Predicting financial stress (and rogue waves) – [FULL INTERVIEW]

0:03
So hi, everyone. I’m here with Ken Doherty. Today, he’s from Cerebreon. And he’s the CTO, and Cerebreon owner in the data analytics, software analytics debt recovery and insolvency space in particular.

So Ken, thanks very much for joining me today. I appreciate you joining really, listen, thanks for having me part of the series. And I suppose, you know, part of part of my immediate questions, we’ll get on to some of the analytics a little bit later. But you’re part of my immediate questions where we’re on things like AI and modelling and some of the analytics and we’re seeing this huge sort of like, increase in the, you know, the amount of focus there is really been on analytics really driven by data, but are you sort of sort of starting to see that that changing? And particularly, there’s been a lot of noise made around AI and AI techniques? I mean, but have you sort of seen that sort of maturing? I mean, what’s, what’s your current view? Kind of that kind of interesting industry?

0:57
Yeah, so that is a really good topic. I guess for me, maybe before we start to just my interpretation, I guess I probably have a wee bit of a gripe about the use, or the misuse of AI. It’s a very, very broad brush. And Katherine added guess it’s a good buzzword, a lot of salespeople can use it to spice up their product. But for me as the proper artificial intelligence as really, you know, as I said, as the machine doesn’t, it can think for itself. So it’s really that unsupervised learning, it’s kind of maybe constrained more to the sort of robotics if you just think of the robotics and the Sci Fi stuff. For me, that’s what the term and the branding AI should be used for. And then the other segment, is the machine learning. And for me, probably, I probably wouldn’t classify that as artificial intelligence, because it’s just a machine that’s been well trained by humans to, to look for patterns and stuff, very powerful. Don’t get me wrong. And so I guess, if we segmented in those two parts, I kind of feel the AI, the robotics, you know, that sort of up to the stuff that Tesla, the Googles, the Amazons of the world, you know, I really think that we’re already starting to see the tip of the iceberg there, I think that is going to innovate, that is going to get bigger, better actors, or, you know, our minds are going to be blown away with what can happen. And it’s probably isolated to those big players, because it’s getting more and more expensive to make any advancements in is it’s not for the small companies on the high street.

2:32
And this was for machine learning. And let me know, let me actually write I mean, like, a lot of this modelling techniques been around for the longest time hasn’t I mean, even back since the 70s. And what’s really changed has been, the amount of data that we can now store, I think, is number one, and then the processing power to then be able to process that and using techniques such as neural networks, which I think they invented in the 70s. Where they, I mean,

2:55
I think so and it goes back even further, I was, you know, I think a lot of the techniques that we you know, the machine learning is based on the mathematics was still kind of figured out in the 1930s and 40s. And it just, it’s, it’s sort of, it’s sort of lay on the shelves there, because no one could really do anything with it, because you couldn’t compute what you needed. So the just the advent of the rise of technology has now brought this to life. And so, you know, as a society, we’re getting to experience firsthand the product can deliver, but it’s not new. And so I guess, people selling it as a new thing is gonna be a little bit misleading at times.

3:29
And a little bit here, there seems like there’s more realism around machine learning now. And people understand it as machine learning. I mean, it’s, but there’s still research out there ready for use cases, right, which is, which is sort of, you know, been been been really the the Achilles heel of it. It’s like, what do I do with it? So great stuff I can do. But what do I actually do with it? And what kind of problems do I solve?

3:51
Yeah, I think simple is better to always start with that, you know, there’s a lot of nice machine learning techniques that people like to rush for the neural networks. I think it sounds fancier. Sounds good. But yeah, let’s start simple, you know, start, you know, start your foundation, just a step up just some good statistical analysis, you could say, but more traditional mathematics, and then go into the narrowly stages of machine learning some of those models, rarely to get a find the simpler the machine learning model. If you know, with a company or an individuals on that journey, you can actually understand what it’s doing. You can as a human interpret what a student, the more and more complex you get, it gets a bit more blackbox. And, you know, that then brings with it a lot of a lot of perils along the way because, you know, if you don’t understand the assumptions behind the the mathematics or the fundable fundamentals behind it, it’s easy to misuse. And I guess that’s the one thing, technology and the advent of bringing machine learning to life has brought the success of all the companies in the US use a story that challenges the use cases. But it has also brought the ability to misuse it quite badly. And the machine will always give View and answer. And sometimes the problem then is how humans interpret that answer. Sometimes you want to find the answer that you were looking for in the first place, and what how they approach it. And I guess don’t think that you need to go all singing, all dancing, you have to have the most elegant machine learning approach for you to win and use it in your company. Start very simple with some of the most basic techniques and build it up. And if it works, you know, what’s it’s always just going to be a means to an end. It’s just another tool in your toolkit to either run your business or find some insights on if you can do that simpler than we always go to that leads.

5:39
And I mean, there’s definitely been a lot of focus, certainly, certainly here in the UK from the FCA around transparency of decision making, I know in the US has been that for a while actually around transparency of decision making and lending. In the background, we’ve got these, the EU AI act that’s coming out as well, which again, is talking about transparency, it talks around sort of like making sure you understand that how do you just sort of see like the regulator, that there was like the regulations haven’t kept up with the technology, but there’s like, it feels like they’re starting to catch up now.

6:08
Yeah, I think that’s very true. And I guess that’s, that’s true with any kind of regulation, that technology always advances quicker than is brought under scrutiny, commercial science, and then hang anywhere, there’s a commercial reason that’ll always advance quite quick. Regulators do have a tough job to try catch off. So I think this EU AIA Act, is good to catch up. It’s very far reaching broad brush under I guess they’re trying to future proof it as much as possible, because they can’t react that quick to changes. So they’re trying to come up with the most broad brush future proof what the result of that is it encompasses a lot.

6:45
Yeah, so do you. Do you think that is there? Is there a sense that the regulation could potentially constrain some of these use cases in the future? I mean, because we come up with all these great ideas. I mean, you look at you know, for example, elsewhere, where they’ve got like, social schools and example, which are definitely prohibited, and they’re prohibited, explicitly prohibited in, in the EU AI act. But I mean, so socialist socialist for them, it’s an idea. So it’s almost like you can come up with all these great ideas or these great use cases. But do you think do you think that it’s going to constrain the regulation is going to constrain maybe rightly constrain, but the development of the sector?

7:21
So I think my opinion, so it is going to constrain the volume of developments in the volume of an event, you know, not gonna say innovation, just the volume of use cases and different permutations that are going to be developed? Is that a bad thing? No, I do feel though, that the quality of the innovation will still flourishes will still thrive. And so I find the Regulation Act will probably in a good way, or most of most of the time, filter out a lot of the stuff that as you say, there’s a lot of these ideas out there, it can be done. But then if you’ve asked the question, like, what problem is that solved, you’ll find a lot of these use cases aren’t really solving a problem, it’s just been done, because it can. So I think the regulations will improve the quality. So anyone that has a good innovation, that fits within the regulations, and ticks, all the boxes, what’s still comes through as a, you know, a proper use case, or a product from commercial case, will actually be a really good product at the end. So I do feel that we’re going to get better quality and less volume.

8:27
And I suppose it needs to fit into into the constraints of what society wants as well, rather than just that it can be done because it because it needs to, it’s gonna like well, is that is there an end use case for the Society for customers? Really, which I suppose comes back to having good use cases, even in business? as well? As does it? Does it help the customer does actually help the process rather than, here’s the technology, let’s find something called to do with it,

8:48
it is good. And listen, you know, I would have a tech background myself. And it would always be that sort of that way orientated. And sometimes you do get sort of caught up. And this is definitely in the software or web developers thing, because they can build it, they do. But if you just ask, Well, what problem does that solve? And you know, if you can strip it back down to that question of why would I put them in there? What problems does that solve? If you can always ask yourself those type of questions along the way, I guess you’ll quickly you’ll quickly filter out the good from the

9:18
bad drilling down a little bit on machine learning. I mean, there’s, you know, there’s always that the split, I suppose between, you know, supervised learning and unsupervised learning as far as categorization, that kind of happens. I mean, how do you sort of, you know, how do you sort of see particularly the, the, the unsupervised pieces, is that, is that really an area for development? I mean, people always looked at that as like, well, I can just kind of throw the data in, and it’s going to solve every problem, whereas the supervised piece seems like Well, look, I’ve got actual use cases that I can actually go against, and I can actually use that to train my models. I mean, is that is, is that is that is that kind of is there a split in terms of those in terms of use cases?

9:55
And certainly, I guess the split for me would be if you have access to historic data, or anything that you can train models are better users, like it’s theirs, you know, it’s use it because you’ll it’ll for your use case that will be giving you more precise, more direct answers. And so if it’s there, if data is available and accessible, then definitely use it would be my approach as your plan A, you tend to go to the sort of unsupervised stuff, if you, if you don’t really have an option, if you don’t have that historic data set, or the use case just doesn’t, it’s brand new, there hasn’t been any data to base any assumptions on, then you go down on supervised, but you’re right, you know, the precision, the accuracy with what you’re going to get, if you take your categorization, approach, or use case, the like, it’s going to give you a ballpark, and it’ll help you segment large volumes, but you’re not going to get that sort of precision on every data point. So I guess, you know, for researchers, I would like to thank him, you know, the academic Institute’s if they could, you know, continue to try develop the unsupervised type approaches, so that they can become better and, you know, commercial companies that you can start to produce, that would be a real, real good.

11:17
One, well, the challenge that we have today, right, is just whether is it historical data actually going to be predictive of the future? Right, so I’m gonna let everything that’s happened over the last sort of, you know, three years or so. And, and is it an or even the timeframe in which that can be predictive is sort of like narrowing down it seems as, as well. So it’s almost like, you know, are the techniques we can use to basically remodel faster to get to information faster, and that I know, you guys look at things like affordability, and you can look at various indicators and sort of like, what happens here, and here does that does that is that a signal something might happen down the road? I mean, he sort of time horizons for predictions are almost like narrowing down because of because of the quality of the data and the history that we’ve got,

11:56
let’s not do that you’re spot on time horizons, you’re right. Maybe Maybe every generation have said that, but it feels like there’s been more major events happening in a quick succession of a couple of years. So you’re right, you can’t base you can’t use Ted, the last 1015 years worth and thinking that, you know, the next 15 years is going to be the same. It shouldn’t be disregarded, though. Like there, you know, there is a huge, huge amount of let’s say that the historic data, let’s say 20 years, that well can you know, there’s patterns in there and life goes on, if we maybe just thinking more in our own sector, in terms of consumers and consumer affordability in the finance sector, there’s a huge amount in there, that will still propagate into the future. But that should only form your kind of base layer of your future predictions, you then have an obligation, I think, as any sort of analyst to then layer on top of that, the different scenario models of what could happen. And you know, you can inflation and you know, gradual creeps and things that maybe have seen before you can layer that on top of some of that. And what do you do need to go to the extremes of what happens if there is a crisis event so and we’ve been through a couple of them and in the past couple of years, so that all has to be layered on top to make your future predictions. And you can’t just you can’t just let it test all data, and just assume everything’s going to be the same. But that’s that’s not to say you shouldn’t use the audit, it needs to be there as your baseline. But then you have to build up an add on to that.

13:29
Yeah. So it sounds like you’re going and I know we had the conversation before around around physics, but they’ve got a bit of a background, although yours is much more illustrious than mine, I have to say, in physics, but you sound like you’re talking about waves there to a certain extent, there’s always like, you have the macro wave, right? Which, and the world doesn’t really change, right? So we always have this like, you know, long period wave of you know, of decades, even right, and then over that, then you have the next wave, which is like a shorter, shorter, shorter period of a higher frequency. And then you got the, you know, the next wave, the next wave, it’s almost like, each of these techniques allows us to basically almost like fine tune in these extra waves to give us more predictive in terms of what the overall piece looks like.

14:09
Yeah, exactly. So yeah, I’ve been on course, we’ve spoke before, but yeah, I’ve done a lot of work in the renewable energy sector, and in particular ocean renewable energy, and oceanography and of modelling ocean waves, as has been a big part of my life. And, but there’s something that you know, fairly unpredictable, you have a base, you know, you can have a good baseline, but the inverse rho wave, shock waves, all that kind of stuff that can just happen naturally in the ocean. So the earthquake driven or anything like that there, but these rogue waves can develop. And so I spent like a decade, probably modelling and figuring out techniques of how you can predict it and model, that type of stuff. And it’s very much a time history type approach in terms of the market time goes on the next wave hits the shore. Next wave hits the board. And so you know, there is a lot of really good techniques in there. already from those type of physical sectors that, you know, in the physical sciences sectors that I think can be applied into other industries. So like the finance sector, and that’s exactly what we do we bring a lot of these techniques that we’ve learned in other parts of life, but are applying them now into the finance world. There’s a lot of

15:20
I think it’s quite it’s quite interesting race. And so we talk about cycle we talk about, you know, credit cycles or predictions of default as well. But they also have, you also have shocks as well. So people have live shocks, or there’s economic shocks. And they always seem to come as a bit of a surprise. Do you think there’s a sense around that? I mean, because this idea of the rogue wave, which would be what I mean, you have these huge waves that basically, I think it’s a superposition of, you know, lots of different ways that are happening at the same time. But yeah, because we can almost like be in a rogue wave now, economically, with all these factors happening at the same time. And so, I mean, is that is that is that a good analogy in terms of like, trying to predict some of those shocks and shocks shocks to the economy? Because it’s almost like this, this, this black swan effect? That is like, well, we can’t predict that. But but maybe maybe we can get a little way along the way. So some of these things can happen when things are happening in conjunction,

16:08
exactly. So you know, that the first starting point, and you know, the statistics and the sciences were like, they’re like, I guess I specialise in a form of statistics called extreme value analysis, which is all about this, these kind of rare events, it takes a little what you can you can, you can assess all the possible things that could happen, you can then assign probability probabilities as to what’s the likelihood of it happening, you can start to Horizon forecast off, you know, the window that you would need to wait until you might see one of these events. So you can get a, you know, that’s well understood all the analysis behind that. And then I guess, you know, there’s a decision to be made, is it a suitably no probability of happening that you’re not going to plan for it? Or, but I think if the scenario if the outcome if the detrimental effect if you know, what can happen? And you then do a quick sort of risk assessment almost on, well, what’s the detrimental effect if it happens, and things like economic shocks, you know, I think there should always be a plan in place for if it does happen. And so I guess that’s what needs you know, I guess that’s what, that’s the missing link. So people tend to go well, it’s unlikely to happen. So I’m not going to think about it. But I think you do need to plan for if it does happen, and the statistics and data analysis out there, we’ll be able to, like we can analyse when it the probability when it could happen. But you do have to have that plan in place. Because you’re actually showing that it is possible for it to happen. And so I’d like I am a firm believer, and, you know, economic shocks, they can’t, they will happen again, they’re going to happen again. I guess maybe every generation just thinks I hope, I just hope it doesn’t happen at all, but you don’t have to go through it. And then maybe people just glass is always have fun, perhaps but then I guess you have to be a bit of a realist. So yeah, like the techniques out there, you know, with any dataset and time series of data, you can use letters, your past your Monte Carlo simulations, extreme value analysis, there’s like there’s well known stuff out there, you don’t have to go into fancy ML to be able to do this sort of stuff.

18:22
And what about So you mentioned time series data there, which I suppose we should kind of what we’re talking about. I mean, that seems like that’s one of the challenges you often have, particularly in the the collections and the recoveries is in the business world, it feels like sort of predicting, if I pull this lever is going to have an effect instantly over here, right? So we’re instant, almost like this. And there’s this idea around like I affect something here, it might have an impact in six months time or nine months time. And it’s the combination of this particular driver plus this driver plus this driver becomes very complex very quickly. I mean, as it doesn’t as it does in physical systems. But it’s it’s there feels like we sort of that’s one of the areas we kind of struggle with. And it’s quite fundamental, I think, to that to the collections process. Really. I mean, I mean, do you sort of you sort of see that in some of the analysis you’re looking at? So for example, if you’re looking at, you know, income and expenditure, you know, what’s what’s the relationship between sort of almost like small indicators today, that could be indicative that this person might have a problem in three months down the road, or six months down the road? And it’s almost like, that kind of prediction becomes kind of critical,

19:20
it does become critical. And you’re right, I think in the finance sectors, it’s I don’t feel it’s used well, either at all or not very well. And it is all about the big, the big number, you know, it’s percentages differences. And as you say, behind that is the assumption that if I do change something that will have an instant effect, that my balance will reduce to zero, but it’s not it’s the causal effect, like humans, we all live in an iterative, you know, we’re all marching through time. And our decision today to spend on something may seem innocent at the time, but it will accumulate that single decision today even if you buy something, don’t shop if something then happens tomorrow. There’s, you know, a night already spent money the day before you like that they’re not unrelated, there is a causal effect. As to and you know, things can contrive in the negative way, and really build up for people and push them into, into more money difficulties. So the time series a time history analysis, I feel is absolutely critical. And that has to be introduced into the definitely into the finance world. And, you know, certainly in terms of debt and collections and affordability. And I guess that’s the approach we take, we do that as our kind of our starting point, we will only sort of start at that level of granularity. Because yeah, that sort of pattern recognition, can reveal an awful lot, and explain what, what will happen. And it’s not just the big number of, you know, this much money going out of my account to this much money coming in at the end of the month. What happened in the days in that month, is critical to knowing what’s gonna happen next.

21:01
And what’s your approach and flight, how you think about how to sort of, you know, tease some of those, when I call the micro indicators, because we like to use some of these indicators out to say, Well, this one here might be particularly indicative. And it’s indicative in combination with XY and Z, for example, use things like open banking at the moment, right? So the fact that I might have gone to the shop before I went to the, you know, before I went to the petrol station could be indicative of something different. You get into all these kinds of complex scenarios. What how do you think about that in terms electronic tea, some of that stuff out and, and analyse it, so becomes representative and actionable?

21:34
Yeah, so I guess that is the challenge. It’s, you know, make it actionable. It’s all well and good having all this data. And you know, I think, as a society, we’ve never been better at collecting data. But, you know, for the volume recollect, and I don’t think we’re actually analysing or looking or using it to do good. So you’re right, it’s the pattern recognition and distilling that down into something that, you know, a decision maker or someone can actually actually, you know, that is key, what, you know, one example will be, and the one thing that we always try to pick out from our analysis is the prioritisation of money. How is sondland prioritising understand, so, you’re right, and that is to entertainment. And, you know, if we have to analyse, they’re either open back in data, or we can examine their bank statements automatically, through, you know, through their sector we work in, and it’s how they prioritise if they have money, what have they prioritised and the timing of when they spent and what they spent it on is indicative, you know, that can be flagged as a, an actual metric. And very often, it’s, it’s maybe it’s not the person is doing anything bad. But you know, some, just sometimes you’re just, you’re not thinking and you’re just impulse by impulse spend. You know, and some people that will have different behaviour between if things are falling into an element of distress, or they can, they can see things aren’t going great. So people react different. So people put the head in the sand, just pretend it’s not happening, and actually then splurge and overspend on elaborative items. Many men in particular are very bad at you know, buying one big, very expensive item, a brand new set of golf clubs are one big points over the top.

23:25
I’m fed up the golf clubs, like that, that’s it, that’s it, you know, life is a heck with it, I’m gonna kind of get myself a new bike, and a set of golf clubs, and, you know,

23:37
I’ve done and things like that, you know, I guess we all like to think we’re unique human beings, and you know, no one, what we do, like as a society, we kind of go for it, there’s a lot of us will have the same reaction and the same behaviour as to how I would react. So there probably wasn’t that maybe not the greatest scenario, but it is possible to pick out that sort of pattern and prioritisation of money, the slow bleed of you know, small transactions is a classic, you know, your yours two pound 50 Every for coffee, but you’re doing them seven days a week, the cumulative build up it’s you know, you don’t noticeable on day to day basis, but if someone like yourself holds a mirror up to you’re going this is actually what you spent maybe a small mortgage payment and don’t take away coffees throughout the year. So,

24:27
but But I suppose is also around change behaviour, things like you might be buying a coffee coffee in a certain spot and then it’s changed you go somewhere cheaper, or it stops or it comes every other day. I mean, all of these can be indicative as like, changes behaviour. I suppose it’s the the piece that’s difficult is what does that mean? And what does that mean? Is it just me just changing my behaviour because I’ve, you know, I’ve moved to a different town or I’ve got a different you know, different job or, you know, something’s something’s happened versus versus actually being indicative of, you know, a greater a greater underlying effects such as you know, Financial stress, all those kinds of things. I mean, I mean, that’s really come back to the supervised learning piece in terms of like, pull some of that stuff out.

25:07
Yeah. Correct. You know, so, and especially when it comes to the time series analysis, you know, if you have on your, on your x axis, you have your time axis, and you have, I don’t know, your income, your disposable income. And it’s just your, it can be this sort of irregular heartbeat. And then, so that, you will build up a sort of baseline pattern that, you know, by and large, we call it week, month or month, this is broadly how someone lives. And then you can start to spot significant changes in that. And when patterns start to deviate from their, their sort of baseline pattern, you always have to have a little bit of knife variance in there, naturally. But you’re right, you know, that pattern behaviour, unusual change? As long as you understand, you know, I guess, for me, I always find our role in our company is understanding what happened, we’re not to say if it’s good or bad, or, you know, what the action needs to be after that that’s maybe not our job, we’re all about presenting the facts and the figures the insights as to why did something happen? And it happened and the and the reasons and the correlations that build up to that. And I guess, uh, you know, I guess that’s maybe before we then sort of draw the line going, like, we’re presented facts and figures, and again, that goes in, in terms of transparency, like, if you’re a data driven, evidence based decision, organisation, you know, you’ve automatically ticked, you know, regulation boxes a transparency out, you know, why did you make a decision? Well, here was the evidence we had at the time to make a decision, and you know, you can back up and you should be able to stand over every decision you make. And like, you know, in our sector with the FCA consumer, God has a new obligation coming in. And it’s, there’s a lot to it for firms, but the main thing is evidencing why you made a decision? And can you evidence that you’ve made, you’ve done the right analysis to make the right decision? So I guess, you know, as long as you take a data driven approach, by default, you’ll take almost all those boxes.

27:12
As I said, the master data guide is just expanding exponentially seems doesn’t it? So I mean, apart. Part of is also like trying to find, what is the signal within the noise as well, right. So that phrase gets used, and I suppose, you know, I love I love just putting on the theme a little bit around sort of like wave analysis for analysis, because it does become like, you know, the crackly radio, right? Because you have almost almost data, it’s like, how do you actually tune into what sort of that signal is actually happening? And then you’ve got, then you’ve got extra data within that itself to then find out what’s the frequencies within within that particular signal that’s coming out, you know, to give you to give you the pattern you’re actually seeing, and it’s sort of there’s quite good, it’s quite interesting analogies around I love the analogies around using like, scientific or mathematical techniques, you know, that can be used to look at some of these almost like adjacent Type, Type Type Type fields. Really?

28:01
Yeah. And I think that, you know, your I’d agree the, I find what may be coming from an engineering background or science background, a little bit biassed, but I find the, you know, the technique techniques used in those other sectors that gets valued financial services and oceanography, you know, you can argue that there’s no relationship there. But there actually is in terms of the the techniques that should be used. And very often a crossover, cross pollination coming in from a completely different sector. And often in the physical sciences, bringing that into the business and financial work is probably just maybe some sometimes the fresh thinking that is needed. You know, yourself if, you know, if you’re, if you’re operating in one space, too much, you get kind of blinkered off to what the possibilities or what someone else is doing in a different sector, that that could be useful. So we’ve certainly found that, you know, we’ve we’ve started to apply all those techniques from our, our different backgrounds. And, you know, and I don’t Well, we looked around enforcement, I guess I didn’t find it was been used very much before. So we thought, let’s try bring in some of this and see what can be revealed.

29:05
In what what are some of the things that the discussions you’ve been having across the industry in terms of, you know, areas that are changing or areas you think are going to change?

29:14
Yeah, I think in our sector, you know, obviously, affordability and, you know, cost living crisis is you know, for debt recovery, insolvency for loans, financial services is still always the hot topic, hot topic, but like, what are they doing different about it? I guess, you know, that that’s one of the questions we’re getting from firms. And what can we do different there, I guess, are busy trying to manage the, we’re already in it. So they’re now already trying to manage the process of people that have kind of fallen foul to it already, and are really struggling. But it’s going to continue. It’s going to continue for, unfortunately, a long number of months yet. So it’s really this horizons gammon is maybe we’ll wait A way to do and it is that service scenario modelling. What if so, economists are predicting the inflation is going to go like this and the cost level is going again? So what if scenario, but what if it got worse? Or what if it was better?

30:14
If you look through it from a portfolio point of view, here are the signals that are out here that are saying, well, this, this, I think might be a problem. And it might, you might might be hidden in your overall portfolio now. But it’s such a clear signal to say that something here is going to happen almost like a diagnostic test to say something actually going to happen. And almost like flagging that from Horizon scan point of view, and then I suppose being able to drill down to an actual level to, to then be able to review those accounts, or to be able to, you know, help them actually take action and sort of mitigate risks as well. Yeah, and,

30:43
you know, I think for the firm’s the the insolvency and debt advice firms that we did with, I guess, they didn’t, they have such a high volume on their hands. They only have, you know, a lab, you know, as much resources they can, they can’t do, they can’t talk one to one with everyone in their portfolio. So to be able to, you know, hyper accurately segment that portfolio that, you know, this is the group of people, most of risk, they will take action, you know, the solution really, in this in our industry, though it still feel is that human to human contact, it has to be like that can’t go away. But if you can segment of God, these are the people, you know, we definitely have to talk to reach out this has to be armour and the shoulder support. And there is there hopefully a bigger percentage of the portfolio’s that are are doing fine. They’re coping, they’re, they’re managing well, there’s no, there’s no risk signs just yet, or alerts flagging just yet. So if you can’t get to everyone, this cohort will be okay for another one. And I guess that’s, you know, that is, that is one of the kind of main value adds to think we’re bringing to the firms that we work with. But I guess, because we start from the granular level of the spending behaviour profile of that individual or that household, it is hyper, accurate and hyper personalised to

32:08
be able to drill down to something that’s very sort of meaningful. Yeah, yeah. I, that’s quite an interesting way of using the math techniques, because it’s all scalable, or, you know, you can, or what’s that will be the reversal scale, or you can use it to basically drill down to the lowest level, right.

32:21
Sure. Yeah. And, you know, yeah, I guess that’s, that’s, and you know, the the key is, so, to two elements with the collection of data is key. If I guess as a data scientists, if you don’t have data to work with, then your skills, it’s, it’s hard to be employed. So collection of data is key. And you know, collection of good accurate data is key. I think a lot of firms, a lot of software, telcos are good at that, and are collecting a lot of data. But it’s not know the use, that’s one thing. And that’s a very important starting step, if you don’t have that, there’s no, there’s no one to talk about downstream. But I think, you know, everyone can do that pretty good note, but it’s making use of it, and making use of it on time, because it does have a shelf life, you know, you need to analyse it, you know, almost real time hot off the press, because it’s going to be different than the events a couple of days later, or a month later in someone’s life. You know, life can be very, very different. So it has to be that incident. And so it’s having the facility that you can set up that you can drill down, this can be analysed in a, hopefully a fairly systematic and standardised way, again, for the transparency as to how we approach it. And then here are three to five actionable alerts or outcomes that have been spotted. And then the treatment plan of that. Well, I guess that’s specific to every use case. But very often, in the debt recovery sector, the treatment partners engagement with the consumer and the right type of engagement. I guess that’s where the human element takes place. And that’s right, why these what these firms are very good.

34:00
Yeah, yeah. Well, it’s an interesting topic. And if you say there’s one area that we need to look next, I mean, what’s coming next? Because it sounds like that, you’re pretty far ahead in terms of like, looking at some of that some of these new techniques, but I mean, what do you think? What do you think we should? Where should we be looking next? I particularly look forward into the I mean, what could have what can happen over the next year or so? I mean, are there other areas which you think are gonna be the next sort of light frontier of development in terms of data analysis,

34:25
and diversity analysis in terms of being applicable? So again, this is a disjoint. Here, I find the research there is some some wonderful that’s innovation and research being done but they’re, they’re still so conceptual. It’s hard for a company or commercial entity or, or it’s hard to bring it to life to the populace, because it’s still abstract. So maybe in terms of getting something actually applied into the sector to make it a reality so that we can all live in touch and breathe it. I think for the In the finance sector and the debt recovery sector, they need to adopt much more of blockchain type technology. Like it’s a game out there, it’s well understood. It’s been used in all sorts of different things. But for the for some of the, for some of the sectors, I think that could use it the most, like payments, collections. Buying and selling off debt buying and selling. You know, that’s exactly what blockchain should be, should be used for. And it’s probably the one area that it’s, it’s not really used. And so that could be a thing I would like to see brought into, certainly our sector. And I think it’s a real element for school. The challenge is, that’s not really a technology challenge, because it’s there as understood, you know, it can be done. It’s a stakeholder buy in, it kind of works, everyone has to buy into it, every side of the equation regulators, every player in an industry almost has to buy into it for it to actually work. And getting everyone to buy into it at the same time as a challenge. Definitely is there, it’s ready. It’s ready to go from a tech perspective. So I guess that could be one thing I would like to see. But hopefully, I would like to see. That would be because it’s, again, it’s Trump’s party, it’s known as a non fungible tokens. But like, it’s, you know, it can’t be interfered with, you know, payments that go through four levels of process and passing from one entity to the next before the cycle is completed. It’s just scope for error scope for crying scope for a lot of things to go wrong. Don’t Don’t say it does that for you. Blockchain type tickets, you can’t you can’t interfere with it, which is beautiful.

36:47
Now, let’s face it, the big opportunity there is like there’s a there’s a big, big opportunity in terms of efficiency isn’t there because you don’t have to go through all these handoffs, and the quality check processes and those kind of things, because you can prove that that’s the debtors example, you can prove that that’s the particular you know, the the particular person they are or whatever it is, in terms of using Blockchain. So there’s, there must be some sort of efficiency gain or but I suppose, coming back to time series of the fact that across multiple different entities makes it quite hard for us to think about rather than something we just control ourselves,

37:17
you know, and that is in terms of our sector and works and the interactions and, you know, for some, it does roll into question, then what you know what your for if one of your functions was just to pass things on, then it would supersede some fault handler. So not everyone would necessarily want it in place, either tech can, can disrupt and can, although Tech Tech can just do the job of you know, what was previously done by humans or by firms. But I guess that’s the march of progress. That’s the march of time. Yes, tech is always going to go forward. It’s always got to get better. And I guess, we always just need to be adaptable and think, okay, that I no longer need to do that, because tech is doing that for me. But what can I do? Like everyone has good, valuable transferable skill sets to do something, and contribute different ways?

38:05
Yeah, and the one thing you can say about humans is if, you know if, if we always find something new to do, yeah. There’s never there’s never a lack of things to do, unfortunately. So the dreams are the dreams of the 60s and 70s, where we just sit back with robots, or of helping us make cups of tea and coffee don’t seem to have happened, unfortunately. So we seem to be busier than ever so well, Ken, thanks very much for making the time. I really appreciate it. As ever, it’s always good to chat about some of these mathematical techniques and techniques and sees here’s some of the things that you guys have been up to from an analytics point of view. So really, really appreciate it.

38:40
Chris, listen, thank you very much. Appreciate that.