Data Brew Season 3 Episode 2: Data Culture Outside ‘The Valley’

For our third season, we focus on how leaders use data for change. Whether it’s building data teams or using data as a constructive catalyst, we interview subject matter experts from industry to dive deeper into these topics.

Have you ever had a spam call automatically blocked for you? You can thank First Orion for that - in one day they blocked or scam tagged over 108 million calls - just on T-Mobile alone! In this episode, we have the pleasure to chat with Charles Morgan and Kent Welch, CEO and CDO, respectively, of First Orion to discuss Arkansan data culture, First Orion’s one hundred day program, and team culture.

See more at databricks.com/data-brew

Transcript

Denny Lee (00:06):

Welcome to Data Brew by Databricks with Denny and Brooke. The series allows us to explore various topics in the data and AI community. Whether we're talking about data engineering or data science, we will interview subject matter experts to dive deeper into these topics. And while we're at it, we're going to enjoy our morning or afternoon brew. My name is Denny Lee. I'm a developer advocate at Databricks and one half of Data Brew.

Brooke Wenig (00:29):

And hello everyone. My name is Brooke Wenig, Machine Learning Practice Lead at Databricks and the other half of Data Brew. For this episode, I am thrilled to introduce Charles Morgan, CEO and Chairman of First Orion, and also a very prolific author, as well as Kent Welch, Chief Data Officer of First Orion. So to kick it off for today, you two have a wealth of experience within the realm of big data. Let's start off with, how did you get into the field of big data? And perhaps Charles, let's start with you.

Charles Morgan (00:55):

Well, Brooke, it started, believe it or not, in the middle 1970s. What is that, 45 years ago? I'm quite sure I'm the oldest subject you've had on this podcast, but I'm also probably one of the biggest geeks you've ever had for a CEO on a podcast. I'm a mechanical engineer, and I went to work for IBM out of engineering school and went to work for a small firm after I left IBM in a few years. I had the chance to get into a new field... We were in political direct mail and we were also in commercial direct mail. It took us into managing and doing a lot of promotional stuff on mailing lists, and we grew into a bit of a national reputation. The first what I would call 'big data project' I worked on was, we had a national client that wanted us to match all the voter registration in the United States to all the phone books in the United States... Phone data, which was already digitized to some extent.

Charles Morgan (02:14):

And we started dealing with the problems of big data. Those big data problems morphed into... By the early 80s, I got involved with, and our company got involved with, Citibank in the very early days of credit card marketing. That really grew into the first big data instances I think are built in the world. I'm not going to go on and on with this story, but early days, everything was tape and very large databases on flat-files. I've never fallen out of love of flat-files because they process a lot of data. I look at Parquet and Databricks and how it handles stuff, I said, "I understand that's cool shit." So anyhow, I digress. I digress a little bit.

Charles Morgan (03:13):

But we built some of the first large relational databases on DEC Alpha in the early 90. And I mean, really big databases; Tried to take all the consumers in the United States, all the five years of credit history in the United States, and all of Citibank's transactions that they have over the last several years for all their customers and merge all that into one database. And obviously that would work a lot better if it could be relational in those days.

Charles Morgan (03:52):

The problem was, I met Larry Ellison around 1995, and it was at a function and I introduced myself. I said, "Larry, I'm Charles Morgan from, and then it was Axiom." And I said, "You don't know me." He said, "Oh, I know who the hell you are." He said, "You're the guy that keeps breaking our shit cause your database are too big." So, anyhow and Kent Welch... Then we tried to solve some problems that deal with that and that's where Kent and I came together and we got married on some buildings of big data solutions at Axiom. I was the CEO, but I was still doing the head of R&D a lot. Anyhow, that's way more than you wanted to know. That's 25 years... Whatever, but I am still, Kent will tell you, I'm in their shit on all their design stuff.

Kent Welch (04:51):

Absolutely. Absolutely.

Brooke Wenig (04:54):

That is one heck of a story and I think that's a great segue into Kent. Could you introduce your introduction to big data?

Kent Welch (05:01):

I came out of college as a young pup working at Axiom with Charles. Charles didn't know me then, but I was one of the little guys and just doing some data processing, but pretty quickly I became a Solution Architect in the realm that we had and started designing some of these databases that Charles was just mentioning. I spent a lot of time in the financial services sector, but we had lots of other sectors. And so I kind of bounced around through different industries, creating these big databases. And so that's, that's kind of how I started and then became a business data expert in the company.

Kent Welch (05:40):

And as Charles started talking about, we had a product that we were trying to build and make our processing be faster and scale better than the way that we'd had it before. Technology we'd called Abilitech. And so I came over to help design and build, and then implement that product. And so we did a lot of design work together. Charles was my number one data scientist. So I had the data scientists team, but Charles has always been my top data scientist. It's kind of hard to wrangle and then stay focused, but he's pretty sharp. So we can usually get good work out of him when when we give him a little bit of data to play with.

Charles Morgan (06:21):

Yeah, I still build prototype stuff.

Denny Lee (06:26):

I have to naturally ask, you're busy being a geek, what inspired you to... And this is back to you, Charles, what inspired you and also how do you have enough time to go ahead and create First Orion, for that matter?

Charles Morgan (06:42):

Well, I got involved at First Orion. Axiom, when I left it, I was CEO and Chairman and had been for 30 something years, 35 years. And when we had a billion and a half in revenue and about 7,000 employees. A CEO of a company that size is herding cats and not doing the fun stuff. So, I saw an idea that was a precursor to First Orion and that was handling call management solutions inside the network of Wireline. And that was in about 2008, about the time I was leaving. And I actually was a founding investor at First Orion. I really did not get involved in the business on what I'd call a daily basis until 2012. And then by 2013, I had so much... I was the primary funder of the business. I had so much money in it, I took over as CEO.

Charles Morgan (07:56):

So I've been the CEO since 2013 and, amazingly enough, we're a startup that's made money virtually every quarter since 2013. So, we're 300 employees strong. I think we said we were 302 on Monday and we've hired 90 people this year. So we're a pretty fast growing startup. And what we're doing is all about data management and doing some very large scale data management solutions again. And that's what I like to do. And that is fun for me. And even though he's not on this podcast, it was Jamelle Brown who brought this whole concept, Databricks, to me. I hadn't spent enough time to really understand it, but it sounded like great promise. And Jamelle turned that great promise into, what I would say, what will end up being a transformational technology for us because we have got an absolutely ridiculous amount of technology to build in a very, very complex solution space.

Charles Morgan (09:16):

And I really think Databricks is going to be a tool that will allow us to do it better, faster, cheaper, and have a more reliable end result. I have studied enough about Databricks, Data Lakes, Delta Lakes, Lakehouse, and all those fucked-up names you all have for your product. But I will tell you the whole Parquet structure and I shouldn't... You won't have any questions to ask because I'll tell you everything I know. The whole Parquet structure is incredibly elegant and the way you all are optimizing it with Spark jobs and the fact that you can do... I mean, you've taken Spark from being a really great tool to being a much more elegant tool that you can fine tune, organize so you could manage data in, in this environment. Large amounts of data.

Charles Morgan (10:27):

We've saved a lot of money. We were spending $430,000 a month in AWS a little over a year ago and we have implemented Databricks and we're now spending about 380 last month. So we've virtually 50% more work and we're spending 10% less. So it's already been a big help to us. We probably got 40 people that got me doing Databricks. 30? 40? Something like that.

Kent Welch (11:05):

Probably 40.

Charles Morgan (11:06):

Is that right? Jamelle?

Kent Welch (11:07):

Yeah, Jamelle, that'd be right?

Charles Morgan (11:08):

Yeah. He said yeah.

Brooke Wenig (11:10):

Well, we certainly love all of the praise for Databricks in the Databricks product, but I have a question for Kent about transformations. So you're a Chief Data Officer and not many companies currently have Chief Data Officers. Could you educate everybody about what is a Chief Data Officer and perhaps something that they don't know about the role of a Chief Data Officer?

Kent Welch (11:29):

Well, so that's an interesting question. If you actually look at the roles out there, if you do searches for a Chief Data Officer in the various companies, it's all over the place, right? There's lots of different types of Chief Data Officers. The backgrounds of the people in the jobs are very diverse. And so it's a little bit different based on what each company needs, is what I would say. The job role itself is hard to nail down. What's something you wouldn't know about a Chief Data Officer? I don't know if I could actually pick out certain things. I know in our area, we're a little bit more product-focused. My group is more product-focused, so we're building a lot of new product as Charles talked about, using this technology specifically for fighting scam in the world of voice calls.

Kent Welch (12:28):

And that's really our big focus, so we're big product focus group. But interestingly enough, as we get bigger and bigger, a lot of the data science that we are looking at, it's turning internal to some of the data that we need to run our business, keep our systems going. And so, it's a little bit diverse on the job role.

Charles Morgan (12:51):

But all our large data assets, in effect, your group manages.

Kent Welch (12:56):

They do.

Charles Morgan (12:56):

I mean, that's really what it's all about. From my perspective, somebody's got to own your data assets at a high level. And there are obviously all these people around the edge. Now we have Chief Security Officer, we have Chief Privacy Officer, but remember, we've got a lot of consumer phone numbers. We've got all the phone numbers in the United States and some in the world. So we've got consumer data and we have call records of a lot of phone calls. So we're very privacy conscious. We're very data security conscious. How we manage that data and keep it secure, it's all... If we have a really big screw up, all I have to do is just call Kent and say you screwed up.

Kent Welch (13:43):

Right. That didn't happen very often.

Charles Morgan (13:46):

Now, no, not really.

Brooke Wenig (13:48):

That definitely makes sense that the role of Chief Data Officer is very diverse. Because if you look at data scientists, they all have very diverse backgrounds. Getting into that field as well, it's hard to pin down what exactly is a data scientist. Makes sense that a chief data officer, same problem there. But now I'd like to segue a little bit, since you talked about spam detection. How exactly do you keep track and build models to detect spam given that the nature of spam changes over time?

Kent Welch (14:12):

Oh yeah, that's a good one. The spam scam business is a big business. These guys are making billions of dollars just in the United States off of U.S. citizens. And it's an arms race type deal, is the way I would describe it. We're focused on bringing all of our data assets in to see... As Charles talked about, we see billions of call messages that we're trying to dig in and determine what's good traffic and bad traffic. It's a big, big data problem.

Charles Morgan (14:55):

Believe it or not, almost 30% of all the traffic on the public switch telephone network that is intra-carrier, between carriers is scam traffic. 30%. It's crazy.

Kent Welch (15:16):

Yeah.

Denny Lee (15:16):

Wow. That's a pretty insane number.

Charles Morgan (15:20):

What's our biggest call block day, 108 million calls in one day we blocked or scam tagged? 108 million calls, just on T-Mobile.

Denny Lee (15:29):

Whoa. Okay. So I think that naturally segues. So then how often are your... How do you build processes to actually handle the, basically what you're just calling out, an arms race here? Are you constantly updating your feature engineering? Are you constantly updating your data processing, your machine learning models? I'm just curious. It seems like there's a lot to go here.

Kent Welch (15:51):

There is. Right. And so we have to look at the problem from multiple directions. And so we have different types of models and processes that we put in place. And the scammers are going down a path and we figure out how to stop them. And they work hard to try and figure out how to get around that. And so we have a lot of different types of models that are constantly being updated and changed. So some of the data that we have goes in and updates on a daily basis and some of them on an hourly basis and some things 10 times an hour, sometimes near real time, right? We're having the machine learning components that are sending-

Charles Morgan (16:34):

Half the calls that we tag or identify are in, what I'll call, a delayed offline process. Half the calls we identify are real time, looking at the actual call, sent message itself inside the network, as it hits the edges of the network. So we're looking at call coming in and we look at it and we do what we call call-print it. We look at the invite, sent message, and there's a lot of data in there and we analyze that, but we also analyze data offline in our Athena AWS database. And we look for different patterns, which may be high-volume scammers that are spoofing, not spoofing, but mostly our offline stuff is the non-spoof stuff.

Denny Lee (17:37):

Right. Well then can you describe to me a little bit about how the spam texts work? What type of models, what inherently is in the stuff that basically you're tracking? Is it just the phone number? Is it the text itself? It's the metadata? I'm just curious what?

Charles Morgan (17:55):

Well, no, it's no text because remember these are phone numbers only now where-

Kent Welch (17:59):

Phone calls.

Charles Morgan (18:00):

Phone calls only.

Kent Welch (18:01):

Voice calls.

Charles Morgan (18:01):

We're looking mostly at the detailed data and there are a bunch of fields that are in there. For example, it says in the phone record, it'll say here's the number that made this call. Well, that may be spoofed, but it will also have the type of device that made that call, certain routing data about the call, here's where the call came from. We also have, from some of the interconnect carriers, here's where they got the call. There's all that kind of stuff. And there are a myriad of other parameters that we look at and very obscure stuff... The only way in the world we could tell you, we'd have to kill you.

Kent Welch (18:58):

But there's some time series stuff that we do, as well, that we're looking at it. And one of the important things in our processes, we're not just looking for the bad calls, right? We're looking for the good calls. And so we're really trying to understand call patterns for numbers and for traffic routes and digging into the details of that. And then you look for anomalies, right? So you're doing a lot of anomaly detection across those good and bad calls. And so, odd things happen in the telecom world where phone calls start getting routed new directions because they have what's called least-cost routing. It's cheaper to go this way. And it changes the way that the traffic goes. And so you got to be able to understand that, even though it changed a pattern, it's still good, right? So it's just a constant pattern anomaly detection process that we're dealing with.

Charles Morgan (19:56):

But I just...So y'all have to understand that, for the most part, we've got a system that works for this, and we are looking at picking up some other carriers, but almost all of our growth and all the stuff we're going to be using Databricks for in the future, cause we already converted a lot of this to Databricks. All of the future growth is going to be in our branded calling solution. We are working to bring every handset in the United States, into our exchange. And so that every phone call you get from every business will say who it is calling, who's calling you, and why they're calling. And that's partially enabled by an expanded caller ID field. And in the future, will be delivered by RCD, which is a rich data content. And that a solution will bring all the carriers together into one centralized platform.

Charles Morgan (21:00):

And we'll be managing policy. We'll be managing billing. We're going to be managing the actual distribution of the content to each of the carriers. And we'll also be collecting all the data from all those phone calls back into an analytics databases that will be displayed or available to the customer. This is not for marketing. None of these calls can be made unless they have a purpose for making them that is a permissible purpose. You're not going to get any unsolicited phone calls from this system or you're not going to get spam calls because they're going to be secure calls. It will know it's not a spoof call or an inappropriate call, so we also have that.

Charles Morgan (21:55):

That merges all our scam stuff back into this solution, too. It's stir-shaken. You don't know what that is, but there's some industry standards. RCD's an industry standard. We're trying to be the first. We've actually demonstrated our ability to produce an RCD call working with T-Mobile. There was press release that went out about a week ago, I guess? Maybe look it up? T-Mobile's RCD, first in the industry to deliver and that was us.

Brooke Wenig (22:30):

Well, all of us will be thanking you for stopping these spam calls from coming through. We definitely have to thank all of you for that. I do want to ask you a little bit more about the culture at First Orion, because I heard that you have a pretty interesting hundred day program. Could you talk a little bit more about your first hundred days at First Orion?

Charles Morgan (22:48):

We have the global expert on hundred day projects sitting right here with me.

Kent Welch (22:54):

Well, so our a hundred day project concept actually started back at Axiom when we started this Abilitech project, and the idea is we had a concept that we needed to be first to market, and how do we get there? Right? So you have to put your blinders on and focus on an MVP product. How do you get the basic out as fast as you can, but it has to be built on something that you can't throw it away, right? You have to be able to build on it. And so the mindset has been pull your top talent in, get them focused on a very finite set of things that you're going to deliver, and put up a really tight timeline on that doesn't allow you to waiver and go off track, right?

Kent Welch (23:41):

And we've, and we focus on doing that. And so we've done that a couple of times. We did that at Axiom when I first came over here to First Orion with Charles, and actually we hired Jamelle. You heard him talk about Jamelle. We hired Jamelle Brown. We actually created the first hundred day project at First Orion then, to build what we call Athena, which is our AWS environment that has all of our knowledge, all of our models on the first scam detection that we did. And now what we have done is we've just launched a hundred day project, I guess we're halfway through now, a hundred day project on the technology that Charles just talked about. So we're going hard to get that full solution in place within a hundred day period. And then you can build on top of that and you can start adding all kind of additional functionality, but the core of what you've got is there and ready to roll.

Charles Morgan (24:35):

We're delivering these branded calls right now, about 30 million a month. So it's not something that's going to happen in the future, but we're trying to build a much more robust multi-carrier platform that can handle not 30 million a month, but three billion a month.

Denny Lee (24:55):

Well, this is super interesting. I love this concept of the hundred day type of project. But we basically are talking about culture and I'm just curious, especially considering everything you guys have gone through. What's the difference of the tech culture in Arkansas versus Silicon Valley? I'm just curious.

Charles Morgan (25:15):

Well, sometimes it's very, very different and other times it's similar. I think every company has got to build their own culture. When a lot of people start building a business, the focus is on getting shit done. Just get stuff done. Hire good people and get stuff done. Sometimes when you can't hire good people, you just have to hire people to get stuff done. So our culture focus here has been to make it very integral with the way we work. Our culture is the way we work. I don't know a lot of ways we're all that different from some of the West Coast cultures. We don't have free lunches every day, we have free lunch on Tuesday. On Thursday, we have beer 30, where everybody can come up to this floor, we're on our fifth floor of our building, and we come up here and we have free beer and no telling what else.

Charles Morgan (26:19):

We may have Mexican food, Chinese food, or a combination of Mexican and Chinese food. You never know. And beer and some wine generally. And we try to create a team-based culture and a camaraderie. It's hard. It's hard, everybody at home, but we've got everybody back to work now, pretty well. Even though Arkansas is a real hotspot, but you know, you can't have everybody back to work unless they're vaccinated. So we've told everybody, you can come to the office, got to be vaccinated. We're not having anybody to offices not vaccinated, and we're not having any issue. I'm knocking on wood really loud right now, but we're not having any issue here because people are vaccinated.

Charles Morgan (27:11):

And I really encouraged that. We do a lot of communication. We have a Monday morning meeting where we have 45 minutes where I talk, all the leaders talk, our leaders have a separate leaders meeting right after that. And we do a lot of surveys, feedback from our people, finding out what they think of their environment. We do great places to work every year. We have a culture that's founded on people first. People first.

Charles Morgan (27:54):

There are four pillars. The upper left hand corner is people first. Guess what the lower right hand corner is? Innovation. And clearly we want an innovative company. We want one that considers people. If you're going to be considering your people as one of your most important assets, then you have to exhibit characteristics like trust and transparency.

Charles Morgan (28:27):

And so we don't have a complex, what I'll call a complex culture, but culture is all about living. And every day we have fun and have it be a fun culture, but it's a culture where everybody's responsible for getting the job done. We are very, very clear that we don't have vacation policy. We don't have sick leave policy, but you better get your ass to work and get shit done, or you're fired.

Kent Welch (29:04):

Yeah.

Charles Morgan (29:04):

And I make that direct.

Kent Welch (29:06):

Yeah.

Charles Morgan (29:07):

You know, you cannot be a member of this community and this culture and have all your work done by your team members. You've got to do your part. Not everybody has to be brilliant, but everybody has to do their part to the best of their ability. And we also try to do a lot of training, cornerstone of our training. I know y'all are interested in that it, and the way we get our... We don't have an enormously large pool of data scientists, but we have apprenticeship programs, which are... Is that a six, 14, how many week program is your apprenticeship program?

Kent Welch (29:50):

Our program for this is about three months.

Charles Morgan (29:53):

Yeah. So, yeah, so we have an apprenticeship program, which is a school all day. It's not work part a day, it's school all day. You're paid. You're hired and paid to the job you were hired into, but you're in a probationary period till you get through the three month apprenticeship program. And the apprenticeship program is a combination of, I say off the shelf education, but we have an education staff of three or four people. And if it's an apprenticeship program in Kent's area, he's going to have a lot of the people actually responsible for that program in the classroom, hands-on. We do a lot of Capstone Project and stuff. And so we're sure when they get out of this program, they're ready to go sit down and help build a model.

Kent Welch (30:59):

Exactly.

Brooke Wenig (31:00):

I love this whole discussion we've just had on culture, especially this... Well, one the beer and Mexican and Chinese food, but two, having this program to help train data scientists, because there's a global shortage of data scientists.

Charles Morgan (31:13):

Yeah, we have 42 openings right now if either one of y'all are interested?

Brooke Wenig (31:24):

Thank you for that callout. We were going to do that at the end, but thank you for putting that one in there, Charles, that you're all hiring. And so just as a way to wrap up the conversation we've had today, I want to leave all of our listeners with a bit of advice. And so I want to ask each of you, Charles, what advice do you have for people that aspire to be a CEO? And for Kent, what advice do you have for people that aspire to be a CDO?

Charles Morgan (31:46):

Well, I think we actually talked about it. Nobody's going to be successful alone. I know so many businesses where people start off a business with a friend or something and get some investors and it's all about building things, getting something done, and it's all focused on getting sales, getting revenue. There's virtually no focus on the culture and even generally the training or all these kind of things, which are related to it. We got to a point at Axiom that we had a lot of things out of control. Our culture was not managed, it evolved and happened. And even though we hired really good people, we worked hard, customer first, but the culture was not formal. It was informal. And it was all over the board. And we didn't have standards for leadership training or any of those sayings.

Charles Morgan (32:55):

Don't wait too long. As your business evolves, have some idea of where you're going. Don't ignore the concepts of having people-related policies and culture early. Don't just think about trying to get through this next month or two months. If you start getting any level of success, try to transport yourself into the future. From the early days of Axiom, we did planing meetings couple of times a year with the top leadership. We still do that here, and doing a lot of planning, doing a lot of thinking about your future, you can always take off a day a quarter, or a day every six months, and peer into the future and try to organize yourself to be successful. So organize for success, plan for success. You've heard those kinds of terms. Really important when you're a young company. Don't ignore those things until it's too late.

Kent Welch (33:59):

For Chief Data Officer or anybody in the data field, the thing that I would say is always be inquisitive, right? You need to be inquisitive. You need to be looking at, "Hey, what does that mean? Or what caused that?" Right? Thinking about those mindsets, we have a saying we use around here and we've used it for 30 years. "It's the data, stupid," meaning it is literally... Data is the foundation of everything that we're doing. And so you've got to start there and you end there. And a lot of times in some of the things that we've seen happen is projects get away from that. And then inevitably what you always come back as, "Well, what did the data tell you?" And let's go back to that point and then let's see where we go from there. So that'd be the focus I would say for Chief Data Officers.

Charles Morgan (34:47):

Yeah, people that don't get a vaccine shot, I say, "Look at the data, stupid."

Denny Lee (34:54):

Oh, this has been a wonderful session. So thank you very much. We're going to leave it with, "It's the data, stupid," and people first and innovation. This has been wonderfully great advice from Charles and Kent. Any last words for that matter before we close off?

Charles Morgan (35:14):

Hey, I hope you could publish this. If you can't it's my own fucking fault.

Kent Welch (35:25):

Yeah.

Denny Lee (35:27):

Thank you very much for your time. Really appreciate it.

Charles Morgan (35:30):

Thanks guys.

Kent Welch (35:31):

Yeah. Thanks all.

Charles Morgan (35:32):

Y'all were fun.