What a learning curve it was listening to Roger Grobler from Ethos AI.
I will write a separate blog on the key points, but until then here is the video of the interview and the transcript.
Colin Iles, Roger Grobler, Shiona Blundell
Transcript generated using otter.ai
Colin Iles 00:15
All right. Good afternoon, everyone. And thanks again for joining us. As always, I'm going to hand over to Sybrin in a moment, just to give us a little Kickstart to give the intros. And then we'll kick off with the discussions with Roger, I'm really looking forward to this one, I'm not going to say too much. Now, we'll come on with that in a minute, just while we're waiting for attendees to get in and well done for getting past the power cuts. And if you're in Bryanston, we've also seemed to have a data cut as well, which is not particularly useful for a call like this. So well done for getting through that. I just want to remind you, though, please do put your questions in. And if you've got questions, put them in sooner rather than later. We always try to finish punctually on the hour. So if you get them in early, we're more likely to get them if you're putting them in in the last 10 or 15 minutes and we have a big rush. Well, we're not going to go and manage that. So with that Shiona, why don't you kick us off?
Shiona Blundell 01:06
Thanks, Colin. Hi, everyone. Thank you for joining another installment in our game changers series. I'm Shona Blundell, the executive head of sales at Sybrin and I'm standing in for our CEO Marius today. Joining Colin today for our session is Roger Grobler. Roger is a partner and co founder of the Ethos AI fund, and he has extensive knowledge and experience in insurance data, artificial intelligence and private equity is helped numerous companies including Crossfin, one of our new shareholders Tymebank and Optasia formerly known as Channel VAS amongst others. Working with their management teams he has helped to improve their operations through the use of AI. And I'm super excited about today's topic, Colin, the art of science and data lead decisioning and how that influences and impacts business and I'm really curious to hear your views. So over to you, gentleman.
Colin Iles 01:58
Shiona. Thank you very much Roger. Welcome.
Roger Grobler 01:48
Thanks. Colin. Thanks. Now,
where are you dialing into from today?
That's why you've I'm actually thinking about putting a jumper and a coat on up here in Joburg, because it's freezing I guess you can get away with a T shirt down there. Anyway, let's let's start I don't want to go and spend all the time talking about Stellenbosch just gets me incredibly jealous. Ethos AI what is Ethos AI?
Roger Grobler 02:29
Currently, we my career has been spent in working with businesses that uses algorithmic and data driven decision making in insurance in banking in telco. And we got to the point where so the algorithmic decision making started to be called AI because the machine start learning themselves and adapting themselves. And we thought about how do we invest behind the AI as a, as a team, it's quite a powerful theme. And our first conclusion was that we didn't want to invest in the tools of AI. So the people actually make the tools that you use to deploy the AI, but rather the companies that use the AI, in my experience, what we've seen, particularly with value chain where you can deploy algorithmic decision making in multiple places in the value chain, you get an uplift, and then a multiplier effect on that you can get significant increases in the value of that business. So our Fund invests in sort of normal businesses, if you like. And what we've also learned in the process is that it might be normal businesses, but they've got quite disruptive business models, that it's let's say, it's in insurance or in telco and medical technology, or whatever the case maybe. And then we assess those companies on their AI journey, amongst other things. So how do you think about AI, where do you deploy it? How do you go about building a data science team? What vendors do you use, how do you engage with them? And then help them go through that go through that journey. So we've made a number of investments to date. And I think we've been quite lucky with lock down because types of businesses that we invest in actually did quite well through locked down because they're very data and technology driven. And and we look forward to doing that for many years to come.
Colin Iles 04:29
Okay. And then just for people on the call some of the companies that you've invested in that I recognize Tyme, for example, maybe Crossfin , and people won't know but some of the CrossFit is portfolio, they might. In terms of the merchant acquiring, Ikokha, Aduma, Retail Capital. And then what else have we got in that in that portfolio? I suppose. Actually, that's not in the portfolio. That's something that you're looking at separate. I want to come back to that. But I can clearly see how Tyme for example, is going to really benefit from this AI approach. I mean, they've now got five 6 million customers. T'hey are probably the fastest growing bank certainly in Africa, if not the world, the way that they're recruiting new customers. So their opportunity to go and use data to actually go and enhance that services must be massive.
I think that the, what's quite important to understand is that you've got a business that does banking, for example, in Tymes example, we've got a medical technology, business, called Vertice, or rain lives in the telco space. So the space in the industry is quite normal. Then you've got a business like Tyme that's got very new technology. So it's banking systems, it's all sort of cloud driven, highly connected, very flexible. You've got data, in terms of having access to data that other people potentially don't have any ability to use that data. But then very importantly, you've got a culture inside the business that can actually use the technology and the data to disrupt their market. And that then leads to sort of a disruptive business model. So what's attractive about time, it's a hybrid digital bank. So it's not a pure digital bank in that you've got a presence in supermarkets you can pay in the Foschini . But that lowers the cost of delivery significantly for the bank, and they can get to a much lower cost to income ratio, which puts them in a in a competitive position.
Colin Iles 06:31
How did you get to this space? Let's just jump back in time now because you know that the message is clear. And I think we're going to go into this in a bit more detail. Data led, AI driven organizations, you're inferring have a massive advantage. And certainly in some of those examples, I can't really dispute that. But where did your journey start? I mean, this is we're talking in the 90s. When you started to get interested in data and thinking about this in terms of a career, what, what was that journey?
I studied actuarial science, and we were still at the university, doing some postgrad work and an insurance business came to the university and said that they have got a serious problem, they've got a insurance license from Lloyds, but they're running at very low, very high loss ratios. So loss ratio of insurance companies, the percentage of your premiums that you pay out relative to the in claims relative to the premiums that you get, and they we're running at like a loss ratio of like, I don't know, 140%. So they get one rand in, they pay out a rand fourty and that's not not sustainable. And they asked us to help them to develop a pricing strategy. So the pricing at that stage, it's hard to believe today, but insurance, you'd have your guy car insurance pricing is printed on a piece of paper, it basically says whatever the value of your car is, that's the that's the premium. And they couldn't charge more because they were already at mark. So we took all of their data and basically developed the first individualized pricing model in South Africa. So look at people's age, the gender where they live, what make and model of car they drive, and so on. And they were quite brave to actually take our model implement it immediately, and in South Africa you have got monthly contracts. So some people's premiums doubled, others, others went down. They last half their book overnight, because lots of opinions went up. A year later, they were at the loss ratio that we modeled, and they were double the size of where they were when we started and we sort of fell into this silver bullet, me and me and my partner, Willem Roos and Willem Roos then went on to be the founder of Outsurance. They went to Australia and created Youi. He Rain for a period of time. And similarly I've had a career using algorithmic decision making.
So that must have been quite exciting when you know 12 months later, you can look and you can show with evidence, the impact and the benefit you've brought. So in you know, to an industry to an organization at scale was that what got you excited and made you want to move into the insurance sector
Well we were suddenly very popular because we had the silver bullet so we had job offers suddenly from lots of different insurance companies. I went to join Hollard and instead of you know I have technology challenges because how do you actually get this pricing out. So we had this wonderful programmer called Tony Mollens. I asked him what languages does he can he program in and he used to like slam his keyboard and says all of them are all the same. So you wrote these black boxes that we then loaded them on to actual stiffy disks. And these stiffy desks will literally be delivered 100 of them every month to all the different places where we need to deploy this pricing model. So it's a black box with other stuffy desk. And then the pricing got deployed and worked remarkably well. And if you've got that kind of thing, then gotta take advantage over your over your competitor. Fast forward to today in some of the businesses in Australia that I was involved with, you would either get online and start your quote or you would call the call center. And by the time that you've given them all the information that they require, that has gone off and looked into banking transactions with your supermarket transactions and got adjusted out of that for your premium based on what they're seeing and what your what your risk looks like, plus those models, self update, so as claims come in they actually adjust themselves. It's become enormously sophisticated now using very wide and big data sets.
Colin Iles 10:54
How do you get the senior management? I mean, you mentioned a word that comes up a lot black box or two words, I suppose one or two, don't know, doesn't matter? How did you get the senior leadership team at Hollard to accept. So you've, you've created this black box information comes in pricing comes out, you get the 100 USB sticks and send them around so that people can actually use it, you're now in production, if you've got this wrong, that's quite a significant risk. coming through, I just think about other experiences I've had in financial services, it's not easy, getting senior leaders and compliance teams and other risk monitoring teams to get behind new technologies. They want to understand it in detail. And it doesn't sound like this black box is something that those guys were going to understand at that point.
When today we are investors. And one of the key things that we look at before we would invest in a company is whether the management team is actually up to embracing these kinds of technologies. And unfortunately, there are management teams that are not not up to it, and are not going to evolve and evolve with us. And if you didn't have a disruptive business model that comes along, that company will really suffer. So if you talk about data driven decision making and AI and algorithms, the very first thing you need to get right is the company's culture. And that does start from senior leadership all the way all the way down. The technology, the data is not going to get you there if you don't have a right culture inside the business.
Colin Iles 12:21
We're gonna get into that in a minute. I want to know a bit more about your journey, though. So I think the first one that really stood out for me where you were making a massive difference, and this time you're taking the credit at least publicly because you are the throat to choke was with real insurance in Australia. How did you get into that? And as you go through the success and how you built this out, be really interesting to hear about some of the products that you actually prototyped out there, because they were quite novel, they still are quite novel.
Yeah. I mean, I had an extremely steep learning curve going to Australia, I came out of Rand merchant bank, did financial engineering and securitization, all kinds of complex things and went to I had a yearning to be a entrepreneur. And it was extremely steep learning curve, it's very hard to come out of a large organizations suddenly stepping into a team that's only got 14 people. And I think what we saw the things that we learned, we went down into car insurance, for example, in in Australia with very, very strong competition and really struggled with that. But we also went into life insurance, we got a very talented guy to run our life insurance business. And that was quite an uncontested market. So I think competition is quite a quite a big thing to avoid. Competition can make a very lucrative opportunity, not lucrative, because there's too much there's too much competition. So in our life, I'll speak about the life insurance business first, and then we can talk about the innovation on the car insurance side, but on the life insurance side. And a period of I think about two and a half or three years, we went from a standing start to selling the most life insurance policies in in Australia. And the reason basically, for that is we had a extremely efficient sales process, all the way through our call center, we had very good systems. And at every step of the way, we were better than another than the rest of the market. And at one point, the biggest player in the market actually came to us and they wanted to outsource the direct operation to us. And we did a proposal for them. And when we looked at it, the embedded value of the policy on our side was six times higher than theres and better value is basically the discounted future cash flows of a policy. And I thought somebody made a mistake. Initially, when you unpack it, you saw that at every step of the way, we would like five or 10% better than they were, but these things multiply out. So we had a lower cost per lead slightly higher and better conversion rates, higher ticket size, better underwriting and, importantly, better retention. And lets say you make 5% profit on a life insurance policy, we can get all these multipliers out Suddenly, we had a significantly more profitable product. And then because we had more profitable products, we could spend so much more on direct advertising than anybody else, and still be profitable for for us. So I hear many years later from Google themselves that for a number of years, we actually the company in Australia that spent the most on Google of any company, including banks and retailers and all of that. So my lesson out of that was the extreme power of a multiplicative value chain, the other product,
Colin Iles 15:45
Why were you able to do that Roger, and the incumbents weren't, because as you said, this was new for you, you're, you've come out of a corporate and now you're the CEO / founder and having to worry about everything, you've gone into a very competitive market in life insurance. And yet somehow, you've managed to put a process in place where you're getting this multiplicative benefit, which is outstripping others. What do you put that down to? Was it luck? Or was there some other.., I don't think so? Yeah,
I think that on a market construct perspective, the six large life insurance players in the market and they all had traditional distribution channels. So they distributed through through brokers and intermediaries. And if that is what you do, you, it's really difficult for you to go direct, because you alienate your distribution channel. So there's only one other entity in the market that actually had a direct operation. So there's very little competition in the direct distribution space. And then we had like, just incredible, incredible people, the guy that we had that ran that business, he actually built the systems himself. Incredibly incisive data driven thinker. So you've got these algorithms, but you also look at the process and you deeply understand the process. If you think about about AI, and about data driven decision making, it's the data and technology, yes. But you need to have people that can actually see through the noise, find the signal, and then pull where the big levers were, and are leader in that business was exceptionally exceptionally good at that.
Colin Iles 17:24
Let's get let's jump into that. Because I think that's, you know, really interesting. So, we've got this, I mean, just looking at your career and your history, you can see just how you can use data to go and generate super profits and outstrip the competition, even in areas which are, you know, relatively competitive. How do you find these data you know guys, these people that have got this ability to go and put algorithms put data components in place, but not just doing from an engineering perspective, doing it from a business perspective, and doing in a way which can then be exploded to actually service these customers? How do you do that? Where are these people, I've never met one?
I think there's a lot of people that can learn how to do this. If you change the culture, to a culture where you use data to make decisions, and instill that in operating procedures of the company. So one of the things that we do quite frequently with businesses is have a performance meeting. And it's a weekly meeting, where the senior leadership dials into a call to take the key performance figures of the business. And you discuss it for 45 minutes to an hour. And what I've seen in those kinds of meetings, because it's all sort of number driven as to what do the sales funnels look like, what is the performance levels look like? How did it change from this week to next week, there's a few fascinating things that that happen. The conversation levels go up, and the use of language goes up and understanding of metrics go up. And then in other meetings, you start also seeing as a conversion rate or retention rate is important thing you start seeping up into the into the business. And the second thing that's interesting that happens is that reporting evolves. So you'll be looking at a report in this meeting, and somebody said, Listen, if we break up those two channels, when we look at the segment differently, isn't there something going on there, and then the reporting, guys, by the next meeting in the next week, they would have evolved, that set of reporting. So I think that's a little bit of a long way to say that there are processes and things that you can put in place inside of a business, and you can change the language and the discourse in the business. And it's not necessarily to say listen, there are certain types of people that are naturally sort of data driven, you need to find those people. I think you can actually change a culture with sort of, for lack of a better word for normal people to have a data driven, data driven process, but that does require a bit of leadership.
I mean, you said that this lots of companies out there, which presumably have got massive amounts of potential, but you won't invest in just because the leadership is what is old fashioned, I suppose they're not particularly data centric. What sort of markers are you looking out for when you're talking to them?
Roger Grobler 20:25
I think the unfortunate challenge with successful old businesses is that they are successful old businesses. So they've got these business models that's worked for them forever and a day. You've got managers that sit there, and you sort of start challenging the goose that lays the golden egg at your at your peril. That's quite a dangerous and brave career moves to make. And therefore, those companies take a long time to change if new disruptive business models come along. And that is, that is the opportunity for new businesses that that do come along. One of the things that's quite important is first principle thinking. So if you step back from from a problem, and you look at a problem with a sort of first principles, then data, data becomes a lot more important. So where are these levers that we pull? What what fundamentally is it that we need to that we need to do here? And then to be honest about your existing business models. And do you need to evolve these business models. One of the most famous examples in the world would be Netflix. To evolve to fundamentally different business models to go from a DVD mailing service to a streaming service, now a content production service and creating massive, massive value through that process. And they found a Read Hastings. Netflix was his second business. And one of the things that he wanted to change from his first business was to have first principle thinkers. Elon Musk is also very famous for placing a high, placing a high amount of value on on first principles. Our team in Rain, our Rain telco, they are incredible with first principle thinking with looking at where the future is gonna evolve, how it's gonna evolve, very strong engineers, very data driven, and then make extremely, I think, robust decisions. That's, that's, that's data drawn. Hmm.
Colin Iles 22:24
So first person, first principle thinking, is, is coming through what else are you seeing in the organizations that you've run successfully, and the ones that your invested in? Are there consistencies in terms of, I guess, attributes or styles of leadership that you can sort of nail your flag on and say, they are significantly increasing their odds of actually using data successfully here.
They're all quite good with technology, I must say. They all have reasonably sort of modern and good systems, and they've got the ability to wield that technology. Very well. And that's, that's not a that's not a trivial thing to do.
And is that typically cloud, this is typically, you know, they've made the shift, they've put a lot of care and consideration to how their data is stored and managed.
Yes, yes, yes. In a big way. I mean, we ta ke, I mean, rain and Vertice, and Tyme, they are all all the data are sort of cloud based. It's available throughout the, throughout the business. There's one source of of truth. They've got very strong data science teams that that work on the data. I mean, the one, the one aspect of all of this is that it is actually quite hard to build a data science team, the data scientists are very similar to senior software developers, a very specific breed. And you should one of the mistakes I've made when I went to Australia is I didn't carefully think through through the employment proposition for the data scientists. So you need to be very careful as to how do you shape that employment proposition so that you can attract high quality people and then put them to work in a productive way. And part of that work, setting up infrastructure, which is, which is not trivial. Infrastructure would be the cloud cloud setup and the data lakes and the tools that you use to analyze the data and and then also, how you deploy it inside the inside the organization, and the deployment inside the organization then depends on your managers inside the organization to be data driven. To actually embrace it.
Colin Iles 24:37
Do we have those types of skills in South Africa to do this? Or is this something we're now really going to have to unfortunately, look overseas and pay premium dollar or pound or euro rates for?
South Africans are very, very innovative and very scrappy. It's one thing that we'd sort of speculate as to as to why that is, so we've got a very large number of very scrappy, very talented data scientist consulting firms. So that's one thing that you can do. There's people that are actually very good at building a wide range of things. If you are a company not going on this route, the first quite difficult question that you got to answer is do good data scientists want to work for me? And the mistake that I made firstly, in Australia, I tried to create a data science team, that my data science team just wasn't very good and took years to realize that they didn't want to work for a two-bit insurance company. So I eventually got a very talented consulting firm, and they came to work inside our business. With Rain, I think we are fortunate in that it's an credibly attractive place for data scientists to work so we can track very high quality data scientists to work inside of rain. So even our business, that's the first question that you need to answer, let's say, let's say you're struggling to attract high quality data scientists, you're gonna have to find a way to work with a consulting firm to come in and sort of establish your models and help you build what needs to be built. And then maybe you can have internal team that does the does the maintenance around that. But I think the key thing is to start with the employement proposition. What is it that you can offer these people and it's not it's not only money, the money needs to be okay. But it's who they are going to work on, what kind of problems if they're going to work on whats the infrastructure they're going to work on are they going to be empowered? Who's going to..are they going to be listened two inside the business? Those are the decisions that needs to be made.
Colin Iles 26:39
Yeah I see, this question is actually just coming in from the audience there. I mean, is that Is it as simple as that though, I mean, if I'm running a really boring financial services company, I'm just never gonna get good quality data scientists, because I can't do these things, I can pay them, but I can't give them interesting problems. I can't give them things that are going to, you know, wow, them, I'm actually essentially going to have to go down the consultancy route.
Okay, you now use the example of a financial services business, if you're going to use technology and data science in your financial services, and if you have the right culture, then you're probably going to do some disruptive things through the use of this data. That's an interesting problem. Therefore, you can attract data data scientists. I think where it's harder is let's say you've got a manufacturing business. And you've got a whole bunch of plant with a bunch of equipment that manufactures things. And there's not that many opportunities to deploy data science, but there is very specific instances, then then I would go to a consulting firm, when an example is there's a consulting firm in Cape Town, they specialize in manufacturing. And it sounds very obvious in retrospect. But if you have manufacturing equipment, that's been many that's been sort of bought anywhere in the last 15 to 20 years, there's a significant number of sensors on this manufacturing equipment. And how each one of those sensors work is they've got a tolerance range between the humidity or the acidity or the temperature, whatever it goes out of whack, and the red light goes off and everything stops. And you've got to fix that instance, first. But this data scientists firm did this, they took all of that data, and they look at the interactions of all of those things. And it turns out, there's many instances where the temperature still in range, facilities still in range, and humidity is still in range. But that particular combination of those things don't work. And then there's a problem. And a manganese manufacturing plant or processing plant, they brought the error rate down from 5% to 0.5%, massive, massive difference on existing data that sits there that they that you basically just needed to go and look at what the correlations were. I thought that's quite a fascinating example.
Colin Iles 28:58
Okay. I think it was, when we chatted a week or two ago, we were going through some of the other I'm gonna call it the cheat sheet, the cheat sheet for people that want to start actually using data in their organizations. You talk about talent multipliers, and I thought, this is fascinating. You want to just go into the details on that.
There's two things. There's talent multiplier and talent density, which has to do that I absolutely love the first time I got around the term talent multiplier was a book by Laszlo Bock, called Work Rules. Laszlo Bock was the head of people at Google for something like 15 years from the very early days. And I think this book is an absolute gift. He just in a very raw fashion, wrote down everything that they learned at Google through all the years, in terms of HR and how you manage people how you recruit and all of those kinds of things. And one of their belief is that a really talented software engineer is worth 100 average, average engineers, and I think, for data scientists a really talented data scientists you can't replace with 1000 Mediocre data scientists. So If you're going to employ these kind of people, you really need to go after the talent of getting the best people you can get and if you can't get the best people because you can't employ them, because they don't want to work for you you have got to insource that from a from a vendor with a vendors got an employement proposition that they do want to work on, because maybe there's more things to work on. So that's the that's the talent multiplier. And I mean, we're looking at one of our companies the other day, and lets say there is a budget for 90,000 Rand, and there was a guy on the market who is 110,000 Rand, if that guy is going to take you from average, to really outstanding that 20,000 Rand is absolutely meaningless, you should get the guy that the talent multiplier is very strong. And the talent density thing is, is, I think, an even more interesting phenomenon. The first time we're going to talk about that was Read Hastings at Netflix. They had a rocky patch at some point, in the early days of Netflix and they had to retrain a bunch of people. And then suddenly, after he retrenched these people, Netflix just rocketed and his reflection on it was in the retrenchment processes at we wanted to retrench and implicitly retrench the people with lower talent. And what remained was higher talent density, so more and more talent on average per person. One of the businesses that I'm invested in had a similar thing with lockdown where we had to let a whole bunch of people go at like half the size today they are more profitable than they ever were. And I strongly believe it's the talent density that has gone up. So I think recruitment is important. But it's also very important to go and have a look at what's your talent per seat, and make sure you don't have and this sounds very machevellion but make sure you don't have seats in your company that's wasted by people of average talent in those seats.
So I can see now where you're going when you were saying there's lots of companies with leadership teams that you'd best avoid, because this is going to be really tough. For traditional leaders, you want to go and minimize the talent pool. So keep it lean, I suppose keep it mean, you're looking for specific individuals that can be multiplicative, in what they actually do 10x 100x. And presumably, you've got to pay them way outside the normal standards and audits that a Hayes market survey or an HR team would normally be comfortable for so I can imagine, that's you're talking about some properly committed leaders to actually go through this. You also say that you need Mavericks. And to Chris's point, yeah, if you can repeat the book name as well before I forget to ask that one because I'm sure there's one to look at that.
It's posted in the chat. The writers Laszlo Bock and books name is Work Rules. An exceptional piece of writing to read. Very, very worthwhile on on talent management. And that I mean, we are at the AI Fund we spend a lot of time on helping companies with talent management. I think one of my beliefs is that recruitment is the single most important thing that a manager does and a CEO does. It's more important even than your than your work and what the company actually does, because if you don't recruit well, then you're gonna have low talent density and then you saw I'm struggling it's gonna be very difficult to succeed.
Colin Iles 33:39
Now, what's this that you need Mavericks in the team as well?
There are many things that are quite specific to Rain so at Rain in the Rain credos that we, we embrace Mavericks, I'm not sure that's applicable to every to every business, but a maverick , the definition of a maverick I think is an unorthodox person, that's a person that thinks differently and they, and they might, might even be difficult. so at Rain, we say that we, there's lots of times we have mavericks thats people who are potentially quite difficult to work with. They're quite quirky. And we actually actually embrace them. I think that's, that's one part of it. The other part, which I believe quite strongly is that you shouldn't define seniority by management. So if you've got somebody that's really strong and doing something, their career path should not be now managing people, if they're not good at managing people, managing people is just a skill set like anything else. And I think if you're in a business and there's individual contributors, like engineer that just codes and he earns more than his manager, that's a very healthy, it's a very healthy sign. The moment you get into business where the only the only path is sort of climbing up the corporate ladder in terms of hierarchy. You'd have heard that thing, the Peter Principle, people get promoted until getting a position that they are incompetent to peform and then they just sit there. It's a very big waste to take individual contributors thats really strong and to make them manage people and they are not good with it.
All right, I'm gonna get told off if I don't start asking some of these questions that we've got coming in now. So let's, let's kick off, Rob Jackson says, How do you convince CEOs to invest in their data for competitive advantage? I think he's making a big generalism on the next one. Most techies are not equipped to do that. I'm sure there are techies who are going to argue against that. But how do you do that?
I've got maybe a bit of a cynical answer on that, is, you pitch it, and if you don't succeed, then you change companies. If you want to be in the space, you're gonna have to be in a gap in a company that inspires you, where the leadership inspires you where the leadership sort of leads the process. I mean, I was lucky in my career. But I, I look back at lots of the things that we achieved. It was due to very strong leadership above me that that drove those changes, I spent a time at First Rand and Johan Burger and Paul Harris in particular, pushed through what was at the time very brave quantitative changes inside of the business done by very young people. And that's, that's leadership. And, and maybe the cynical point is try if you can't do it, go and find a CEO that can do it and wants to do it?
Colin Iles 36:39
Question from Tony, how do you identify people with this high talent multiple, I'm assuming it's just not obvious from the CV and the resume.
So in a lot of businesses that we are involved with, we've got extreme recruitment processes. So the one part that I'm not, I think we still lacking was the sourcing side, so how do we actually get those people, but once they are in our recruitment process, then we would have a whole bunch of steps, you'd have a, what we call a sort of a culture questionnaire, so somebody actually fills out a answers set to probably about 35 different questions that test them on various cultural aspects, so we ask them, what's the biggest failure that I've had, or who's the greatest person they've worked with, and you get a sense of where they'll fit into your culture. That's, that's very important. I'm not a very big believer on interviews. But we do have interviews. And we have, I think, at least four interviews to get a sort of a proper view on a person. And then we design tests. So a person will come in, and there'll be a host of things that they will do, that's quite specific to their job. One of them is an inbox exercise. So they'll sit down at a computer, and they'll have all bunch of emails that they need to respond too. And they don't have enough time to get to all of them. And it's a four hour exercise, and need to deal with everything from client problems to in our analytics business, which is quite amazing as we had this 80 Page deck PowerPoint deck that they needed to fix. And what you literally did afterwards just counted the number of fixes that they did. So what would have been wrong on this thing that would be let's say, scales on graphs, that's wrong, would it be font problems, or they'd be just missed missed spellings and those kind of things. And what was amazing was how high the correlation was between the number of fixes somebody did and how capable they ultimately, ultimately were. And then, since a very, very robust recruitment process, and then a very strong realization that a false positive does a lot more damage than a false negative. So if you appoint the wrong person, you're in for hell. And everybody will know that like for nine months plus as it will take you a long time to actually get to the point where you make peace with the fact that you now appointed the wrong person. He is occupying this seat and is sort of destorying the work of the people around them. And then you've got to get to exit that that person. So your recruiting process needs to be very, very strong at avoiding false positives. And you lose a few good people in the process because your recruitment process is too severe. That's not, that's not not not so bad.
What do you do with them when they arrive? So you've gotten to this huge amount of effort to filter and to pull in the best? Now how do you make sure over those first three to six months, that they blossom that you really learn what they're about?
I've got a little bit of a simplistic view, it feels to me like the speed that somebody arrives in organization has got a big impact on setting their speed for the for the next period, so very keen on a very good lands process. If the person lands they properly inducted into the organization, they've got the tools that they need to work with the are empowered quite quickly and they, and they get running quite quickly. Well I'm all for over ping pong tables, and coffee, and all those kind of things, and that makes the workplace good. But ultimately, people want to work with great other people and want to work on problems that's worth solving. And they want to learn and be empowered to do their own work and to get them quite quickly into doing meaningful work.
Colin Iles 40:42
You mentioned credo, a little while back. So a charter a credo six to 10 points, piece of paper, you can wave around with a list of things in terms of what you want for the characteristics of your staff. Is this a bit gimmicky? Or do you believe in it, you think it works?
I think I think values are gimmicky. So those are the three values, we are trustworthy, innovative, and fun, or whatever the whatever those values are, and if you can look at the likes of everybody from Steinoff to Enron, you can look at they've got value statements, it's it's not like they lived those, those value statements. So what we've done in about 12 organizations that we are invested in is help them develop a credo as a sort of statement of beliefs. Some of them are quite long. I think one of them has about 36 different statements and statements and I mean, every every what's amazing to me is that you'll have different companies and and that are very successful with very different cultures you must realize there's no magic culture. The culture that you've got inside of your organization is right for your people doing your your kind of work. But what we then do in that credos, too write down how we believe what is the best way of doing something. And it's everything from let's say, you believe this would be a single, single source of truth on data, then that might be a belief that you've got, and then write down what what that actually means inside the organization. So what the credo then allows you to do, let's say, you're a new person coming into the organization, Colin, and I'm going to work now with you. I know what to expect from you, because I know now you signed up for the credo in the recruitment process, you've actually read the credo, you've been inducted into the credo. So there's a certain set of expectations. So there's a certain set of behavior that I expect from you, which means I immediately trust you more than that it's more smooth between us. And in the absence of a credo, les say you and I are coming at a problem from two different perspectives because that's what we believe the right way of coming at it. In the absence of a credo, we have an interpersonal conflict. With a credo we can refer to the credo as sort of a third party agreed standard. That is how we tackle the problem and that sort of smooth things out quite a bit. So if you take somebody like, like Ray Dalio's company, one of the most successful investment companies in the world, their version of a credo is their principles, which is an incredibly dense set of operating procedures that they've written up over a long period of time. I like sort of the Netflix example a bit more, which is a set of statements with some texture as to what it means. Something you can read in 10 to 20 minutes. But it needs to be it needs to be needs to be as detailed as it needs to be for your specific business. You shouldn't simplify it beyond what's necessary,
Colin Iles 43:50
Would you or have you fired people that you feel that aren't living up to the credo that they've essentially signed up to, even if they're like the best salesperson that they're actually super in terms of their activities, but they're just not living to the credo?
sign of a good culture is that the culture ejects people that are inconsistent with the culture, your culture is weak. And there's somebody that is inside the inside the business and they they behave inconsistently with the culture and they don't get ejected, you've got a weak culture. It's almost like the immune system in your body. But like, germs coming in but your immune system is not strong enough to push them out. If we sit on boards, or we discuss what's happening in the company, and we hear, people get fired because they didn't act consistently with a credo, that's a good sign.
Yeah, that's, it's, unfortunately, I get the feeling again, when you're looking at the senior leadership teams of a lot of companies, one of the observations is you can actually have a look at that culture and you can see the bad apples sitting there and they talk about the credo or the the values that they've got, but actually it's just lipstick on a pig and they don't really live to it. And that must be a big turn off in terms of investment opportunities, a big giveaway.
And that kind of thing, in the fullness of time you struggle more and more to appoint good people. And it just creates a massive sort of downward cycle. Whereas if you can have a strong culture, strong people, new people coming in, so lifting the standard all time, that's quite a self reinforcing positive model to get on to.
Colin Iles 45:29
Let's take a quick turn now back to some more data related pieces so that you know, a lot of culture. Maybe we'll come back to that. This one, I don't know, I think we're just going to slot it in here. I don't know if we'll get to a theme, which we can slot it in. But it's a great question from Phadke as a new entrant in the market, how do you acquire data to build and train your models? And is it POPIA Or is it POPIA? I don't know. There's a debate on itself, isn't it? But is it poppy? Is it still possible to actually get the data you actually need to train?
I mean, there's fascinating things to do. If you are in the banking space, for example, legislation requires everybody that provides credit to give the data to the credit bureaus. So strangely enough, credit data is not really a competitive advantage for a bank, it's the data that you can put around around that data. And it is tricky when you start a business and you don't have access to your own data. So then you've got to be a little bit scrappy. In Rain, for example, we are going through a vendor, that satellite imagery on the satellite imagery that will look at houses and they've modeled that out of the satellite out of the satellite imagery, they can sort of put values on houses, which then you overlay that with a network map, in your sense of what the market looks like and insight that you might that you might put up. So that's, that's external data. And there's all kinds of other data that you can buy externally from third parties, you can then overlay on that and build quite a quite a rich model. Example, it's actually fascinating is as a company in Centurion, called Merlin. And they've got this AI that they called Tom. So Tom basically builds AI model based on an expert that you've got inside of your business. Let's say you've got an expert and this expert needs to look at a certain case. And Tom gets used, for example, in the US, it gets used at border control. There's, there's a certain set of data. And then your expert says in this data, I will do this. And then Tom will give him another set of data and expert will make decisions that Tom serves up, as the expert makes these decisions. Tom learns from the expert, and at some point tells you that he can accurately emulate the expert in 90 or 95% of the cases. And then you can take TOM and you go and deploy TOM, you can go and multiply your expert because you can multiply your expert down in a whole bunch of different places. So that's, that's quite a great example, if you don't have data, but you've got some experts internally that can actually make a call on a set of things out of out of out of experience. And then you can roll that out into the, into the real world.