and it really helps on the management side to know the architecture. I used to be an applications architect, a solutions architect, and I’ve also been coding for 20 years. So within that vein, I like the synergy it gives me to know technically what everyone’s doing. I can make some Best Practices recommendations, but I can also code every once in a while to test out new technologies. So yeah, I do love to be hands on, and probably 25% of everything I do now is still code or creating.
Roxanne: That’s awesome.
Stephen: I love to code in Java. I’m learning Python now for machine learning. I have a statistics background and I want to start doing that whole machine learning Python thing. It keeps me interested every day.
Matt Vaughn: Were you involved with the analytical and statistics side of things in your at Nielsen?
Stephen: I actually worked on their nielsen.com website. That’s what I was a software development manager for – but I did no coding. After I left the Tampa Bay Times downstairs, I decided to give myself – I’d been going from job to job for over 20 years now, it’s always from one right to the other, and I didn’t really know where I was going to go. So after I left downstairs, I said, “well let’s try Nielsen for a little bit as a consultant more or less, and let’s see what I like and what I don’t like.” And Nielsen was great. But I did realize, you know what? I’m at the point now that I’d really like to get a little bit more hands on. I love to design, I love to build. And so, that’s where we are now.
Roxanne: You had mentioned machine learning. Are there some possible applications you could think of The Penny Hoarder (that you’re allowed to talk about)?
Stephen: It’s very generic. Machine learning is more or less a regression analysis, which is trying to solve for something. So what are we trying to solve for? Well, we’d like to figure out a lot of things, one of which is how to optimize and increase engagement. Machine learning allows us to create what’s called a null hypothesis, which is “I need to disprove the fact that these variables have a correlation to engagement.” So then you say “Well what do I think will correlate to engagement?” Maybe it’s how fast the page loads. Maybe it’s the context of the text. Maybe it’s the color of the page. Maybe it’s the sentiment of the text. Is it happy? Is it sad? Maybe it’s how things are designed.
You need to get all that data (and you need to get lots of it) on the left side of the equation and then you need to use it to predict engagement. Now it could be it’s not engagement I want, I could want pure revenue, and to maximize conversion rates – that’s just an example.
Machine learning is going to allow us to each create sort of a hypothesis of “I believe these variables correlate to what I’m trying to solve for.” And machine learning will allow you to say, “This does not correlate whatsoever, so don’t worry about it. This does.” It also allows you to make predictive analysis which is, if I decrease page speed by a second, what do I think will happen to engagement or RPM? Our initiatives are based around that: how do we figure out what are the most significant factors in determining an outcome-driven model? I want people to be on my site, and I want them to be on it more. I want them to enjoy themselves, I want them to click more, and I want to convert them more. I want to do everything I can to make it very pleasant and an optimal experience. And in the end, obviously, there has to be some type of engagement to revenue to something we’re trying to attain.
Matt: How long someone freezes on one section of a page to lead into a click, or learning more, or a sign-in.
Stephen: Exactly. Because really, that last variable is just a culmination of how well your site’s designed. You just need to know how to measure it. So yeah, to your point, if you see someone getting stuck there and going back and forth for a while, and then never going back down, there’s something on that page which is breaking off their concentration or it’s telling them, “I’m going to stop here. I don’t want to scroll down to the next 90% of the page.” How do we stop that? How do we get them to look at the whole thing?
Matt: That’s where your friction point is.
Stephen: Exactly. Machine learning is a good tool to use to solve all that. You use it and then you match it to resources you have, and you say okay, where should I put my resources? It lets you optimize from a department level what the resource should be used for, and where it offers value.
Matt: So, as you were kind of talking about earlier, you started at UnitedHealthcare, and now you’re working for a startup. Can you speak to the differences between the two when it comes to culture, your management style, and the overall 30,000 foot overview?
Stephen: A lot of it has to do with innovation and risk. Healthcare is very heavily regulated. It costs more, whether you want to put a dollar amount or hour amount on it. Development costs more because there are many more processes in place in a healthcare company to, in essence, prevent you from doing things you should not.
Matt: The dreaded compliance nightmare.
Stephen: Yeah, so that’s the compliance part. But then there’s the implementation of compliance where they refresh everything in healthcare from the diagnosis codes.
Roxanne: ICD codes?
Stephen: ICD, yeah. There was ICD-9, 10 – that happens every once in a while. Proc codes as well – things like that change. All those are things that you have to put on top of your heavy development costs. It costs a lot to handle these and the risk of doing any of it wrong is so massive. The controls are in place where, in my experience, the cost to produce anything in healthcare from a resource perspective is very high. Whereas when you talk about lean and agile, in terms of producing things quickly, I don’t want to say ‘easier’ – let’s just say the environment is set up better to produce code very quickly because our risk tolerance is different. We don’t have to worry about PHI (personal health information).
Stephen: Yet. Maybe we have a vendor that has email addresses. We have to protect that, but that’s different than a person’s claim history over 30 years and every information about their family members.
As a startup, you have built this infrastructure from the ground up to be optimized for doing things fast and keeping downtime low, as opposed to healthcare where you already have all these legacy systems you must support. These may have been built 20 years ago, some of these by developers which are no longer there.
Matt: Who doesn’t love a basement full of ancient AS/400s?
Stephen: Exactly. Every one of them has cost-benefit analysis. Do I need to transfer this? You’d say, “Well yes, once I do that, I will be able to find a resource to code it, but for now, it works and I have other things to take care of.” Legacy system A connects to legacy system B connects to our nice Amazon AWS cloud provider.
Roxanne: Ooooh, you guys use AWS?
Stephen: Yeah. Here, we started with a blank slate and we said, “Okay, let’s build this to be scalable. Let’s build this to be quick and lean.” All those fancy keywords, right? But the fact of the matter is there’s this demand curve on our site. Like every site, there are certain times of day when things are heavier on the process than others. Along with that, there are certain times of the year where we have more work for the developers than others. So what we like to do is not waste those troughs right there because that’s wasted processor and memory you’re paying for. From a demand perspective, we do use some consultants, we do use mostly here, but also, that needs to scale on the demand curve. We try to sort of hug that demand curve all the time so that we’re not being inefficient with the money. That is a big difference because at UnitedHealthcare, I’d have a very large budget for something, and there’s a lot of room for error in that budget. Being our size, here, we do well, but there’s less budget. We want to be more efficient and we want to have less error in everything we do, and less waste. That’s what we try to attain in IT.
Roxanne: Curiosity then, how much staff total do you have?
Stephen: I think we’re at 100 to 110.
Matt: And then in IT?
Stephen: We’re at 12, and we have two contractors, pure software developers. I’ve been in offshore-modeled places where we had to offshore a lot of our development resources. That works in some ways, it doesn’t work in other ways, and in my world, I’d rather have onsite resources all the time. Where they do come in and play nicely is when you have the subject matter expert on site and they are able to use the offshore resource as an extension of what they do – they manage them. I have something I want to do, I have someone who can help me with that, but I’m the subject matter expert because I’ve been doing this for 10 years, and I can see what they’re doing, how they’re doing it, and I can direct them. That sort of extension has worked nicely.
Roxanne: We talked with Chris Jenkins as one of our other interviews. He’s the Chief Digital/IT/Technology/Everything Officer at The Symphony Agency, and he mentioned that they used to have offshore staff like that, but they decided to bring everything in-house because when you have that offshore staff, they’re not invested in the product whatsoever – they just code and send it to you. If they have any good ideas, they don’t come to you with them.
Matt: You don’t get innovation from offshore.
Stephen: Yeah, I like the onshore model a lot better.
Roxanne: So, you mentioned AWS. What services do you use? And what other technologies?
Stephen: Everything within AWS. S3, EC2. We’re looking at Aurora Serverless Database right now, which I’m very excited for. It’s a distributed database where it’s no longer on an actual physical server you can see. It is an endpoint. That’s all it is. It is distributed all over, and you can scale it up and down depending on demand. When it is not being used, it actually drops itself. It’s always readily available within 25 seconds to come back up, but the pricing model on demand, if your caching is such that you’re not always hitting your database, actually saves an awful lot of money, not just from an up and down standpoint, but it’s cheaper from an hourly basis compared to the RDS we have right now. We’re also using Cloudflare – not Cloudfront – for our CDN.
Roxanne: It’s cheaper, right?
Stephen: It is cheaper, yes. Cloudflare is one of my favorite vendors I’ve ever signed up for. They just have a product that works, they have a pricing model which gives me a lower unit cost than anything, and you just plug into it. I mean, I can sit there and block IPs. I can do a static site with the press of a button. There are all these little things they keep adding from a product enhancement standpoint, like Argo. If I pay an extra little amount, I can make my whole site faster. They just have so many enhancements they keep adding. I love the product.
What else do we have? RDS Redshift on postgres. I know I’m missing quite a few. Machine learning, but really, we don’t use that too much. We don’t use Glacier yet, we have another archival system. Oh, Lambda and API Gateway! How did I forget those?! I love Lambda. Serverless, everything’s serverless.
Matt: The dream.
Matt: Excellent. Among the many awards The Penny Hoarder has won, such as #1 fastest growing private media company on the Inc. 5000 list in 2016 and 2017, Tampa Bay Times 2017 Top Workplaces, and Entrepreneur 2017 Top Company Cultures, and with the company’s 4.5 Glassdoor rating, in your own words, why is it so awesome and rewarding to work at The Penny Hoarder?
Stephen: We’re all working on things we like to work on. You are at the office for 8 to 10 hours a day. You are working. At this part of your life, it’s almost the majority of your time when you average out the week. If you are doing something you find enjoyable and you are given the ability and the ownership to do it, that’s fun. That creates a good culture.
Also, in IT specifically – and I think this is generally throughout the company, but let me speak to IT – I hire people who do not get offended by conflict – and it’s good conflict. We all don’t know everything, right? We cannot get offended when we sit down and balance technical problems off of each other. And sort of that law of economics: when people are forced to ‘negotiate’ and do a trade-off, a cost and trade-off on things, they generally come to the most optimal solutions. I have made sure to hire people who are very open to constructive criticism. It’s not even criticism, it’s “Have you tried this? Why don’t you do this? Tell me why this is the best solution.” It’s not a stressful situation, but we openly encourage it in IT. So we come up with good solutions. Again, I’ve just been doing this for 20 years, it doesn’t mean I’m the best at any of these things. I look to everyone else to tell me when I have no idea what I’m talking about. I think that’s the same model across the board which is we’re just very open with ideas.
Roxanne: That’s awesome. You see it so much in IT where you have the one manager that does think he knows everything and has no shortcomings, it’s like his way or the highway and everything gets bottlenecked. So thank you.
Stephen: I’m at the top of IT and then there’s lines on the hierarchy chart that go down here, right? If every decision runs through me, that’s the most inefficient management style there is. If you give the ability to the people who work for you as small production units to make decisions and guide / design strategy, and you have 9 of them, from a number standpoint, you are producing 9 times more than I would produce. So, you need to push the decision-making down. You need to make sure that they have your back and to remove impediments, but you need to create the process, the technical architecture, and give them the ownership to make decisions without me.
Roxanne: You mentioned earlier about the outsourcing. Other than those two people, have you found Tampa / St. Petersburg to be adequate for your tech needs?
Stephen: Yes. Without a doubt, actually. There is no shortage of applications for any positions I post right now. Now, that may be different for other companies. We do have an advantage here as The Penny Hoarder is known as a fun place to work. But I’ve never had trouble filling. If there’s ever any difficulty, it’s because we’re very stringent on who we hire, and we’re very specific.
Roxanne: We’ve had a few of our interviewers say that they have to expand their searches nationwide.
Stephen: We do not have that problem in IT, luckily. Now, I’ve been looking to fill a QA position for a while, but it’s because I want my QA people to be part developer. I’m not looking for the normal QA person. So, you have to find a developer who wants to be a QA person.
Matt: That was about to be my next question. What is the room for growth on the QA side? There are very few shops in the Bay that actually find passionate QA people to run their automation processes.
Stephen: In all my years of software development, having technical QA people as good as I have right now is one of the greatest benefits. Because we’re all CI/CD, everything gets committed to a branch, and then QA will say, “I accept.” Then, they are the ones that promote it. I’ve never done this before. I have strong enough QA folks that I have removed the infrastructure portion of a DevOps person I’m going to employ, I’m going to let QA deploy.
Whenever they are comfortable that it’s ready, they take that branch and they deploy – and we deploy many, many times per week of production-ready items. And it’s because they’re technical. And it’s because they’ve created all these regression suites which is, more or less, almost like a game. It’s like, I’m going to create these automated regression suites which enable me to tell a developer when they broke my site, and they’re constantly running all the time to make sure things are running as they should. Other than me being here for the rest of my life because they’re the best company ever, if I do happen to work somewhere else, I will be sure to have automated QA engineers.
Roxanne: I’ve never seen QA do deploys. That’s awesome.
Matt: What training initiatives do you employ to ensure your tech staff keeps up-to-date with the latest technologies?
Stephen: I don’t have a good answer for this because I don’t really officially train them. I have hired people who love what they do, and specifically, I try to hire people who love what they do so much that they are naturally advancing themselves all the time and coming to me with different types of technology. Really the new one is Python. I’ve gotten a lot of folks here who just started messing around with it, and they’re like, “Steve, why don’t we try Python?” I’m like, “Okay, why?” Java, that’s my core. I program in lots of languages, but Java’s been mostly what I’ve coded in in 20 years, and I realized that if I wanted to code a service, right now in Java, I have Spring, and I go and code my Spring service, right? I have my libraries, I bring in through the palm, I have all these standard tools that I’ve built on frameworks over 20 years, right? So it takes me a little bit to get it set up, then I take the API, I throw it in Lambda.
I could do that same thing in Python when I messed around with it in less than half the time. Now I can see the fact if it becomes really, really complicated – I use Spring security to encrypt things if we have to, and that’s going to be different in Python. But to get basic things up and running, Python is a very light, scalable language. It’s also the machine learning language of choice, so everyone around here is very excited to do it. There have been a lot of people just messing around with it. As for official training, we give them a training budget, and whatever they want to spend it on is good with me.
Roxanne: Nice. They can take their own classes on Udemy or whatever?
Stephen: Yep. That’s generally what they go to, or Coursera. I let them do what they want with that.
Roxanne: So they get creative time here?
Stephen: No, not yet. I try to give them things they would like to work on. I meet with them weekly and I say, “Look, is there something technically you’d like to try that’s in our backlog that’s coming up?” Or I can say, “Hey, is there an alternative solution that you see in our backlog that you’d like to try?” And I’ll let them do that, but I don’t give them specific creative time. I do try to match it up against the work.
It does come down to the fact we have an awful lot to do. Maybe next year, but right now, I need to be very efficient and make sure that they’re always fulfilling something of need from the business.
Roxanne: What do you hope to see in the next few years in Tampa Bay when it comes to technology?
Stephen: That’s a great question. I’d love to see a huge tech conferences. I always fly to San Diego for my tech conference.
Roxanne: I think poweredUP is the biggest one here. It happened a few months ago back in May, and there were 1,200 people there.
Matt: Synapse is another larger one. Did you go to Synapse last year?
Stephen: No. But maybe these are the things I’m missing.
Matt: poweredUp was in Mahaffey a couple months ago, and then Synapse is coming up again in January. Florida Funders put the first one together last year pretty quickly and it went great. We were heavily involved in it from the volunteer side and we’re doing it again this year – but it was crazy because they threw it together at Amalie Arena in the span of a few months. The entire outside of it was all breakout talks, and half of the ice and half of the interior of the actual arena section was put to the bigger speakers like Arnie Bellini from ConnectWise, Jeff Vinik, high ranking military tech leaders, and more. So, it was a little bit more Tampa-centric and not “large tech” like a Microsoft-specific event, but it was great and I highly recommend it.
Stephen: Yeah, I’ll try it. I love going to conferences. I love to learn what everyone else is doing, because you get siloed here, but everyone else is moving at different paces on different things, and so just getting to a conference and talking to folks is helpful.
Matt: Considering you’ve been in tech for almost 20 years, is there anything you’d like to say to anyone thinking of going into tech? Not even necessarily a student, it could be for changing careers. What are words of wisdom you could offer from your years in the industry?
Stephen: Gosh, I did a 20-minute presentation on this for the company the other day. The last I looked, there were 220,000 jobs not filled in the United States for software developers. I was sort of researching it and I have my theories, some quantified in white pages, some not – but in essence, why is that the case?
Demand is scaling, and supply is not keeping up. But the problem is not that the supply isn’t keeping up, it’s that the supply has fallen back. The ratio is not there. What is it that’s causing the lack of supply of developers? There are a lot of things, one of which is that programming is a trade. You don’t need to go to college for programming. It’s a big misconception people have. You are valuable in the IT market as a developer with a single tool. If you want to learn only Python, you can get a job doing specifically that at a company. Now, I don’t want the college folks reading this to get all mad at me. But you do not need to invest 150,000-200,000 for a college education to become a developer. And lots of that supply is coming out of the colleges.
You can be producing a very good salary right out of high school with a trade school, or you can teach yourself. Like the guys who sit there and program video games at 14 years old, they could probably go right into programming as a job without going to college. College delays your entry into the market by four years, and it also puts a much bigger financial burden on you.
Also, you don’t need an engineering degree. People think it’s all math and science. Now I’d like to tell everyone, actually, yes, we all did physics 3 and calculus 4, and I can do a thermodynamics equation with no problem. But that’s not really the case. You learn programming by iteration. You learn it by doing it wrong so many freaking times that you start understanding the way this works, right? So do it wrong 10 times, 20 times, 30 times, and you will slowly teach yourself how to do it and become good. You do not need to be an engineer, you don’t need to be math-oriented.
It does help to be able to be logic-oriented. What I mean by that is, every scenario in programming must be accounted for. There are no gray areas. It’s the most literal thing in the world. You either do this or you do this. If this happens, you do this. Every single thing must be accounted for. And if you can think like that, to be able to analyze the situation and say, “What are the variables I need to determine X? Well there are 10 variables I must know.” If you can quantify all the things you need to know to perform an action, you’re probably going to be a pretty good programmer as long as you put in the effort.
The last part of the equation is demographic participation. Women used to make up 37%. They now make up 22-25%. If you could take that part of the market and equal it to the male side of the market, you would solve a lot of the supply problems. If you can increase it just to where it was in the year 2000, where it was 37% of the market, that would make up a lot of the crunch of demand.
So there are certain demographic participation variables which are below where they should be, and that also causes an inaccuracy in the market. Right now, there are an awful lot of jobs out there, and there’s just not enough supply, and there are a multitude of reasons going all the way across, but I could talk about this all day.
Matt: Same thing with us. That’s how we exist, for the most part. We see all the trends, we see all the – not to speak of anyone, or even the Tampa area, specifically – roadblocks companies make that keep them from the right hires. Companies mandating that you have a 4-year computer science degree even though their tech stack is not taught in schools, or needing to have four years of experience to be a dev in languages, frameworks, and libraries that have only been commercially used for a year or two.
Roxanne: It’s sad, but we see that so much. You could have a candidate that’s 100% perfect, but they will not let go of that ‘you need a degree, and this person doesn’t have a degree’ mentality.
Stephen: My degree in college was not in computer science. I am here because I did night school after college. I was in New York City, and what do you do in New York City after college? Work for a finance company. I was a stock analyst for Union Bank of Switzerland. When you’re in New York City and you’re out of school, you go out, have fun, do what you do in New York City – the bars and all that.
I started talking with the IT folks, talking with the business folks, and eventually I’m like, “God, I really want to do this. I love this stuff.” But they couldn’t give me a job there at the time. My wife and I’s rent went from $1,800 to $2,400 and we were like, we want to start a family some time and we’re living in a studio, we can’t save anything, we can’t invest anything. That’s when I said “let’s go up north,” and that’s how I ended up at UnitedHealthcare. I was up in Albany, New York. That’s why I started programming.
I sort of went the non-traditional route as well, but you do not need a college degree for it. I call it the citizen developer, where you don’t need to dedicate your full life to programming to be a programmer. If you choose a problem you want to solve, you can probably choose a language to solve it. You can sit there and use it to enhance what you’re doing as opposed to becoming a full-time developer. If you find a process that’s repeatable and consistent, and you do it more than once, you will be able to apply that to making whatever you’re doing more efficient, and that’s sort of where programming people can just use it to solve problems. You don’t hire people who know a language, you hire people who know how to solve a problem.
Now, that’s a different equation for United Healthcare and Boeing. If you’re coding something that runs airplane engines, I care about the efficiency of the algorithm, it being perfect, and an engineering mindset to do it. But in most IT, that is not the case.
Roxanne: Any further thoughts or insights you’d like to share? Anything exciting coming down the pipeline for The Penny Hoarder or for yourself?
Stephen: We’re trying to figure out how we can apply machine learning to our systems, how to optimize our user experience, how to be very efficient with the software development we put in. Python, bringing that up, I’d like to make sure that we have a language which is very readily available supply-wise in the schools around here. I don’t know what South Florida does, but a lot of schools are going to Python. I want to make sure that we have languages which are in supply in the market. There’s nothing worse than getting stuck with having to support a COBOL legacy system.
As for The Penny Hoarder: we recently published “The American Nightmare.” It’s a look back on the 2008 financial crisis. On the 10th anniversary of the national housing crisis, four families told us about losing their homes and how they’re still struggling to recover. “The American Nightmare” explores what the 2008 financial crisis looked like for the people who lived through it and how it’s changed the way they define the American dream. It was a cross-departmental effort. Our writers, Desiree Stennett and Lisa Rowan, wrote the story. Our photographer, Tina Russell, took the pictures. Our design team created the webpage and my team developed the webpage linked above. Finally, our video team published a documentary based on the stories.
You can check out Part #1 of the documentary here: https://www.youtube.com/watch?v=iKPG_l1P7lk
About The Penny Hoarder:
The Penny Hoarder is one of the largest personal finance websites. They help millions of readers worldwide earn and save money by sharing unique job opportunities, personal stories, freebies and more. In 2017, Inc. 5000 ranked The Penny Hoarder as the No. 1 fastest-growing private media company in the US for the second consecutive year, and 25th on the overall list of the fastest-growing private companies in America.
Want to nominate a tech leader for us to interview? Fill the form below!