
Exploring the Future of Data: Regulations & Managing Analytics Teams
Summary
Transcript
Between GDPR, CCPA, and more regulatory frameworks on the horizon, the landscape of personal data—and how it can be used in business—is shifting.
On this episode, John Thompson, Global Head of Advanced Analytics & AI at CSL Behring, joins host Dave Cole to discuss that shift, and a potential future in which we as individuals could be compensated for the use of our data.
Plus, John shares the two types of analytics teams he’s seen work well during his career as a data science leader.
Topics covered include:
- If and how individuals can know what companies are doing with their data
- How GDPR and CCPA portend the future of data
- Structuring, growing, and managing different styles of analytics teams
Check out these resources mentioned during the show:
DAVE COLE
Hello, welcome to another episode of the Data Science Leaders podcast. I'm your host, Dave Cole. Today's guest is a good friend of mine, John Thompson. John, welcome to the podcast!
JOHN THOMPSON
Hey, Dave. How are you doing today?
DAVE COLE
Great. I'm doing great. I'm looking forward to talking about the future of data with John, but before we dive into that, I want to give a brief introduction to John. John is a bestselling author of two books. His first book, let me make sure I get this right, John, is “Analytics: How to Win With Intelligence.” His second book is “Building Analytics Teams.” So I have read your second book but, sadly, admittedly, I've not read your first. It's on my backlog of things to read. I also hear that you have another book coming out soon, too. Can you tease that book? I think we'll be talking a bit about that today.
JOHN THOMPSON
Yeah, absolutely. Thanks Dave. It's so great to be here with you. I always enjoy our conversations and our time together. The third book is called “The Future of Data.” I've just started writing it so I'm very excited. The second book, “Building Analytics Teams,” I wrote in the span of three months, a hundred thousand words in three months; I cranked that out. This book, I'm taking a little bit more time on. I'm spending time looking at what the EU regulations are. Some of the new proposed regulations on AI, the California Privacy Act, what has been done with data since online properties have come around.
I've written two nerd books, that’s what I've done. What I want to do is I want to write a book for general society. When I talk to people, they don't really understand what's going on with their data and they're absolutely freaked out or concerned or fearful. I want them to understand a little bit about what's happened with data, where we are today and where we're going. I really want everybody to be empowered about their data and have a voice and go to the government. Rise up, people. Let's do something about what's happening with our data. Let's take back the power of/about our data. Let's get a data dividend. Let's get paid for our data. I'm fired up, I'm ready to go. I want this to be a movement!
DAVE COLE
This is going to be interesting. I definitely want to delve into this because I have also worked for a third-party data provider and I have a lot of experience in that world and understand a bit about it. Probably not as much as you because you're writing a whole book on it. I think there are a lot of misconceptions out there about how their data is being used. I also think there's a lot of things to certainly be aware of and we'll dive into that.
But writing books is just what you do for fun. You do have a job and you are the Global Head of Advanced Analytics and AI for CSL Behring, and have been for almost three years now. So you're also a data science leader and you just write books.
So let's dive in. Clearly you're concerned about how data is being collected, probably unknowingly, about the general populace. Your book sounds like it's going to be a clarion call for all of us out here to wake up to the data that's being collected about us. Help us unpack that. Let's start there.
JOHN THOMPSON
Yeah, absolutely. Thanks for the opportunity and the opening to talk about it. Our children are now, my and Jennifer’s kids, are 23 and 21. We used to sit at the dinner table and they'd be talking about this game that they played with these little puffles. And they were giving the people in the game their name, their address, their interests, and all this kind of stuff.
So we would talk about that and I would say, “Okay, you're giving them your information. What are you getting back?” And they're like, “Oh, we get to play the game. It's free. It's just our name and our age and stuff like that. We're not giving them anything. We're getting this game.”
And we would have conversations and say, “No, that's not true. You're giving them access to who you are so they can sell you things in the game.” Then they're like, “Oh, what are you talking about?” And I would say, “Well, let's sit down and look at the game together.”
You would see how the game would slightly change based on what information they gave them: he likes football and she likes soccer. The game slightly changed based on how they interacted with it. Over their lifetime, we educated them: you’re not giving something away for free. You're actually giving something of value to them and they're monetizing it in different ways with advertising, product placements, and those kinds of things. Then the kids started coming back at dinner and saying, “Well, that's not really cool. They're not telling us how they're using our data and what kind of trade we're making. They're just putting this up in front of us and we're doing it.”
We were having a conversation the other day. I asked people, “How do you feel about the fact that you've told Amazon you live in this house, you have a second house, what your mortgage is, the fact that your Aunt Matilda has just developed cancer, that you ride a motorcycle, that you have a child that has special needs. How do you feel about giving all that information over to Amazon and allowing them to monetize it?” They responded, “I don't do that.”
“Really? They know where you live. They know all your purchases. They know how you like to get things delivered. They know you're sending things to Aunt Matilda. They know all these things and you're freely giving it to them.” Then they respond, “Wow. I had no idea.”
DAVE COLE
Well, I would say, there is a distinction between Amazon inferring things about you, versus a company like Facebook, where you are freely posting things on their platform that says, “Hey, my Aunt Matilda just got diagnosed with cancer.” It’s a little more definitive than inferring. Where do you draw the line then? If you're an analytics leader, and you're making certain inferences based on purchase behavior of your customers, are you saying that it's unethical to come to certain conclusions and then market to that person based on those conclusions?
JOHN THOMPSON
No, not at all. I'm not saying that at all. The analogy or metaphor, I guess it's an analogy that I like to use is that, how many builders in the United States get all the lumber to build the houses for free? None.
Google, Facebook, all these people get this data for free. Now they would say it's a fair trade, “We're giving you Gmail,” or “we're giving you this,” or “we're giving you that,” for your data. I don't think that's a fair trade. I think that if you're using my data to build your business then you should give me a chance to get paid for it. I think that the data that those companies ingest should be an expense to them. They should pay everybody for the use of that data.
No other company gets their raw materials for free; none that I know of anyway. That's my point. I'm happy to use the data. I'm happy to be ethical and happy to take transactional data and do things with it as long as I'm helping my customers, my donors, my patients have a better experience with my company or to have less costs incurred. My point is that if you're using it as a raw material in your business, you should have to pay for it. There have been some people that have done analytics and they're like, “Oh, well each person is going to get $52.” That $52, is it going to make a difference to them? That's not the point. The company should have to bear that expense is what I'm saying.
DAVE COLE
It sounds like you're almost advocating for a model that says, “Look, if I give you my data, you should have to pay me in return.” What happens if I say, “I don't want to give you my data, but I still want to use Gmail for free,” as an example?
If I'm Google, I would say, “Look, if I'm not paying you for your data, then why should you get Gmail for free?”
JOHN THOMPSON
You shouldn't.
DAVE COLE
So then we're moving to a model that's very different. I mean, obviously there's the free content model and the free service model is pervasive, like on Facebook.
JOHN THOMPSON
My point is that it should be optional. You should go to Google and the first thing, when you're signing up for Gmail, should be a toggle button. I'll give you my data for free. Gmail is free for you. I won't give you my data for free. Okay. You pay $52 a year for Gmail.
DAVE COLE
Okay. All right. I think that's fair. I mean, at least have that option.
JOHN THOMPSON
I think so.
DAVE COLE
Yeah, choices. Now, Google is probably one of the more above-board companies out there in terms of collecting data. There are a lot of websites out there that you can happen upon that won't be nearly as ethical, in terms of the data. Some of those fly-by-night game companies that pop up. Now I know that those companies are not just collecting the data and using it for themselves. They're also repackaging it, selling it down the river too.
JOHN THOMPSON
Yeah, absolutely. There's no doubt about it. The whole data syndication ecosystem, that you and I both know well because we have worked in those organizations and know what's happening there, that whole industry is kind of like we sold our souls back with Arthur C. Nielsen when he was walking around counting each box of Tide on the shelf or something like that. It all started way back there, that the grocery stores in America, chains and grocers, or the people that committed the original sin in this industry, because they gave all that information to Arthur C. Nielsen, with all the rights to own it and resell it and repackage it. That's the basis of Equifax, Epsilon, Google, Facebook, Netflix and all these companies. It all comes from that origination of that business model.
At one point it was probably a fair model. I did a ton of work through the ‘80s and ‘90s in the retail, CPG, and FMCG market. I was always asking the retailers, “Can't you just stop doing this? Stop giving your data to Nielsen, stop giving your data to IRI? The amount of money you get is a pittance.” And the only people that have ever really done it was Walmart.
If you remember back in the ‘90s, I think it was, Walmart just said, “Forget it. We're not giving you their data.” IRI and Nielsen both freaked out because of how much of the US market is that retail channel. So it can happen. It has happened. If people actually demand it, it will happen again.
DAVE COLE
When I think of Nielsen, I think of the television ratings, the Nielsen ratings. In my recollection, I thought the Nielsen families that were used as the basis for determining what programs people were watching, those deals and families did get paid in some form to get the data.
JOHN THOMPSON
That's the media measurement side of the business. Yes, I think Arthur and his children did realize that they hoodwinked the grocers. Hoodwinking Joe and Sally, probably not very smart, and the people wouldn't do it either. They tried to have people fill out the diaries, because that's what he did.
DAVE COLE
That's where it all started. It was written down. Now it's more automated.
JOHN THOMPSON
And they wouldn't do it, so he had to pay them. But on the grocery retail side, that's still the business.
DAVE COLE
I do want to talk about GDPR. I do want to talk about the CCPA, the California Consumer Privacy Act, and how that impacts us as data science leaders. One thing I've always said is this: I'm going to get ads. That's a fact. If I'm going to a content site, ESPN, watching programming for free—I don't know that people don't have over the air TV anymore, when it was free; now it's all cable and you cut the cord etc.—then I’ve got to suffer through some ads. If you're going to get those ads, wouldn't one prefer to get ads that you might actually be interested in rather than ads that have no relevance to you whatsoever?
JOHN THOMPSON
Yeah. I would suppose that people would make that trade off. Personally, I know I'm not normal. I don't mean that in a derogatory way. I'm not. Every time I take a personality assessment, I'm in 1% of the population. I'm just not normal. In the Gaussian distribution, I'm always out in the tail somewhere. Everything in my life: I pay for everything I possibly can to never get an ad served to me. I never watch live TV. I never listen to terrestrial radio. Any channel there is in the world where there's an ad, I assiduously avoid, so maybe I'm not the right person to ask.
DAVE COLE
But you still see the ads when you go to any website where you're reading content right now.
JOHN THOMPSON
I never look at them.
DAVE COLE
That's the thing though. You might not look at them but it might be the corner of your eye and it might have some impact on you. There might be some incremental lift if you had not actually seen it at the corner of your eye.
But at any rate, I hear what you're saying, which is, “I do not want to participate in the whole advertising business. I'm there for the content. I'm there to view the show. If I have to suffer through some ads, I will block it out of my mind or what have you.” I don't know that you're alone in that thinking, that ads are a little bit of a nuisance. You talk to folks in the UK where you don't have to deal with as many ads. There isn’t that culture of ads, like we have in the US, and they think we're crazy. Maybe that's where you should move to, John.
JOHN THOMPSON
I did live in London for a few years.
DAVE COLE
So there we go. Still, if you walk around London, you see the billboards, ads all around the sides of buses and things like that. It's not like they don't have an advertising business. Would you suggest to a retailer to get your word out and to sell John Thompson something that he doesn't know he needs until he sees this ad.
JOHN THOMPSON
Wow. That's a great question because it would be really hard to get to me, to tell you the truth. I don't really know how you could get to me. I'm trying to think of where you would be able to do it. I don't look at anything on my phone. I don't look at anything online. I don't watch anything on satellite television. I guess physical direct mail, which is kind of weird. I do get some really interesting high-end expensive packages from people, and I open every one of them.
DAVE COLE
I went on a bit of a tangent on John's advertising peccadillo, there, but I think there's a lot of truth in it. If you were to step back and understand: your data is being used to better target you and to hopefully give a retailer some incremental lift, but there's a portion of the population that sees ads as a big nuisance, and you're trying to balance these things. You're trying not to frustrate, because you can absolutely inundate. You have retargeting, where you search on something, like a football for your kid, and then all of a sudden, all you see is football ads and sports equipment ads following you all over the place, and that's creepy. I don't know if your book talks about how retargeting works.
JOHN THOMPSON
No. Well, it might. I've only written the first chapter, so I may talk about that in “The Future of Data.” I just want people to be aware of what's going on. I don't want them to have this niggling fear that something's happening to them, that they don't understand.
Now, I'm cool if people go in and say, “Hey, I want the free services.” I talked to someone the other day, I guess it was about a month ago, who's a very well-respected analytics leader. I put this forth and he was allergic to the concept. He was like, “What are you talking about? Google's great. They can have all my data. Facebook is awesome. I love to get all these free services. You're off your tree and it will never change.”
And I'm like, “I'm not telling people to change. I want them to make informed choices.”
DAVE COLE
Yeah. You want them to have the choice. They don't have a choice today.
JOHN THOMPSON
Exactly. If they're into it and that's what they want, then cool. If you understand it and you know what you're getting into, then you're going in there eyes wide open. I don't want people to go into it and give things away that are valuable, that they should get money for, or some compensation or consideration, at least.
DAVE COLE
There are many data science leaders out there who I'm sure are purchasing third-party data to use, to augment their models. What is your perspective on that? I mean, because from the third-party data perspective, from data sets that are collected, many of them would tell you that they were collecting it in an ethical way and they will explain how it is being collected in ethical way and that they are only doing it with the people's willingness and sort of opt-in, or maybe opt-out.
We could talk about what is ethical in terms of opting in versus opting out. That's a whole other conversation but would you, as a data science leader, be okay using these data sets or would you do your background research to make sure that they were collected in an ethical way? Where do you see that impacting you as a data science leader?
JOHN THOMPSON
Yeah. As a data science leader, yes, we use those data sets and we're doing some really intriguing things with the data set right now that's probably going to give us a competitive advantage in our industry that no one else has even thought of. When I exposed it to some of our leadership they were just mind blown. They're like, “Wow, you're kidding. Can we do that?” We can, absolutely. Now, we don't go out there and dig through every agreement. People say that they have informed consent from all the users. Wow. You probably have consent. I think ‘informed’ is probably a stretch.
I know that we have to do a better job on the click agreements and being able to simply and easily tell people what's going on with their data. I'm not such a zealot that I won't use the data, that I'm absolutely pure and pristine in my approach. We do use data, we are just assiduously careful about PII and PHI. We're a pharmaceutical company. We can't go stomping all over the place and be careless in that area. So we're very careful there, but there is lots of data. I think there can be lots of improvement in the industry but, yes, we do use data. Absolutely.
DAVE COLE
You mentioned PII. Let's talk a little bit about PII, GDPR. I don't remember it exactly when it sort of went into practice.
JOHN THOMPSON
15 years ago now.
DAVE COLE
Really? Okay. That's longer than I thought. Most of our audience, if not all, should know the basics of GDPR, but maybe not CCPA, unless you live in California. Maybe you can briefly give us an overview of both.
JOHN THOMPSON
Yeah. Well Dave and I have been around long enough that we understand these things. I've been doing this for 37 years. I remember when GDPR (General Data Protection Regulation) was being talked about. Everybody was saying, “Oh, no one's going to do business in Europe. No one will have a server in Europe. All the data is going to be in the United States and Middle East and Asia, and Europe will be a desert.”
The idea was that you as a consumer have the right to be forgotten. You have the right to go in and look at the data someone has about you. You have the right to delete the data that is inaccurate. These are things that we talk about now that are fairly reasonable but back then it was like, “Oh my gosh, the Europeans are insane! This is just terrible. This can't happen.” But it's really turned out to be a fairly reasonable, fairly decent basis of doing business. It's nothing crazy.
Last February there was a proposal to the EU to create a pan-European data commons, where every organization, retail, manufacturer, consumer: all their data would flow to a central repository. Again, people were all up in arms and, “Oh my God, this is terrible.” And, “Europe is going to fall apart. No one will do business in Europe.” It was the same thing, almost the same conversation. I looked at it and said that this is the future.
Every European is going to get a data dividend and maybe it's a hundred dollars. Maybe it's 72 euros. Who knows how much it is, but you are going to have the opportunity to go in and say—I'm just going to put out some scenarios—“I don't like companies that contribute to climate change.” All right. Well, some people would say, “Check, check. I don't want the airline industry or any airlines to be able to use my data.” That's one way to look at it.
You don't like companies that contributed to climate change. Maybe you don't like cattle farming, maybe you don't like airlines. What you should do is click the button that they can use your data, but they have to pay you 100x more than anyone else. If they want to pay that amount of money, great, you suck more money out of their organization and send it to the places that you support.
This data commons is going to be a way for the European Union to do that and it'll take 15-20 years for this to come to be. It's kind of funny that people that are futurists are saying, “This is madness. This will never happen.” Well, over the long run, it will be the norm. I started talking about this stuff maybe 13, 15 years ago, and people were saying I was nuts.
DAVE COLE
You are nuts. But yeah, but not about this.
JOHN THOMPSON
That is happening. The regulations are happening. The basis of what's going on in the CCPA as well. As we all know—you live in California, you know this, everybody knows this—they are the leaders in the United States, in regulations for emissions and privacy. This will sweep the United States over the next 15 years. This will be the norm. So people, as consumers, should get on board with it because it's great for them.
DAVE COLE
As a consumer, it makes a lot of sense.
JOHN THOMPSON
Why would you fight it?
DAVE COLE
I can't believe I'm standing here, playing the role of the defender of these companies out there. But, as the company, it's certainly going to make business harder. I remember when I was reading over, hearing about GDPR, my first reaction wasn't like, “Oh, this is a crazy and a dumb idea.” My first reaction was like, “Oh, this is going to be a nightmare having to create systems to delete all this PII when a user or a customer calls in saying like, ‘Hey, can you forget about me?’” I'm just like, “Oh, I can't imagine how much work it's going to take to build the systems out, to be able to do that in an efficient way.” But if you want to do business in Europe you have to be able to do that.
I know there are companies out there that specialize in helping make that all happen. At the end of the day, do I want my data to be used in ways that I don't approve of? Of course not. If I could wave a magic wand, of course I wouldn't want that to happen. When you unknowingly consent to these things, you don't realize that you're consenting to the company using your data in all sorts of ways.
JOHN THOMPSON
One thing that I think is going to be a huge market, and you should probably be the chief customer officer for whoever starts this company, is going to be a data lineage company. It's going to be someone like what you were talking about earlier. All this data has been collected through all these different channels and used in different ways. At some point, someone is going to regulate data lineage. If it was collected from John Thompson through this app, then it went to this aggregator and ended up in this company, and then it ended up in this prediction. If John Thompson objects, that model is no longer valid because all that data lineage says that he doesn't want to be part of it and it has to change. That's an obscene and extreme example, but I do believe that whatever company can do extreme data lineage and do it well at scale, they'll be huge.
DAVE COLE
All right. Your first two books were definitely more focused on, like you said, the nerds, the data scientists. I just called all my entire audience nerds. I love you, nerds. I’m a nerd, myself. I mean it in the best possible way, a loving way. This sounds like you're hoping to open people's eyes and show them that there are ideas out there. There are alternatives that you could get behind that would make it easier to give up data in a way that is more transparent, in terms of how it's being used and what you get in return.
JOHN THOMPSON
You're absolutely right, Dave. I want this book to be a learning experience. I've written two how-to books. Now I want to write a book about the fact that this is happening. You should be aware of it. It'll make your life better. Make informed choices and do business with people that are aligned with your values and are using your data in an honest, transparent way.
DAVE COLE
You mentioned data lineage. I'm hardly a blockchain expert, but I have to think that my PII being stored in some way, and then having some transaction log as it moves around, seems like that could be a potential vehicle for making that transparency of where your data is going, a little easier.
JOHN THOMPSON
Yeah, distributed immutable ledgers could definitely be a value-added underpinning technology. There's no doubt about it. Any time you have multi-party transactions, and some component of trust is required, blockchain is definitely a runner and possibility to be in the mix. You're spot on.
DAVE COLE
I want to switch gears. I want to talk a little bit about your book “Building Analytics Teams,” which I have read. One concept of it that I found, that I use to this day: there's one chapter that talks about the artisanal teams and the factory teams. Tell our audience a little bit more about what you meant by that in the book and help us understand the different types of teams that you see out there.
JOHN THOMPSON
Yeah. Thanks for the opportunity to discuss it. It's a concept that's come about, as I noted, doing this for almost four decades. As I said before, many times, I've made every mistake known to man and woman. I realized that data science teams and data scientists are very creative people. They need to be handled and managed in a different way. I was casting around, trying to figure out ways to structure teams, so that we could get resilience, responsibility, accountability, individualism and creativity all in the mix.
I generally think of things as two ends of a pole. The artisan or artisanal team was where you have highly skilled, highly autonomous, highly trained individuals who could do everything. They could acquire data. They could integrate data to get/do feature engineering. They could build models. They could do subject matter expert interfacing. They could talk to executives. They could work with the IT department to get models into production. They're kind of like data science superstars and that kind of team resonates with me because I liked those kinds of people. I like to give them direction, let them run and make mistakes, blow shit up, and come back with some really creative and interesting things.
On the other side of the equation is the more modular factory team. That's where you have a piece of work and something like two people that do data acquisition, two that do feature data integration, two that do feature engineering, some people that do model building and so on. The work moves through that group.
I've been part of different organizations, including seven different startups. I've been at Dell and now at CSL, a $10 billion biopharma company. You can't just hire a team in your image. If I hired a team in my image, that would be the Uber artisan team, that would be like John Galt and all that kind of crazy stuff. There are companies that run more like factories even if they're not a manufacturing plant. You need your team to have a model that integrates well with the broader organization.
You can have a hybrid team as well, which I'm building now. I've got an artisanal team or an artisan based team and they do a great job. They're incredible. Now I've got a bunch of work that's scaling up, that's repetitive, that needs to be done over and over again. I've got some other visionary ideas that I want to go out and collect a ton of information, primary research, and I want to bring that back and integrate it together. Well, I can tell you that the artists and data scientists don't want to do that work. They just sniff at it. They turn up their noses and walk away. They're a bit of prima donnas too. You’ve got to deal with that, so I'm building a hybrid team now. I've got a team that does all that mechanical work, and then I've got an artisan team above it. So I've got both a modular and an artisan team. So it's a great way to do data science, to tell you the truth.
DAVE COLE
I think there are certain types of problems. If you have a model that's been in production for five years, the business problem is well known, you're approving loans or something like that. All you're really adding to the equation is new features and there's new approaches to new algorithms you want to try out, that seems to be more of a factory thing. If there's a business problem and you don't know how to apply data science to it, that's really where I would imagine the artisanal team shines, where their creative juices flow and they're whiteboarding, riffing and putting together POVs, and things of that nature. Do I have that right? Is that how you would describe it?
JOHN THOMPSON
Yeah. You do. In fact, Dave, there's another concept that's tightly tied to it: the personal project portfolio. I was talking to someone else the other day and they said, “Yeah, we've taken your personal project portfolio and we've called it the prioritized personal project portfolio.”
And I'm like, “Wow, how many Ps can we stream together in one acronym?” But anyway, the idea of having these portfolios—you've experienced this, I know you have—is that we have data scientists working on a project and we're encouraging them to be on the edge, experimenting and trying things. That project blows up because they tried something with an algorithm or they tried to integrate data and it didn't work. What happens when that blows up, if it's the only thing they're working on, is a massive freak-out.
So these people love what they're doing, they’re way into it. They're super committed and when they freak out, it's almost like a super meltdown. You get a call in the middle of the night, “Oh my God, John, it's not working!” And I'm like, “I don't want to hear this at midnight. I want to be sleeping.” There's a little bit of personal preservation in there too.
If you give them two big projects, a few small projects and some service requests, and you have a portfolio they're working on...when that project blows up, the last thing they want to do is tell you about it. When it blows up, they immediately switch their focus to a different project and work on it. More often than not what happens is that, if you give them enough space and autonomy, they'll come back a week later and say, “You know what, I was working on that project and it blew up and it's totally a mess, but it was working on project two. All of a sudden, when I was out fishing or boxing or taking a shower or whatever, this bolt of insight came to me and now the project's back on track.”
I didn't have to hear the freak-out and they were able to gestate on this idea for five, seven, 10 days, two weeks, whatever it was. Then they could talk about their failure and own it in a very exciting way. So I created that construct for these teams and it's been great.
DAVE COLE
How do you track that? On an individual basis? Is there a way to track these things and figure out when somebody's portfolio has too much, to dial it back? That's the job of a manager. I imagine that's our job.
JOHN THOMPSON
This happens all the time, too. Everybody listening to this: if you're going to try this, be ready for it. Every time I brought someone into my team, I said, “You're going to have a personal project portfolio,” or, “You're going to have two major projects.” A major project is something like six months to two years long. It's a big project and takes a long time to work on it.
“You're going to have a few small projects, two or three. They're going to be four months, something like that. The CEO shows up every now and then and says they need to understand this for the board meeting. Guess what? We're going to work on it, so you're going to get some of those, too.”
The first thing they say is, “That's too much. I can't do all that,” but once they get past the kvetching, bitching and moaning then they come back and they go, “Ooh, I kind of like this.”
We have weekly meetings. The whole team has weekly meetings and we never say, “What happened to that other project?” I always have that conversation in a one-on-one setting. In the team meetings, everybody's talking about what they're engaged in. So in their portfolio, they may be engaged in two projects or one and a half projects, and four of them are on the backburner or something like that. We never talk about the backburner stuff in the meeting. We're doing code reviews, talking about projects, roadblocks, data algorithms and all that kind of stuff. It's a very active discussion. Every two weeks I have a one-on-one with everybody. That's where we talk about projects that are off-track, our daughters and dogs and sons and car restoration: all the other things that people want to talk about. It works out really well that way. It's amazing. After someone's been doing it for two or three months they're like, “I never want to go back to just having a project again.”
DAVE COLE
Yeah. That's great advice. Just me personally, the way I am, I like having multiple things for exactly the same reason that you described. I know that maybe one of those projects or ideas might be a little bit of a moonshot and fall flat, and you have other things you can point to. You can say you’re working on five things and three of them are going well; two are off the rails and here's why.
JOHN THOMPSON
We all know that too, because every project we're working on, and every project you guys are working on out there, should have a business-oriented subject matter expert engaged. Sometimes those people go on vacation. Sometimes they get sick. Sometimes they are on maternity leave. There's lots of reasons that your projects slow down or stop, that are out of your control. If that's the only thing you got going on, you kind of get hosed.
DAVE COLE
So there's another topic that's come up on the Data Science Leaders podcast before, which is balancing project-based work and research work. A lot of projects are blessed by yourself or the business side, or both, before they get underway. Sometimes projects come from just messing around with data and obviously you have to have some business acumen. You can't just build art out of this data or something like that; it's got to have some value. How much would you say that ‘not blessed by a business user’ or maybe that which you're even skeptical of yourself, as a data science leader, do you allow in, and allow your team to go on?
JOHN THOMPSON
We have said that we want to have somewhere between 10-12% of your time when you're just goofing off, trying stuff—
DAVE COLE
—and learning, obviously.
JOHN THOMPSON
Yeah. One guy was playing around with neural nets and he wanted to try a long short-term memory architecture and I said, “Sure, go ahead and do it.” He came back to the team with a really nice tutorial on where it worked and where it didn't. It didn't really work as well as I had hoped it would, because I wanted to have that time when people could feel unstructured, relaxed and expansive in their thinking. But life gets in the way, your projects get in the way.
I changed it around to when we're scoping a project, you're building the project charter, that the front part of the project should have a component where you're doing blue sky thinking and just out-of-the-box innovation. We finally got the executives at CSL to understand that we could not, as a data science team, tell them that the project was going to finish in six, seven or eight months. We could tell them it was probably going to finish somewhere between eight months and a year, but we didn't really know how much experimentation we were going to do on the front end. They were like, “This is not the way we usually hear executives position it.”
And I'm like, “Well we could just do it and get it done, and you would get 2x improvement or you can let us experiment and we might come up with something that is 100x better than you've ever seen before in your life.”
And they finally said, “You're going to do that in every project, and we may have a delay of a month or two or three or something like that?”
And I'm like, “Yeah so you either let us play around and innovate and screw up and fail a few times and come up with something mind-blowingly good, possibly, or we just get on the hamster wheel and do it and improve it by 2%.”
And they're like, “Okay we hired you to do that, and we believe you're going to be able to do that and we think that team can make it happen. We understand there's a little bit of a squish in the beginning of your projects and we're cool with that.”
DAVE COLE
I'm hearing a little bit of your artisanal bent coming through there because I think the artisanal model allows for that spitballing, creativity and that shoot-for-the-moon mentality. I think if you have a factory-based approach, you're going to get the 2x improvement. That is a little bit more formulaic but you're not going to get the big step-change improvements that you might want to have happen as a data science leader.
JOHN THOMPSON
As a CEO, COO, or CFO all read about these huge improvements, in airline magazines or Forbes or wherever they are reading these stories, they're like, “How come we haven't transformed the whole company in the span of a week?”
DAVE COLE
Yeah. I know. “Why isn't an AI embedded in everything we do?” I hear you. I think there's a whole other podcast episode in terms of setting expectations. I think you touched on a bit of it that we could probably talk about for another time but, John, this has been awesome. If you haven't read John's books, you can find them on Amazon and check them out. I look forward to reading “The Future of Data.” Do we have any ETA on that, John?
JOHN THOMPSON
It's one of those things that, as we talked about, I wrote the last one in three months. My publisher's like, “Are you going to have it done in three months?” And I'm like, “No, no.” I think it's going to take me about a year to write this book, so probably February of next year, something like that.
DAVE COLE
Great. Well, I look forward to reading it and I thank you again for coming on the Data Science Leaders podcast. I had a blast. It's been fun. Just a lot of fun, John.
JOHN THOMPSON
Always fun. I always enjoy the conversations with Dave. I think we probably do get off the rails a little bit more, but it's always great.
DAVE COLE
Yeah. No more advertising to John. That's the big takeaway from this episode. It doesn't work. It's a waste of money. All right. Well, thanks again, and we'll talk to you next time, John.
JOHN THOMPSON
Take care, everybody. Glad to be here.
Popular episodes
What It Takes to Productize Next-Gen AI on a Global Scale
Help Me Help You: Forging Productive Partnerships with Business...
Change Management Strategies for Data & Analytics Transformations
Listen how you want
Use another app? Just search for Data Science Leaders to subscribe.About the show
Data Science Leaders is a podcast for data science teams that are pushing the limits of what machine learning models can do at the world’s most impactful companies.
In each episode, host Dave Cole interviews a leader in data science. We’ll discuss how to build and enable data science teams, create scalable processes, collaborate cross-functionality, communicate with business stakeholders, and more.
Our conversations will be full of real stories, breakthrough strategies, and critical insights—all data points to build your own model for enterprise data science success.