Value Driven Data Science: Boost your impact. Earn what you’re worth. Rewrite your career algorithm.

[00:00:00] Dr Genevieve Hayes: Hello and welcome to Value Driven Data Science, brought to you by Genevieve Hayes Consulting. I'm Dr. Genevieve Hayes, and today I'm joined by Douglas Squirrel to discuss strategies for making data science teams profitable. Squirrel has been coding for 45 years and led software teams for 25. He uses the power of conversations to create insane profits in technology organizations of all sizes. His experience includes growing software teams as a CTO in startups, consulting on product improvement, and coaching a wide variety of leaders in improving their conversations, aligning to business goals, and creating productive conflict. He lives in Frogholt, England in a timber framed cottage built in 1450.
[00:00:49] Squirrel, welcome to the show.
[00:00:52] Douglas Squirrel: Glad to see you Genevieve.
[00:00:53] Dr Genevieve Hayes: For many people, data science is synonymous with machine learning and many data science courses are little more than overviews of the most used machine learning algorithms and techniques.
[00:01:06] Douglas Squirrel: What a waste. That's so depressing.
[00:01:09] Dr Genevieve Hayes: I know, isn't it? And this is great. If your primary motivation in becoming a data scientist is to participate in Kaggle competitions, but less so if your motivation is to take those skills and apply them to industry.
[00:01:24] Where the majority of data science courses fall short is that they fail to bridge the gap between data science theory and business reality, resulting in data scientists who are technically strong, but unable to create value from their work.
[00:01:39] Douglas Squirrel: They're stuck in the ivory tower is how I often describe it to business leaders who hire me to go and look at what they're doing and figure out what on earth is happening in the tower.
[00:01:49] Dr Genevieve Hayes: And one of the sets of techniques that you use is action science, which according to your website, you've been using to make technical teams insanely profitable since 2001. Well, we're going
[00:02:02] Douglas Squirrel: That is right. Well, I haven't been using action science to do that. I didn't know it existed until 2011 or something, but I used a lot of the techniques without knowing it.
[00:02:16] Dr Genevieve Hayes: into that, Squirrel, backstory. What led to you going into the business of helping tech teams to become insanely profitable in the first place?
[00:02:28] Douglas Squirrel: Well an awful lot of failure to make tech teams insanely profitable. So I have a wide and deep experience of screwing up and I make a tremendous use of that. So the brief squirrel history is that I was a CTO or VP engineering. In a series of startups here in England.
[00:02:47] And at each one I built a team and sort of got it going. And then I got fired and I got fired in this really nice way. Every time founder would come to me and say, squirrel, you've built this team and they're doing amazing work. There's somebody you've trained to lead it. Their processes are great.
[00:03:04] We're doing fantastic stuff here. I couldn't be happier, but there's not much for you to do anymore. And you're kind of expensive. So would you please go be wonderful somewhere else? And it was a really nice way to get fired the third or fourth time. And they each got shorter. So it was 10 years and then two years.
[00:03:20] And then one I said, gosh, I really should kind of plan for this. This is something that I seem to do. And maybe instead of just getting fired, I could make it part of what I do. So I started as a fractional CTO. And so I would work with teams two, three days a week. I'd have two or three clients.
[00:03:37] But I kept getting faster. So those initial 9 month engagement between 6 and then really the time that I take now is is very short 2 to 3 months. My record is 3 weeks for turning somebody around. That was an extremely motivated person who did all his homework and met me every day. Not everyone does that.
[00:03:56] I coach people. I transform teams. I evaluate organizations. I do strategy. I do all those sorts of things to help technology be part of the business, be a profit center rather than a cost center.
[00:04:09] Dr Genevieve Hayes: How do you define a tech team? Because obviously we're not just talking about data science teams here.
[00:04:15] Douglas Squirrel: Absolutely. So tech team for me is, and I intentionally use that term, not an engineering team or a software development team or something. It's all the people who are doing technical work. And that's usually, as you say, much broader. First of all, some people say, yes, we've got our tech team and then we've got our data science team, that's a problem that sometimes people say, yeah, we've got our tech team.
[00:04:33] And then we have our QA team, we have our tech team, and then we have our product team. We have our tech team and then we have our technical customer service or technical support. So I don't want to distinguish any of those things. I want to make sure I'm evaluating and coaching and transforming all of the people who are doing technical work,
[00:04:50] Dr Genevieve Hayes: So would a finance team count as a tech team?
[00:04:54] Douglas Squirrel: not typically, because they're not usually doing something that is in any sense technical, but there have been exceptions. So you know, there are companies who do a highly financial task for their customers and their finance teams are sometimes a big part of that. I have one who Reconciles accounts payable and figures out, you know, make sure you don't pay people twice.
[00:05:15] That sounds like a good idea and guess what their finance team uses the software and is a big part of making sure that they do the right thing.
[00:05:22] Dr Genevieve Hayes: But data science, software development, those teams are definitely in your room if they're tech.
[00:05:27] Douglas Squirrel: Yes.
[00:05:28] Dr Genevieve Hayes: Okay. Do you work with the team members themselves, or do you work with senior management? It's
[00:05:34] Douglas Squirrel: Yes, so all of the above. Increasingly, I find that I get the most leverage from working with, say, the CEO, or sometimes the CFO or the COO. People are outside the tech team as we just defined it, but who are its major customers and funders, right? They're the people writing the checks. And they often find, as I was alluding to before, that there's some strange thing going on over there, and everybody tells them it's fine, and they have no idea what that is, but they keep writing the checks.
[00:06:01] And usually I find that there's a lot of value there, and then a bunch of things that they shouldn't be writing checks for, and we do something about that. But in order to do that, I need to work very closely with the people in the tech team in order to understand what's happening, whether it's efficient, how it could be better.
[00:06:16] And so I'm coaching and leading and encouraging the senior leaders of the tech organization, CTOs, VP engineering, heads of data science, folks like that, and individuals within those teams.
[00:06:29] Dr Genevieve Hayes: only really been in the last decade or so that organizations outside of Silicon Valley have really started to have data science teams. When was the first time you worked with the data science team?
[00:06:40] Douglas Squirrel: That's an excellent question. Now I'm going to make a radical statement. We didn't know it was a data science team, but it was back in 2011 or 12. I'm never good with dates. And that's the second of those employment journeys. I was fired from the first one and went on to the next one in that friendly way I talked about and that e commerce company.
[00:07:02] Had 4, 000 new products arriving in the warehouse every week. Even, you know, we didn't have the sophisticated tools that we have today. There's still pretty good Shopify that would deal with your t shirt store or something and could run something with a reasonable product line, but we were completely unreasonable.
[00:07:18] So we had to build all kinds of tools ourselves. Oh, and we sent I think it was a million and a half emails every day to people to tell them to buy our stuff. So it had to be a very efficient operation. There were lots of problems with it that I helped solve, but we had People who functioned like data scientists, but we didn't think about it in that way.
[00:07:36] For example, we had people who did data cleansing, but they weren't people who'd gone to university to learn data cleansing. They were folks who used to work in our warehouse and they were called the processors because nobody knew what else to call them. And an entertaining thing is that when I first arrived.
[00:07:51] Somebody pointed out there's the processors over there. And I said, what do the processors do? And they said, you know, none of us know. We're not really sure it seems to keep working, but they do something and it involves all the new products that arrive and they're doing stuff with it. And they sit over there and they do something.
[00:08:05] And you know, I'm not really sure. And eventually I sent our head of product. Over there, and she sat with him for a day and she said, my God, these people are amazing. What they're doing is taking badly faxed invoices and details and comparing to the actual items, which they bring in from the warehouse and they look at them and they do a ton of data cleansing.
[00:08:26] And they have a whole bunch of sophisticated tools they used to do that. And none of us had ever noticed because they used to be off in the warehouse. They had come into our office when we outsource the warehouse and that they were doing a tremendously important. Data cleansing job that no one knew was happening.
[00:08:43] Dr Genevieve Hayes: I remember at the start of my career working in an organization that I now realize had all these data engineers and people like your data processes in the background. And I realize now how much I took them for granted
[00:08:59] Douglas Squirrel: course,
[00:08:59] Dr Genevieve Hayes: I know it's such an important role, but me, 15 years ago, just didn't recognize this.
[00:09:05] Douglas Squirrel: absolutely. And a lot of folks didn't. And the fact that we have names for these things now and that we have processes and degrees and stuff like that, doesn't really change what they're doing. We have a lot better tools now, but people aren't necessarily using them as well. And especially when they are, as you say, outside Silicon Valley, they don't have AI in their title or something like that in the name of their company the tools and the way they're used is still pretty primitive.
[00:09:34] Dr Genevieve Hayes: You've been doing this for over 20 years. How have things changed, particularly with regard to why tech teams fail to be profitable in that time?
[00:09:44] Douglas Squirrel: Oh, well, they're not profitable for all the same reasons. Things like not talking to customers and building things that make them happy rather than customers. So the reasons for not being profitable have remained the same. The industry landscape has changed in a couple of ways. First, we've Systematized and automated a number of things that would have been horrendously complicated before, because we kind of conquered that territory.
[00:10:08] We understand it now. We've mapped it. We know how to do it when I first started setting up a website where people could give you their credit card and buy something was a massive effort. He needed tens of really experienced engineers for at least a year to set up something that would do that.
[00:10:23] And now you just go to Shopify and click a few buttons and you can run a website. It. So we've kind of got that to a state where we don't have to worry about it. But there are lots of other things, the boundaries have moved out and people are now operating in very different ways. For example, we've discovered this discipline of data science.
[00:10:41] We've given it a name. We've put people in roles with data science in the title, but we still don't have the kind of Shopify level. Understanding of what the methods are, how to set something up. And so you're still doing a lot of things from first principles. Every time you get going.
[00:10:58] Dr Genevieve Hayes: So what are the main reasons why you see data science teams failing?
[00:11:01] Douglas Squirrel: Well, the very most common one I alluded to before, and this is where you get a divide between some very theoretical folks who are doing wonderful things, absolutely exciting, cutting edge, change the world kind of stuff, but it's unaligned from what other people are doing in the rest of the business.
[00:11:21] And so you get this phenomenon I call the walled garden. And so inside the walled garden, you have a tech team. And in the case we're talking about perhaps a data science team that has a beautiful topiary hedges. And they just trim them beautifully. And they write fantastic papers about all the great things that they're doing.
[00:11:38] They're really pushing ahead human knowledge. They're trying things that nobody has ever tried before. The unicorns really like the grass in the walled garden. You know, it's a really comfortable, happy place. And you go see the engineers and they say, I've never been anywhere better. I'm so happy here.
[00:11:52] We're doing such wonderful work and the code is really clean. Everything's great. You leave the walled garden and there's a zombie apocalypse. There are angry customers chasing down customer service people who are running in fear. And nothing is as it should be for the rest of the business, but there's been this complete divide.
[00:12:10] There's this wall that's gone up between the two. Now that's an extreme example. There are certainly many cases that are not quite as pathological. The wall's not quite so high and their unicorns maybe haven't arrived yet. But when you get that very significant divide between the data science team and what they're doing, and business results that always spells trouble.
[00:12:29] Dr Genevieve Hayes: How does it arise?
[00:12:31] Douglas Squirrel: I think the main reason is our fault, yours and mine, and all our listeners the folks who participate in this, because we've allowed this Sort of mystique to grow up around technology. And it's got this special characteristic that nothing else does. You talk to people who are outside tech, you talk to say the CFO, like we were talking about before, and that finance person says, yeah, you know, I read all these checks and there's all this very expensive equipment the hosting center and so on, and I go and ask how we're using it and , why is the cost tripling in the last year, what's changed.
[00:13:04] And I can't understand the answer. So it seems to work and, you know, nobody's died yet. So I've kept it going and I don't know any other alternatives, but I definitely have no clue what's happening there. And that person would not accept that in any other category. So if the customer service people said, Oh yes, we're doing special customer service things.
[00:13:23] You know, we need fancy phones. That's how we work. You know, we can't explain to you. The special computers and phones and phone lines that we need in order to talk to our customers. Finest person never accept that. They show me the phones. Let's see if we can get some cheaper ones. Maybe we can send the call center overseas.
[00:13:37] We can do something. The finest person gets involved. Same with sales. Same with operations. The people outside the business are willing to participate in, but somehow technology is so complicated and so special and so different. It's not, but people think it is. That they can't engage, and that's a real problem.
[00:13:54] Dr Genevieve Hayes: One thing I find interesting is my, I started off as an actuary before I went into data science and you don't tend to have that same walled garden effect in actuarial.
[00:14:06] Douglas Squirrel: Of course not, because people think they can be actuaries. Hey, I know that, you know, if I just sat down and added up enough numbers, I could figure out how often people are going to have accidents. I kind of understand how that would work. So if you come along and say, oh, got a special thing, I need.
[00:14:19] To go to a special conference in Bali in order to learn about this, you say, yeah. So tell me what you're gonna learn data scientists comes along and they say, okay, you go to Bali. I don't know what's going on. I can't understand anything. You're saying that's nuts. That's wrong.
[00:14:33] Dr Genevieve Hayes: Yeah. Is I've found even in teams that are both actuarial and data science combined, you tend to have. Part of the team ends up in that walled garden situation, but part of it doesn't, and it's weird to see.
[00:14:47] Douglas Squirrel: And unfortunately, the frequent recent phenomenon is the data science are in the ivory tower. The data science somehow get more isolated because it's perceived as, and we reinforce this perception special tools that are only available to wizards with fancy names. You know, we have to have attention and neural networks and gradient descent and all kinds of other clever complicated things.
[00:15:11] Dr Genevieve Hayes: And we don't explain, what we're basically doing is showing a lot of pictures to a computer. We're having it get better and better at figuring out which pictures are right. And we want to spend more money on more pictures. One thing I've observed, and I've seen this in more than one organization, is with the actuaries, the organization will allow the actuaries to interact directly with senior management. With the data scientists, they'll often create some sort of bridging team between them.
[00:15:41] Douglas Squirrel: Because they're too geeky. They're too scary. They might make a bunch of Star Wars references. And the board of directors might be disturbed by that. Because these special people over here are just too geeky and weird. And this persona, this kind of patina of nerdiness and somehow specialness that we're somehow special wizards or something.
[00:15:58] It's just so damaging. So if listeners don't take anything else away from this, I hope they take the view that it would be really helpful if Data scientists, whether that's them or people they work with we're spending some time, not only with senior management, but with actual customers. So put them on the phone lines in customer service, have them ride along on sales calls.
[00:16:18] And you think, why on earth would that be useful? Don't I want these special, highly paid, very expensive fancy people putting their hands on the keyboard all the time? No, you don't, because they're going to steer in the wrong direction, going to build the wrong thing that you don't need.
[00:16:31] Dr Genevieve Hayes: A lot of the data scientists I don't think are any more geeky than your average person in the organization.
[00:16:36] Douglas Squirrel: There's this myth that somehow they are.
[00:16:38] Dr Genevieve Hayes: But I've actually had situations where people have actually said, I'm not going to bother the data scientists with my request because it's not good enough for them. I'll just go to other people who are less highly paid and less. It's highly qualified because we need to reserve the data scientists for these special tasks which never actually come along and so the data scientists just sit there and do nothing.
[00:17:05] Douglas Squirrel: It's a sad story and repeated far too often.
[00:17:09] Dr Genevieve Hayes: And then the data scientists quit because they have nothing to do. And it's like, yeah, Hmm.
[00:17:13] Douglas Squirrel: Don't do these things. These are, these are damaging. I think we agree.
[00:17:17] Dr Genevieve Hayes: So you've probably seen it all by now. I'm guessing you keep going into organizations and seeing the same thing over and over again based on what you were just saying. Do you ever go into organizations these days and find yourself thinking, Hey, that's something I've never seen before?
[00:17:35] Douglas Squirrel: Let's see. When's the last time I got surprised? I get surprised all the time. So there are variations in every organization I go into, but let me try to think well, no, I know the one that I was most surprised by recently. There, there's an absolute technological explosion in Africa. And there's incredibly qualified people who haven't sort of had the brain drain, who haven't disappeared from their country for whatever reason that might be.
[00:17:57] But I have a client in Africa who is taking over a market kind of in a Facebook style growth that they're just going market to market to market at incredible pace. They hired 50, five, zero engineers in two weeks in order to capture Four or five different markets. They're not just going after one.
[00:18:15] They're going after multiple countries and multiple solutions, multiple customers. So that really surprised me. First of all, could that possibly be a good idea? Yes. In their special case, it was. And can they make it work? Well absolutely. Astonishingly, they have succeeded and are succeeding today.
[00:18:30] So that's a client with a totally different situation. Cause I normally never say, please hire 50 engineers. That would normally kill most companies, but these guys are just eating it up.
[00:18:40] Dr Genevieve Hayes: Which country are they in?
[00:18:42] Douglas Squirrel: They're in Nigeria and Kenya and expanding into South Africa.
[00:18:46] Dr Genevieve Hayes: Yeah. It doesn't surprise me because I was thinking those would be the countries.
[00:18:49] Douglas Squirrel: Those are the ones that are advanced, but man, there's a lot catching up. So I predict great things on the continent. They're really, really skilled people who are a little less mobile. So they're not able to just get on a plane and go to France or something and go live there and work there. So they're at home, they're ready to do things and they're tremendously skilled.
[00:19:09] Dr Genevieve Hayes: That's fantastic. Fantastic.
[00:19:10] Douglas Squirrel: I think so, too. So there's a surprise.
[00:19:13] Dr Genevieve Hayes: So. We've talked about some of the reasons why people fail and there's the walled garden effect. Are there any other key effects that we haven't mentioned so far?
[00:19:24] Douglas Squirrel: Well, it's related to one that you brought up. But the failure to have important conversations is is one that I focused on a lot in my career. And you mentioned it in the introduction and the one you alluded to where somebody says, well, I'm not going to bother. Those people is just 1 of a host of different ways that you can avoid difficult conversations and really get yourself tied up in knots.
[00:19:43] So another example Would be a data science team who has really severe technical challenges. You know, they don't have data engineers who can provide clean data. They're they're suffering a lot of drift. So their concepts are changing underneath them. And nobody's going back to train the model.
[00:20:01] So I'm just making up things that. Could be happening and they don't tell anybody, or they try to tell people and no one listens. So those conversations about what is going wrong in the team might be happening internally, but the news isn't reaching the rest of the organization.
[00:20:15] That's another sadly, all too common failure mode.
[00:20:18] Dr Genevieve Hayes: Yeah. So I've understood what the problem is , and I'm sure many of our listeners are painfully aware of the problem because, you know, there's This is their lived reality. The big question is, what do you do about it? How do you change the situation and make tech teams, particularly data science teams, insanely profitable?
[00:20:40] Douglas Squirrel: Well, so how to do it soup to nuts would take us the rest of the podcast, but here's the brief outline. The typical situation. Is that nobody inside or outside the team has any idea whether the team is profitable. There's just no measurement. There's no way of telling whether there's a connection between all the money that you're putting into a bunch of very clever, very capable data scientists and customers who give you money for products.
[00:21:03] So that's a normal situation and the thing you want to do is to make sure that you have some way of measuring whether there's been an effect. And one of the biggest challenges that typically comes up in trying to do that when you say, well, gosh, how is the latest new model that we've created or the fantastic new marketing that we're able to do as a result of analyzing our customers or whatever it is the data scientists are doing.
[00:21:27] The problem is even if you tried to measure it, you couldn't measure it very frequently. So you can't tell whether the thing you did six months ago that you only released yesterday actually has had any effect. So one of the things I work on most frequently is decreasing the cycle time.
[00:21:43] So the time between when you would like to work on something, when you think, hey, there's a great idea, this could make customers lives better, it would probably lead to upsells, reduced costs, whatever it is And the time between when you think of that and the time when a customer actually sees it and when you can actually measure a change in pounds and pence, that time is way, way, way too long.
[00:22:00] It might be six months. It might be a year. It might be a month. It might be three, four weeks. Those are all way, way, way too long. And a sprint is still too long. If people were working in one or two weeks sprints, I'm aiming for every day. I want to make sure that the cycle time, the time between when you're thinking of something and the time when you measure it is as short as possible.
[00:22:19] Even in my biotech clients, where they're building medical devices, things that cause people to cut off parts of their body. If the test comes out wrong, those folks still manage to get something new when I work with them every two weeks. And that's totally revolutionary in their world. Whereas in say e-commerce or finance or music or fashion or other places that I've worked in every day is absolutely possible.
[00:22:43] And it's possible for data scientists. It may take you weeks to train your new model. It may take you a huge amount of time to determine which of many new innovations. To implement. So I'm not suggesting that you're going to make every specific action happen in a day, but it's certainly possible to have new things coming out.
[00:23:03] With tremendous frequency from your data science team and to measure them. And guess what? If you measure the actual profit and loss, what's the result in pounds and pence of what you did? And you do that very frequently. It's actually really hard not to improve. And so that's kind of the engine that drives getting to significant profit out of your data science team.
[00:23:26] Dr Genevieve Hayes: this sounds fantastic, but based on some of the organizations I've worked in in the past, often these have had 12 month cycles because of things like insurance renewal. I can't envisage how this might work in practice. So suppose you're dealing with an organization that did have a 12 month cycle due to 12 month renewals.
[00:23:48] Douglas Squirrel: Does everybody renew on the same day?
[00:23:50] Dr Genevieve Hayes: Yes, in this scenario, I'm thinking of
[00:23:52] Douglas Squirrel: that's why I asked, so that I could make sure about it. So there's April the 1st, everybody in the whole country gets their new maritime insurance, or whatever it is.
[00:23:59] Dr Genevieve Hayes: Same state. So one place I worked for was WorkSafe Victoria, which did the workers compensation insurance for every business in the state of Victoria. We had about 200, 000 clients and a hundred percent of those clients all renewed on the one day.
[00:24:17] Douglas Squirrel: Got it. Except some of them didn't, because I'm sure they went to some other provider, got
[00:24:23] Dr Genevieve Hayes: monopoly provider.
[00:24:25] Douglas Squirrel: it, okay. So there was somebody else who was deciding on you, I assume the government said, you can be the provider, and that they could have fired you at some point, that they could say, oh, Squirtle's company is a better provider of insurance, or were you working direct for them?
[00:24:37] Dr Genevieve Hayes: I was working directly for the government.
[00:24:39] Douglas Squirrel: I got it. Okay. I just got to get the situation clear. So what you don't have is a situation where you can have people go through the renewal process on a different day. Each time. First of all, this is unusual. It's a great unusual case, but most folks listening to this, I think, will not be in a situation where every one of their customers buys every single year is forced to do so and does it on the same day,
[00:24:59] Dr Genevieve Hayes: Yeah. That's why I wanted to see your answer to
[00:25:01] Douglas Squirrel: which is great.
[00:25:02] This is a wonderful example, but the thing that you can do here is there's 364 days. On which you can make changes in preparation for the only day of performance, a singer doesn't only sing when he or she goes up on stage with thousands of people in the audience.
[00:25:17] The singer sings in the bathroom, the singer sings in practice, the singer sings for his or her coach. There are lots of places where that person gets feedback on what they're doing. If they have one performance a year. They don't wait till that performance to practice what they're doing. So you can do exactly the same thing.
[00:25:34] Dr Genevieve Hayes: ones. Yeah.
[00:25:35] Douglas Squirrel: You do dry runs and you have some mechanism of getting feedback on those. So this dry run had a 2 percent error rate. And this error rate means that we have to do manual steps to fix up the insurance policies that were, that 2 percent error rate is way too high. If we can get that cost down.
[00:25:54] Then we will have fewer manual correction steps when we actually do the real performance on April the 1st or whatever it is. And so you could do those practice runs and get lots and lots of feedback from proxies for your users from real users. There are probably people who buy an awful lot of workers compensation insurance and would really like to have fewer errors because it's lots of manual work on their end.
[00:26:16] It's a big employers for example. And so you might be able to get a user panel of some variety to say, okay, here's the results of the latest dry run. And they say, this is wonderful, but you've really messed up this part. And if you don't get the premium scales, correct then that's a huge amount of work for us.
[00:26:32] And you say, Oh, I didn't realize that was a problem for you. Let me fix that. So when it's time to renew, we can do something different. So you've given me a hard case, which is great. I like that. Okay. But it's eminently possible to have very frequent test runs so that you're getting lots and lots of feedback before your big day.
[00:26:48] Dr Genevieve Hayes: I just finished reading the lean startup and a lot of what you're saying.
[00:26:55] Douglas Squirrel: Reese was ahead of his time, had lots of really good ideas there, and these things definitely work.
[00:27:00] Dr Genevieve Hayes: Yeah. I finished reading it about a week ago and I've been thinking about how this could be applied to data science teams. And it sounds like this is actually what you're doing.
[00:27:09] Douglas Squirrel: But Reese assumes something, which we've been exploring a lot here. He assumes very naturally that you're in a startup, that you're in some kind of environment. In which it's relatively easy to get data about what's happening. So he tells the stories about his chat tool and how they rolled it out to 5 people.
[00:27:27] And they were really excited that 5 people used it. And then it was 10 and so on. You could go and look at that stuff you could participate in. I imagine when you were dealing with the government and the workers compensation information that was all locked behind closed doors. There were rules and regulators and there were committees and things that made it much more difficult for you to get at the information.
[00:27:45] So you need a broader set of steps to not just say, well, we're going to actually do this frequently. We're going to release frequently do a lot of dry runs in the example we were talking about. But also you need that feedback. So the fancy term for this, which comes from a guy called Dave Snowden.
[00:28:02] Is to probe sense and respond. So you might be able to get the probes happening frequently. But unless you know, you send the space probe out to Venus, if it can't radio back to you, it doesn't do you any good. So you need something that comes back and says, yes, now you've gone from a 2. 2 percent error rate to 1.
[00:28:19] 9%. That's this much better. That allows us to save this much cost and you say, great, we're going to stop worrying about cost. We got to our goals there. Now, we need to be able to provide our service to a broader section of customers. So let's go for that. Now, you need that kind of feedback.
[00:28:33] Dr Genevieve Hayes: So it sounds like there are two key aspects to this. One is increasing the frequency of what you're sending out into the world. But the other thing is having those metrics in place to measure it, because otherwise you could send things out every hour and it would be worthless because you're not seeing, are we driving down this error rate or Whatever it is that drives costs.
[00:28:56] Douglas Squirrel: I have a client here in Britain. With our NHS. We're doing a new program where we try to track people over a long period and learn a lot more about them. It's called our future health. So it's this massive program expanding at a tremendous rate, not quite as much as the Nigerians, but not far off.
[00:29:12] And I was coaching someone in that organization who has to connect researchers who are using this data and the engineers who are creating the tools that allow the clinicians in the field to actually gather it. And yeah. One of the biggest challenges there is actually keeping those folks in sync and having enough contact with both of them.
[00:29:32] So I had to get him out of the building and get him going off to talk to actual researchers in actual wet labs who were using this data because that hadn't happened. And in starting to get that feedback, he started to understand what those folks needed and what he could do differently in the engineering side.
[00:29:50] So they actually got the data that was valuable to them. That feedback loop is often missing, especially in government, especially in large organizations where it's organizationally more challenging.
[00:30:00] Dr Genevieve Hayes: So I suppose understanding that is what would drive which models do we need to release on a frequent basis and the metrics that you measure when you're releasing them.
[00:30:12] Douglas Squirrel: Exactly. And whether we need any models at all, maybe what we need is better ways for users to interact with the software and better ways for them to search the results that we're providing if you sort of limit your probes and limit what you're going to be sensing then you're limiting the Types of options you consider.
[00:30:30] So you want to make sure it was as broad as possible, including, Hey, wait a minute. We don't need any more models. This actually is giving us what we need. But what we desperately need is a better way to search better way to slice the data, better way to graph it. Something else that's more meaningful for us.
[00:30:45] There was a great example in finance where I really thought speed was the most important thing for this organization. It's finance. People need to do things quickly. Why on earth wouldn't they want more data more quickly so they could act better? And I figured out that there were some people who had a job which it must have involved lots of drinking of coffee because you would click a button and you would drink a full cup of coffee.
[00:31:07] And then you would go get more coffee and then you would drink another cup of coffee, and then finally your report would come back. And I said, you know, we can make this happen in seconds. Wouldn't this be wonderful? You don't have to pay those people to sit there and drink coffee. Cause they really couldn't do anything until they got the report back.
[00:31:20] And I went to the folks who were funding it, the company that had brought us in to implement this software for them that we had, and I said, this is great. We're going to make this faster. Isn't this wonderful? Thank heavens I asked because they said, my God, don't make that faster. Those are folks who were in another country.
[00:31:36] They work overnight. We love them. They're great, but we'll buy as much coffee as you want for them because we don't care how fast they get it done. We just need it in the morning. So it's perfectly okay. We don't need to make that faster, but what we really need is better reports. We need fancier charts, fancier graphs, better things that will allow us to sell our services better to other people using the information these folks produce overnight.
[00:31:57] We don't care how fast it is. So thank heavens I asked. Thank heavens there was the feedback loop in place because had I come back and said I made it faster, I would not have got their renewal business.
[00:32:07] Dr Genevieve Hayes: So it's just having conversations with the right people who are the people who the data scientists probably haven't had the conversations with to begin with.
[00:32:15] Douglas Squirrel: Yep. And some listeners are probably thinking to themselves, Oh, it's okay. You know, I have my product people. I have my, I don't know, my agile coach. I never know what those people do. The, you know, I have somebody who's in between me and the customer. So they come and represent the customer to me. So I can stay here in my ivory tower.
[00:32:29] I can deal with my models and I don't have to talk to any of those smelly users that, that I don't like talking to, Maybe they're thinking, I would love to talk to those users, but there's these people in the way. Whichever attitude they have. It's wrong because there's tremendous value to having individual engineers talk to individual customers and I always have to really emphasize that.
[00:32:50] That's what I mean. I don't mean sort of. They read a bunch of reports from user researchers. You've gone to the field. It's not that I mean that they have really good product managers who reflect their ideas back to them. Although, Okay. It's criminal. Too many data science organizations don't even have any product manager people.
[00:33:05] There's no feedback at all to them. That's terrible. But even if you have that kind of indirect feedback, the problem is it doesn't allow you to use the things that only humans can do. And we do have computers that now can can learn right they can use gradient descent, get to , at least a local minimum that allows them to have a really good cost function.
[00:33:24] So there's some sense in which we can kind of set the computer off to go and discover the best solution. But the problem is that a human will be able to think outside the parameters that we've defined. And I just gave you an example of that. I'll give you another one if you don't mind. There was another finance company, but very different finance company where I was CTO.
[00:33:43] Our audience was eight to 12 year olds. So we had young children who had debit cards. I don't know if this is true in Australia, but these days you can get these pretty much at your local bank, but when we were doing it, it was totally new and there were a lot of very nervous people at the card issuers who said, well, my God, is this going to cause a problem?
[00:34:00] It doesn't of course, but we had to overcome that. And in the corner, I had someone who was doing what we would now call data engineering, but also a lot of system administration. And, Nick had all the characteristics that you associate with sort of the bastard operator from hell.
[00:34:18] You know, he wore the Star Wars T shirt every day. He sat in a corner that was very dark. He never talked to any other humans, but he was a wizard with the machine. And the problem was that he was being a wizard on things that mattered to Nick. And they weren't as important to all the rest of us.
[00:34:34] So I said, I'm going to solve this problem. And so luckily we're in a relatively small office. I said, Nick, I don't want you to sit there today. I want you to sit over there and I want you to answer phone calls from customers. And he grumbled and groaned and said, why are you making me do that? Why can't I sit over here in my dark corner?
[00:34:49] I said, just go over there. And he became a different human being. So when he got phone calls from crying nine year olds, I remember one particular one, this kid was crying in McDonald's because the kid could not get their Happy Meal or whatever it was, and they were hungry. So they were sad, and they were phoning us saying, you know, my card doesn't work.
[00:35:08] He figured out, because of course he was a wizard with a machine, he could figure out how to put more money on the card and unblock whatever was causing the problem and so on. So he did that in a few seconds. And he did it in a really sympathetic and energetic way that made the kid really happy.
[00:35:21] And the kid's laughing and, saying, this is great. I'm going to have my happy meal. And, Nick was really gratified by that, but the other effect it had was Nick was then able to really empathize with the customers and do different things. He was able to build the software differently and work on cleansing data and administering the system and setting up servers.
[00:35:39] So that we could deal with unhappy, crying children who couldn't get their money off their cards much more easily. That kind of direct contact unlocks things for engineers that you would never capture in a requirements document.
[00:35:52] Dr Genevieve Hayes: What if the customers don't have time to spend with the data people?
[00:35:58] Douglas Squirrel: So I never let my coaching clients use the phrase. I don't have time. And the reason is that you have 86, 400 seconds in a day. I have 86, 400 seconds. Everybody has the same amount. So it's not that you don't have time you're not prioritizing it. Now, some of those folks, including the people in the financial organization I was mentioning before, are making a million pounds a second, and it really doesn't make any sense for them to prioritize talking to data science people because they're really high powered folks.
[00:36:25] Most people listening to this podcast are not in that situation. So most of their customers would actually really benefit. From talking to data science people because they would get immediate benefits. Immediate results. Just like the customer service people who had to train Nick and sort of sit next to him and make sure he pushed the right button on the phone.
[00:36:42] Those folks could have been answering customer calls. But my God, I'm really glad Nick helped the 9 year old because then he was able to come back and say, we need to add this and that and the other thing so that we can. Deal with these kinds of issues much more quickly, make more happy nine year olds.
[00:36:54] So that's the normal situation. You're good at coming up with the tough cases. So I'm going to do it for you here. So there are cases where, you've got some high powered trader or somebody else who's, a doctor often kind of thinks they're one, one step down from Jesus.
[00:37:07] So they can't spend any time with those nerdy people in the corner, just have them do their job. So there are still tons of things you can do about that. First of all, you can watch videos. I've had folks who are spending time with those folks. Those folks usually do call the customer support line, or
[00:37:22] it's sometimes they're too far away, they're in another country. But you can watch a video, you can listen to recordings, you can participate in that way. I remember learning at that e commerce company a very important fact about how people used our app our website on a phone, because somebody said in a focus group that I was watching And then our marketing team, she said, now, the most important thing is that I'd be able to use your website with one hand.
[00:37:44] And I said, that's really interesting. And I kind of look carefully and rewound a little bit. And she had two hands. She definitely did. I could see that. So, it wasn't that she had a disability. She just really wanted to be able to use our website with one hand. I thought, wow, I wonder why that is. Luckily, somebody asked her, went forward in the video a little farther.
[00:38:01] And the person said well, actually, you know, I need to hold the baby with one hand. But then I need to be able to scroll and find my half price Louis Vuitton handbag with my other hand, because , that's what I do. To kind of keep myself entertained when I've got a crying baby jiggle the baby and I use your website.
[00:38:16] Well, that made a lot of a difference for our user interface that we had to make sure it was usable in this very noisy environment that mothers had because mothers were a lot of our customer base. So, that's 1 example where you can get lots of information, but there's other things that you can do, like looking at system logs or tools that tell you where people are clicking.
[00:38:35] Data science, of course, would be wonderful at statistically analyzing large amounts of user log information that can tell you a lot. And very frequently overlooked your company will have proxies. Your company will have account managers, customer service people, operators who have done this job.
[00:38:51] If your company, say, doing actuarial work for the government, it might be that the government has their own actuaries, but you probably have your own somewhere. Okay. So go find those people and have them be proxies to talk to your data science people. So don't tell me that there's no way to do it.
[00:39:05] And if you really think there isn't, then you should get in touch with me. Go to douglassquirrel. com and I'll explain to you how you can do it. I'd be happy to help any of your listeners who think they can't.
[00:39:14] Dr Genevieve Hayes: So let's say you can get hold of one of these customers to talk to. What happens if they tell you that they want something? You go off and provide it and then once you've provided it to them, it turns out actually they didn't really tell you what they wanted. They thought, they told you what they thought they wanted.
[00:39:30] And those are two different things.
[00:39:32] Douglas Squirrel: I have wonderful news for you. I can see that you're sitting down. So that's good. So stay sitting down. Okay, Genevieve, don't stand up because this may be totally shocking to you. I have the complete cure for this. This will never happen to you again. If you do what I say, you're right with that.
[00:39:45] That interesting.
[00:39:46] Dr Genevieve Hayes: Yep.
[00:39:47] Douglas Squirrel: Excellent. So the thing you do is you show them progress every single day and the good thing that will happen to you is you will get it completely wrong. That's going to happen anyway. So let's say that in your actuarial case, you're going to build a new model that will improve the pricing for the workplace compensation and you're going to have much better pricing and somebody has told you that you'd like a pricing model that works this way.
[00:40:10] And they're looking for you six months from now when it's time for everyone to renew to roll that out and we'll make lots more money from it. And what you can do is every day show them your progress toward that pricing model. And it may be that your progress is very small, maybe that your progress is not very meaningful to start with.
[00:40:28] But if you structure it right, you can do it so that you get that frequent feedback and there's a way for you to get that frequent feedback. So you're building it in a pattern by structuring your work so that you may go slower overall.
[00:40:42] But you're getting very frequent feedback, and it could be, for example, that you say I'm determining all the different inputs that I'm going to use. And this is my first input. I'm going to be drawing this piece of data from this set of people that we're ensuring. And there are 17 different things we're going to use in the formula and we're gathering number one.
[00:40:58] And they say, my God, number one is completely wrong. You've got the wrong model. You're analyzing this incorrectly. You're using the wrong type of payroll information. And you find out on day one that you're headed the wrong direction. Thank heavens. What a great result. And so if you structure your work in that way, and you get this very frequent feedback, it will be impossible for someone to come to you at the end and say it's wrong.
[00:41:20] They will come and tell you on day one or day two or day three that it's wrong. And then you can avoid the additional cost of investing more in doing the wrong thing.
[00:41:30] Dr Genevieve Hayes: How do you stop them from thumping you? Because they want you to go away and you're annoying them by showing up every day.
[00:41:36] Douglas Squirrel: You need more senior profit and loss focused buy in and agreement. And so that's where your conversation might be missing. So you might get somebody who has a. Understandably short sighted approach who has a narrow focus and that person says, look, my job is to answer these customer service calls from unhappy people whose workers compensations claims have been dealt with wrongly.
[00:42:00] And that's what I need to do. So you keep bugging me about how to reduce claims. Don't bug me about that. I need to deal with the phone calls that are coming in from all the angry people who are unhappy about their claims. But somebody who's thinking about profit and loss will say, I could have the size of the call center if we just didn't screw up the claim so much in the first place.
[00:42:16] And we had an app where people could check them. This is valuable. Great. You should go do more of that. So you may get thumped by people at a level where they can't think in profit and loss terms. I've never been thumped. And I wouldn't predict that you would ever be thumped by someone. If you bring them a profit driving initiative, they might not agree with you.
[00:42:34] They might not want to do your initiative. But they're not going to thump you
[00:42:37] Dr Genevieve Hayes: And that is metaphorical thumping because if they actually did thump you, you would
[00:42:41] Douglas Squirrel: that they'd have a worker's compensation
[00:42:43] Dr Genevieve Hayes: then you've gotta work as complicated
[00:42:45] Douglas Squirrel: There we go.
[00:42:46] Dr Genevieve Hayes: Your website mentions that you make use of techniques from action science. Are these action science techniques, or is action science something different?
[00:42:55] Douglas Squirrel: Well so action science is a lot of names for things that people already do and claim they do, but typically do very badly that have to do with conversations. That's a kind of very broad definition off the cuff. That comes from the work of a guy called Chris Argyris. And was a social science research.
[00:43:14] He died a few years ago having written an awful lot of very academic papers on very abstruse and complicated ideas in the kind of overlap psychology, business, topics like that. And nobody could really follow what he was saying. The good news is that he did discover a lot of really useful stuff.
[00:43:31] It's just buried in these academic papers from the 1970s. So, a number of folks have found and kind of unearthed and done archaeology on what he did. And it turns out to be really valuable. For example, Genevieve, do you know about something called test driven development? Is that a familiar concept to you?
[00:43:48] Dr Genevieve Hayes: Is this with software development, how you have all those tests to make sure you're not introducing bugs every time you update something?
[00:43:54] Douglas Squirrel: And even more than that, you do the tests first. And actually, in data science, often you kind of, you can't not do this because you need some kind of ground truth in order to make your models work. But those of us more benighted in the non machine learning data science part of the world have sometimes either skipped our tests or written the code first and then tried to make the tests work against it.
[00:44:16] The value that people get from test driven development, where you write the test first, is you actually know what you're trying to make it do, and you're validating your hypothesis. So, the reason I ask you about that is that there's a version of test driven development for people. And I teach people about this in chapter three of my book, the principle is that it would be really useful if you had frequent feedback in the middle of a conversation, for example, a conversation with your boss about whether you should invest in new pricing models and techniques to help in your workers compensation example and if you had frequent feedback during a conversation Then you could adjust your conversation and you could change how you were having it on the fly in real time as you're actually doing it rather than going away.
[00:44:57] And I'm sure we've all had this experience and you say, I was trying to convince my boss of this boy, she really didn't want to hear it. And, I really should have said X. And you think of that two days later, the French have some term for it. I don't remember the French, but it translates to what you thought of on the stairs.
[00:45:12] It's like you left the party and you thought, oh, this is what I should have said. The thing is, you can actually get to that. You can actually unlock that in your own brain and you can actually get to even better things you might not have thought of even on the stairs during the conversation. But it's a skill.
[00:45:25] Something that you have to practice is something you have to work at. And if you develop that skill, then you have better conversations. So guess what? You can extract more from your boss, more from your customers, discover more ideas more quickly. And that has a tremendous impact on the bottom line.
[00:45:38] Dr Genevieve Hayes: So action science is basically all about having better conversations.
[00:45:43] Douglas Squirrel: And having a mindset that helps you to have the better conversations. So for example we, we're all watching some crazy sort of orange hair guy who has some legal trouble in the United States. You might have heard of this guy and as those cross examinations happen, what you have is lawyers saying things like Mr.
[00:46:00] Trump you had the goal of influencing the election unduly. And, you're doing bad things, aren't you? So that's a very leading question. Very good that lawyers ask that. That leads them to succeed in what they are trying to do, because they're trying to give the jury their point of view.
[00:46:16] They're being good advocates by doing that. It's the worst thing possible for the rest of us to do. The problem is it's all over television. It's, in all the dramas and so on to behave in this way and ask these leading questions. What's much more effective is to change your mindset. And ask genuine questions.
[00:46:30] A genuine question is one that might genuinely change your mind. You might discover something different, but this never happens on television. Right. Oh, yes, Genevieve, you know, I'm wondering about whether we should have more immigration in Australia or less, it's a big, big issue. And so I'm on the less immigration side, but you're on the more immigration side let's talk about it.
[00:46:48] And, you know, at the end of this program I might actually change my mind and say, could I join your party? Genevieve? That never happens, right? People say, I'm right for immigration. That's immigration. This is the point of view. This is what we should do. More boats, fewer boats, whatever. And they never changed their minds.
[00:47:03] But in fact, what you'd like is for both you and your boss say to come to the conclusion that maybe you need not three new models, but only one that would be a change in your position that your boss to say, yes, I'm willing to hire a new data scientist to build that one new model. And that would be a change in her position.
[00:47:19] That's where you'd like to be. That's what you want to create. The lawyer in Trump's trial does not want that. They want exactly the opposite. But that change in mindset and the techniques that you can use to create that and to sort of resonate with the other person and help them to move in that direction to a productive conflict rather than an unproductive conflict, those techniques are things you can learn.
[00:47:39] And that's what action science is all about.
[00:47:41] Dr Genevieve Hayes: Okay. And is elephant carpaccio part of action science or is that something different again?
[00:47:47] Douglas Squirrel: So that's something different. We were talking about it before, but I didn't use the term. So do you know what Carpaccio is?
[00:47:53] Dr Genevieve Hayes: I think it's something you eat.
[00:47:54] Douglas Squirrel: It is something you eat very good and something I have never eaten, but I have to talk about it anyway, because it's the standard term for this. It was come up with by a guy called Alistair Coburn.
[00:48:02] Very, very clever radulist he's been around for a long time. Carpaccio is meat that's so thin that you can hold it up in front of you and you can see through it. That's how thinly sliced it is. And what you want to do is take your elephant, your great big project, whatever it is that you're building, and slice it into extremely thin slices that you can release every day.
[00:48:20] Which is what we were talking about before. So that's a way of describing that process. And the connection to action science is that one of the most challenging things when you are building a process like that, when you're helping a team to go that direction is they don't believe you because engineers all know how to do this.
[00:48:38] Machine learning folks, data scientists also know how to do this. If I said, it. Look there are aliens parked with spaceships parked above every city on the planet. We must release our new workers compensation system every single day. The data science in your team, they would have known how to do that.
[00:48:54] Why aren't they doing that? Because nobody has credibly explained to them why that would be a good idea. And nobody has said to them, you know, I'm happy if it takes you nine months to complete the new system, as long as we get feedback every day. Instead of going away into your ivory tower, sitting in a corner for six months and building a thing that none of us want.
[00:49:10] And you have to be really clear about those kinds of trade offs and the kinds of things that you would have to do the changes to work.
[00:49:16] So you need to have that kind of productive conflict encourage the genuine conversation, ask genuine questions about what would work or what wouldn't work in order to unlock the knowledge that technically exists for producing an elephant carpaccio type process.
[00:49:31] Dr Genevieve Hayes: And the thinner the elephant is, the better, obviously.
[00:49:33] Douglas Squirrel: Well, the thinner the slice is, you can have a very thick elephant, you have a great big elephant. The other important thing about the elephant carpaccio is you have to slice the right direction. Now, this is harder on audio, so I'm going to do my very best.
[00:49:42] But imagine you've got your elephant, your elephant standing on the table, or maybe on the floor. It's probably too heavy for your table. So you've got an elephant standing on your floor. And if you slice horizontally, so if you slice so that you get the feet of the elephant first, it's And then you get another slice that has maybe the underbelly of the elephant.
[00:49:58] You have another one that has the tail and the nose and the the trunk and sort of the middle. If you slice that way, none of your slices look like an elephant. They look like four circles. That's the feet. And then it looks like kind of a big blob. And then at the top, it looks like a little blob.
[00:50:12] But none of it looks like an elephant. And the problem is that people often say, Oh, great. I'm going to release every day. I'm starting with The data cleansing process. And then I'm going to move on to the data analysis feature identification and so on.
[00:50:23] They move through kind of the pipeline that they're going to be performing. That's useless to customers because you go to a customer and say, what do you think of my features? Why do I care about features? I don't even know what features are in data science. You know, why am I interested in that? If you say, look, I'm producing results here that when we analyze them, when we get the neural network driving through and doing gradient descent and so on we're going to get these kinds of results, but here's the first cleanse to data.
[00:50:49] Here is something that I can show you that is meaningful. Then it kind of looks like the elephant. Say, we haven't analyzed this yet, but here's the data that's going in. We've now analyzed it, and we've got some early results. They're not very good, but is this in the right direction?
[00:51:03] That's slicing the other direction. That's slicing vertically from nose to tail, but with the knife moving up and down. And so you always get the silhouette of an elephant. Yeah, it's going to leave out the trunk. Yeah, it's going to leave out the tail for certain slices. But most of them will be elephant shaped.
[00:51:17] And that's what you want in your elephant carpaccio.
[00:51:20] Dr Genevieve Hayes: Yeah. It makes sense. So if you had to choose just one technique for our listeners to get started with that would result in them achieving the greatest bang for their buck, what technique would you recommend?
[00:51:32] Douglas Squirrel: Now, that's a fascinating one. So I'm going to cheat and do two because I already talked about one. So the one that you can do kind of instantly. I mean, it takes a long time to get used to this, but you can understand it straight away. And that's the genuine question. So moving away from leading questions, which are very dramatic and which kind of make your point and try to convince people.
[00:51:52] Into a question like you know, it seems to me that it might be better if we made this report faster and that's something I'm interested in exploring with you. And I'm willing to change my mind. And if you think something else would be better, I'd like to hear it. I'm getting a lot of signals there that if as did actually happen to be in the story I told before.
[00:52:10] The person on the other side says, no, don't make the report faster. That would be terrible. I actually listen. And I'm encouraging them to tell me that rather than leading them down the path of isn't faster, better. Wouldn't you like it to be faster? You know, I've got a great idea for making it faster. So that's 1 method.
[00:52:24] But when I haven't talked about, is the process of joint design. Because what happens far, far, far too frequently, and we've even institutionalized this, we're so insane that we've set up these processes where somebody goes away, has a conversation with users, comes back with a high level specification that's handed to somebody else to make the low level specification, that's handed to somebody else to make the acceptance criteria, by the time engineers see it, it's so processed that, you can't tell where it even started or whose idea it was.
[00:52:51] If we were doing a process of joint design, where we involved all the people in the process. And we had a mechanism that kept it from being inefficient. We had good, strong decision makers. If we had mechanisms for soliciting opposing views, and these are all tools that you use in this process of joint design if we built our software that way, if we built our machine learning models that way.
[00:53:12] We would have much better results much more quickly. In fact so I'd say that technique, is probably the one that could unlock the most for your listeners.
[00:53:21] Dr Genevieve Hayes: So getting all the right people in the same room at the one
[00:53:24] Douglas Squirrel: Without getting stuck, you have to make sure it's very efficient, well run that it involves people in the most important decisions and people at the right level. And you make sure that you're hearing all the points of view. You don't necessarily agree with them, but you're including them in your decision making.
[00:53:38] It's so different from how most folks run organizations because they're afraid of conflict.
[00:53:43] Dr Genevieve Hayes: On that note, what final advice would you give to data scientists looking to create business value from data?
[00:53:50] Douglas Squirrel: I'd encourage them, first of all, to talk to lots of customers and do that even in the face of opposition. So one thing I often tell people that I'm coaching is that it's very helpful to do stuff that creates conflict as long as you do it in an open way and you do it in a way that invites interaction.
[00:54:05] Feedback and information. So if data scientists out there listening to me and we're saying, go talk to more customers, interact with them, they might say, my boss would kill me. I'd be in so much trouble if I went and talked to customers. You don't know that.
[00:54:18] And very rarely do I find somebody who has had a direct order not to talk to customers. Typically, nobody bothers to give such an order 'cause their data scientists aren't very interested in doing that. So, first of all, by going and talking to some customers, you're indicating that there's an interest, whether or not you talk to the right customers or do it the right way, or you annoy some account managers in doing.
[00:54:36] So you're going to create a very interesting conflict. And so, if you'll indulge me, I just have one brief story, would that be okay.
[00:54:43] Dr Genevieve Hayes: Please do.
[00:54:44] Douglas Squirrel: And it will seem unrelated but I promise that it's got a real kicker that's valuable here. So I live, as you mentioned in my 600 year old house. in a tiny, tiny village.
[00:54:54] My main neighbors are 700 sheep, but I also have 18 or so neighbors in eight houses along one road. We don't even know what the name of this road is, because it's just the road. It's the road in Froggle. And this road ends in a turnaround. It used to go through, but it's been stopped up.
[00:55:08] It's been made into a turning circle. And so at the end of our road there's this place to turn around now on a Friday night, we sometimes get people who come there for various we might say slightly shady purposes. They want to smoke certain things, maybe they're not supposed to, they want to hang out with their chosen beloved person and we're perfectly fine with that in the neighborhood.
[00:55:27] They don't cause any trouble, they leave, they're okay. We don't mind that they use our dark turning circle for this purpose, but there are some people who come and they stay. And they park like several campers there. They block the whole thing. So the sheep farmer can't turn around his tractor. They make noise, they have parties and barbecues and things in the forest, which is dry in the summer, so they cause a lot of trouble.
[00:55:46] And so when this started happening, when I moved to this village, I went around to all my neighbors and I said, who owns that land? We got to do something about this. This is really a problem. They all check their deeds and they went, well, I don't seem to own that land. And it seems like whoever stopped the road never bothered to sort of resolve who owns stuff around it.
[00:56:04] And probably somebody did, but there's no record of it. So they said, well, I can't really do anything about it. I don't own it. And so I said, I'm going to do something about it. It's very clear. I don't own it. I live a long way from it. But nonetheless, I hired somebody who's good at putting up fences.
[00:56:17] And I said, build a fence over there and put up some signs on it to say private land, no trespassing. I have clearly no right to do this. This is clearly not my land, but I did it anyway. Now, one of two good things will happen as a result of this. Either someone will turn up and say, Hey squirrel, this is my land.
[00:56:33] What are you doing on my land? This is terrible. You know, get off my land. I'll say, what do you like a free fence? Now they might want the free fence or not, but I'm going to get their phone number. So when these people who cause a lot of trouble turn up, I can phone and say, Hey, you know, that land you own, what can we do about it?
[00:56:47] And the other good thing that could happen is nobody ever turns up. Nobody ever says, Hey, this is my land. That's fine. Cause I will keep booting people off it. And eventually I'll go to court in a process called adverse possession. Also known as squatter's rights.
[00:57:00] And I'll go there and I'll say, Hey, I've been taking care of this land. No one else claims it. It's my land and the court will give it to me. And the result in either case is the problem of these folks who come and camp in our village will, will end. Now, the point of the story and the reason I tell it about your listeners perhaps talking to more customers is if you go talk to a customer, one of two good things will happen.
[00:57:18] Somebody will say, Hey, I'm in charge of customers and you're not supposed to talk to that customer. You say, great, you're in charge of customers. Where are some more I can talk to, right? You're now having a productive conflict because you've done it openly, right? You know, I put up the signs I didn't do it secretly.
[00:57:32] I made sure people knew about you should make sure people know so they can come and object or nobody will ever say it. They'll say, well, it's you know, Genevieve's off talking to customers. Boy, that's strange. And then you can point to things that were really good results that you got from it. And maybe you'll be in charge of the data scientist customer contact program.
[00:57:48] So I'd encourage your listeners to create productive conflict in that area and all the others we've been discussing. And as long as you make sure it's productive conflict, that's the fastest way to resolve challenges like we keep building stuff that nobody wants.
[00:58:01] Dr Genevieve Hayes: One thing I've found is often if you go to someone and say, I have this thing, I would like feedback on it. They'll tell you, no, I'm too busy. Go away.
[00:58:11] Douglas Squirrel: Yeah, don't do that.
[00:58:12] Dr Genevieve Hayes: But if you go and say, I have this really fantastic thing that I think will help you, then they'll often say, well, I have a few pieces of feedback on that.
[00:58:20] Would you like to hear them? And it's like, if I asked for feedback. You wouldn't give it to me, but I'm not asking for feedback and you're volunteering it and it's incredible how people work like that.
[00:58:30] Douglas Squirrel: Because and this is absolutely straight out of action science. In the first case, you're imposing your will. You're using what Argyris would call unilateral control. You're saying this is the way I've produced it. I want your feedback. I want this for me. Whereas if you do it in a, in a mutual learning way, you say, I'd like to learn from this, and I think it will help you, and it will solve a problem that you have identified and told me about.
[00:58:54] This is why it's so valuable to talk to the customers, because the customers will tell you about actual problems they have. Oh my God, this thing just never produces the right graph. I'm sure the data is in there, but I can't get it out. And then you'll discover you haven't got a model problem, you've got a visualization problem.
[00:59:08] And that's what you can then take back to them and you say, I have things that might help you to visualize better. So that is absolutely a fundamental core principle that you've alluded to and described very well there. And you can apply that all over the place to get much better responses from people if you use their language and talk about their problems.
[00:59:27] But first, you have to meet them. First, you have to understand what their problems are.
[00:59:30] Dr Genevieve Hayes: And one of the things I've found is, it's what you said before, if you go in and, You want whatever you produce to be perfect and you get upset when someone gives you feedback, then that's not a productive situation.
[00:59:45] Douglas Squirrel: Well, and it wouldn't be perfect anyway, because you're building it for them, not for you.
[00:59:48] Dr Genevieve Hayes: Yeah, exactly.
[00:59:49] Douglas Squirrel: So perfect in your view,
[00:59:50] Dr Genevieve Hayes: Yes. But if you go in and once they start giving you feedback, you sit there and start scribbling down like crazy, they'll actually give you more feedback because, Hey, this person's actually listening to me.
[01:00:02] Douglas Squirrel: and what this person is doing meets my needs and is in my language. And that's what I think a lot of data scientists miss, and that's where we contribute to this kind of mystique of the wizards in the tower who are somehow speakers of a different tongue, and none of the rest of us can understand them.
[01:00:18] If we were speaking in customer language, and we described what we did in terms that were meaningful about problems that are important to the people we're talking to, we'd have much better results.
[01:00:28] Dr Genevieve Hayes: So for listeners who want to learn more about you or get in touch, what can they do?
[01:00:33] Douglas Squirrel: Well, there's really two places to go. The primary one is douglassquirrel. com, so if you remember my name, you can always find me. My email is there, information about my book, videos and articles I've written, all kinds of good things, and of course you do have to spell Douglas and Squirrel correctly, but luckily they're all spelled the standard way, so this shouldn't be very hard to do.
[01:00:56] And I'm not a small rodent who lives in the Pacific Northwest of the United States, so as long as you can differentiate those two things when you google. You'll be just fine. But there's another place you can also go, which is squirrelsquadron. com. And that's my community of tech and non tech people. I get thousands of people together to discuss and debate these issues and discover new things.
[01:01:14] And we're looking into all kinds of crazy new ideas from Dave Snowden right now, for example. And there's a weekly newsletter, which I know you've been reading Genevieve. So that's a great place to start as well. So if you just want to hear a little bit more from you periodically, go to the squirrel squadron.
[01:01:28] com join up and I'll write to you every Monday and you can also participate in other things like my forum and my free events that I run every week.
[01:01:35] Dr Genevieve Hayes: I'll also add, as you said, I read your newsletter and I recommend that. But even if. None of this appeals to you. I recommend you check out Squirrel's website because it has the most beautiful artwork I have ever seen on any website.
[01:01:51] Douglas Squirrel: Which I get no credit whatsoever for, except that I happen to know the designer who does a wonderful piece of work for me. And I decided to make my website different from every other consultant, because I wanted to stand out and I wanted people to know that I'm all about these conversations and productive conflict and things like that.
[01:02:06] And not wearing a suit and tie and giving you a two by two chart with the solution to the world's problems on it.
[01:02:12] Dr Genevieve Hayes: Well, you will never see a website like this. It is gorgeous. even if it's just to check it out and see the pretty pictures, go and have a look at it.
[01:02:21] Douglas Squirrel: Fantastic. DouglasSquirrel. com, I'd welcome you for any of the purposes Genevieve just described.
[01:02:25] Dr Genevieve Hayes: So squirrel, thank you for joining me today.
[01:02:28] Douglas Squirrel: Oh, it's been fun!
[01:02:30] Dr Genevieve Hayes: And for those in the audience, thank you for listening. I'm Dr. Genevieve Hayes, and this has been value driven data science brought to you by Genevieve Hayes consulting.