Avoiding Harm – are we thinking about vulnerability back to front? – [FULL INTERVIEW]

0:00
So hi, everyone. I’m here with John Willoughby today who’s the director and co founder of Elena. And so Elena over in the sort of the data science, dynamic scoring, you know, data analytics kind of space, you know, very much my sort of dynamic basis. So, John, thanks very much for joining me today. Really appreciate it.

0:16
Well, thank you for the opportunity. It’s good.

So the first question, I suppose I wanted to ask was, you know, have you found the marketplace changing over the last sort of two years, particularly, overall, quite close to the data in terms of like, how that’s been sort of changing for companies? I mean, how have you, have you seen the data change? Or have you seen the dynamics change over the last couple of years from from your datasets?

Yeah, I think we’ve definitely seen in the data, particularly with regard to, to financial vulnerability, we’ve definitely seen that change, we’ve seen it, we’ve seen it, you know, as you’d expect a degradation in it within financial stability within over the country. So and, and in areas where perhaps, you know, originally, you’d think actually, that was a fairly resilient area, we’re seeing some some reduction in a resilient. Yes, we’ve seen an impact.

1:16
And is that is that pockets by market, i remember you’ve got quite a good sort of like geographical detail around sort of where it kind of sits? I mean, do you see I mean, there are definitely areas of increased vulnerability or increased financial difficulty or resilience. So let’s, let’s call it that, versus others? And whether some of that’s kind of surprised you? Or have you changed from what it would normally?

1:37
Yeah, so one of the So yes, we do we do kind of, we’re Geo, geo spatial, kind of scoring shops. So we don’t don’t take individual customer data. So it’s, we can step outside GDPR. And we, we take a large set of data across the UK, postcode, suburb postcode level, and just just really above the individual, so we’re gonna get together we got, we were very conscious that we don’t want, we don’t want to know who’s behind that, that data. So we take those large datasets, and we take outcome data from our, from our client base. And we apply machine learning and allows us to say something about, about the UK as a whole at that kind of granularity. And, you know, we use that for for a variety of services to to clients,

2:40
it feels like the pandemic sort of thrown that throwing things up in the air a little bit, and some of the preconceptions we had before maybe aren’t quite right, in terms of how that might have affected people recently, but then, does that really impact us going forward as well? I mean, what’s your what’s your kind of your mind you think things have changed for good? Or do you think sort of now we’re going to sort of get back to, like, the old ways of working, it was it was like two or three years ago,

3:04
I think it then we’re gonna come on to this, and I probably said exactly the same thing in a minute when we come on to it, but you could take you take, you know, call centres, and we will probably talk about call centres during this. A lot of the call centres relocated to to places like Glasgow, and that’s driven by by cost the cost base. Now with remote working, those, those individuals that are servicing those contact centres can now drop the Glasgow contact centre work for for London, and they still do exist, you’d be surprised the contact centre and get paper. And what that’s going to do is going to force him we can already see this, you know, he’s going to get somebody advertisements for for contact, you know, outbound outbound telephony staff. And you can see, you can see the salaries begin to arise in areas like Glasgow, because they’re now competing with remote workers that are that are able to service anywhere in the UK, I think levelling up that that levelling up agenda is going to the government have, I think, in some, in some ways, it COVID is going to the the impact of COVID is going to kind of force it.

4:29
Yeah, yeah. I think what do you think that means for the contact centre industry, I mean, sort of there is a sort of amounts of sort of comfort around having people in your contact centre in physically in terms of like security, data security, privacy, policies, procedures, those kind of things. You know, we’ve all been working remotely over the last two years, essentially, I’ve been meeting nearly all of us. And it’s so attractive and yet you don’t get some of those security Do you think do you think there’s going to be a? How do you think that’s going to evolve? Do you think we’ll end up going back to the office? Or can contact centres or just using contact centre really exist? Do you think?

5:09
Yeah, I think I mean, we’ve seen a distributed contact centre in the US, they know they have them there, they have quite some very interesting, you know, as a result of that, they have some very interesting services where they could, they could do load balancing, essentially, of all the Contact Centre staff. So when when demand peaks, they could just push out of the out of the, you know, the contact centre, if they need more stuff, they can just push that into into the kind of the home worker type, type space to meet to meet the demand. So I know the rules, and the rules and regulations in the US are different different to in the UK, but, you know, they’re able to do some interesting things, things that have distribution network, I think, in the UK. So, you know, I had a similar question to some of my banking clients, at the beginning of the of the pandemic, you do inbound and outbound dialling? What are you going to do? And they said, you know, what’s the, you know, heads about risk? I talked to him about it. And I said, Well, you know, we’ve got, we have, you know, to be, you know, we have the digital infrastructure infrastructure to increase security, we’re rolling out out with with the kettle, on the laptops, the various VPN, dedicated VPN type services, etc. And also, you know, we get stopped, you know, to sign at the stations to say that they are, you know, adhering to to the kind of policies and procedures that we require of them. So, yeah, I think it’s an increased risk, but I think businesses are responding to, to that. I think we will slowly over time, see, move back to, you know, the consolidated. Centre centre. I think, you know, I mean, we’ll, in terms of the other thing is, in terms of that increase in cost base isn’t, I mean, I think we’re gonna see salaries rise, I think that we’re going to see that for two reasons, I think, because the distributor network, and also because of, we’re under increasing inflationary pressure, we’re going to see sea salaries rise, and I think that’s where so much the plug, isn’t it? Yes, where the banks can come in and support, you know, with their data insights, you know, support more, you know, increased productivity increased, you know, contact rates, for example. So, you know,

7:55
that’s gonna be my next question, which was, what’s the role of data? Do you think that it’s going to play sort of going forward in the future, especially if it gets distributed contact centres? You know, everything’s much more remote. And even even the, the consumers, or the, you know, the people are borrowing money. And I suppose in the, in the financial in the financial services world, they’re much more emotion interacting more digitally? I mean, what’s the role of data or the importance of data? And how does that sort of like flow through our processes?

8:24
Yeah, I think I think I mean, this is, this is really interested in you know, there’s, there’s, there’s, you know, multiple angles to this, there’s, there’s an angle I have as a data, data seller. But there’s also the angle that you see from the regulator. So, you know, because the duty act is really quite interesting from a data perspective, they’re the proposal that came out earlier in the year and and, you know, that the outcomes testing is that that every step of of the customer journey, the FCA is going expects outcome outcome sustained, and that puts a it puts the data to match it’s not just data demand on the on the Almanack you know, that data that data is there anyway, they capture for a greater or no, they catch that data, but it’s actually now more about investing in that data actually spent time analysing that data and, and, you know, questioning it from a validity point of view when they think it where it could be wrong. So, you know, for example, you know, some does capture the vulnerability of a customer and they capture that fruit through interactions with the customer. And of course, they’ve got to request from the customer that they can record that detail. And then one of the one of the firms I was working with that they when they when they talk about the vulnerability identifiers they’ve got. And when they compare their number of clients, that number of customers they’ve got that self identifies as vulnerable, it’s a lot less than what say, the regulator is saying or other external heard they say. And so, you know that that almost caused them to question the mechanism by which they’re capturing that information and the validity of that.

10:32
But there’s a bit of an ethical, there’s a bit of an ethical question there as well, to a certain extent it was like, so if I don’t want to identify myself as vulnerable, but all of the indicators are saying that, you know, I likely am in probability, do you recall that person? Is voluntary treatments vulnerable? Or do you only do it when they actually when they actually say, Well, yes, I am. And I need the help. And at what point does that line kind of split? Right. So so, you know, when when does it become such a compelling case that you say, well, like, we’re going to have a bit of an intervention here?

11:01
Yeah, I think I think there’s two, there’s two answers to that. I think the first I mean, if we just if we just split, we’re talking about financial services, financial services. And this is, this is where we kind of come back to, you know, one thing talking about fungibility, but actually, what we’re, what the regulator is more interested in is the harm that we can, that financial services could cause to people and there’s broadly two, well, I would argue there’s two types of harm, because we can cause them financial vulnerability, we, you know, all we can, we can cause mental harm by by the stress of the situation, and those, those are the two harms that we can, you know, that we can we can cause now, there are other types of vulnerability, there’s, there’s physical health, there’s, there’s, you know, capability. And non English speaking Thai those vulnerability metric, but they’re more for the customer journey, they’re more how the customer wants to interact with with the institution. And I think if the customer doesn’t want to say that they’re physically impaired, or doesn’t want to acknowledge, you know, difficulty with language, English might be a second language doesn’t want to acknowledge that, that that’s entirely up to them, because they choose to interact with, with the, the lender, I think, when when we get to the harm that the lender could potentially through a lack of service cause an individual so it might be it might be, you know, financial accountability, the lender can see that, that, you know, especially the ones, especially organisations, where banks that are taking in primary accounts, they can see the, the financial stress that individuals are under, and they have that data, I mean, you know, and so and so, I, you know, at that point, they could be monitoring financial vulnerability. Yeah, you should be calling it out. So, so

13:09
do you think we’re thinking about vulnerability wrong? In a way? I’m not saying we should change it as much as just like, do we need to, like change the paradigm and how we look at it? So really, yeah, I mean, it really is about outcomes. Right. So it’s around outcomes and, and avoiding harm. And, but our language has been sort of at the front end talking about vulnerability, and the causes, rather than the outcomes and a certain extent do you think?

13:32
I think there’s, I think, I think there is a it’s easy to just talk about vulnerability, and it’s nice, it’s nice and, oh, don’t meet, don’t we do a great job on vulnerability? It’s hard to talk about the harms we do to people. And I think really, the, it should be an about, and we should be talking about actually, if that individual banks with us, if that individual is our lending services, What harm could we cause them? And how do we support? You know, how do we ensure that we reduce that because I think Tom is the key, identifying what harms you can cause which is going to be stressed, or financial, financial hardship.

14:18
One of the most interesting things, I think there’s is almost like this feedback loops, which I think is going to come out. So we’re talking about outcomes here. So we’ve gone from the front end of the of the customer journey, talking about, you know, causes to outcomes, which you can measure, then it’s almost like it’s the feedback loop to then say, well, then, okay, you can’t stop it happening with this particular customer. How does it how do you then feed that back? So it goes back up the process, but it’s always like that feedback loop piece, which is perfect for machine learning to certain extent, training against outcomes, right. It’s like, how would you do that? Then identify things earlier? I think that’s that’s kind of fascinating. I think

14:53
it is. And I think you know, I think that’s what that’s what’s interesting about the CP is FCAC vfca CP and increasing increasing use of use of outcome, and reporting?

And how much how much do you think that’s going to play into machine learning? And how involved do you think we are in that emanates across stones and stuff that you guys do? But I mean, it seemed like it’s a it’s something that can be done. I mean, I know in places it is being done, but it’s like, how much more can it be done? Or how evolved? You think the industry is on doing that?

I think I think I think we have pockets of it. I don’t think I think, you know, we all talk a good talk about AI, you know, AI? It’s a set of if statements, really, because and you can see that because when you ask it a question, it can’t answer. He can’t answer us is do you want to take somebody I’d love to? So yeah, so I think we are seeing a AI and machine learning interchangeable, but you know, increasing in, in this space. In terms of in terms of, you know, drilling into into internal data? That’s, I think that’s got some way to go. Within businesses, I think, I think we’re only at the start of that, you know, I’m not talking about Lnf here, but I think I think a lot of the vendors are really starting, still trying to understand, understand what that how to how to

apply it. And it does feel like almost like the the use cases are sort of like almost like they’ve almost like struggled to a certain extent, but it really does come down to, it really does come down to, to outcome, isn’t it? So we’re trying to get through around, like, what’s the outcome, you can measure an outcome as being a contact, where you can measure an outcome being a, you know, vulnerability, or you can measure it as being you know, someone in death or financial harm, and it’s all those outcomes can then drive the methodology. So it’s so it’s the, it’s the outcome that’s driving the, the need as much as the technology looking for the need.

Yeah. I mean, that’s certainly how we went about our, our, you know, our, when we started out how we start to develop our product suite, it wasn’t, it wasn’t, Oh, we’ve got we need to apply machine learning AI at any cost, we must do it, it was more rewarding, what what can we provide our clients that will reduce their, their, their cost base, really, and so and that’s really been at the heart of everything, we’ve kind of approach? So an example is capital modelling that we do. You know, our view is very much why why why hire a quant to do a particular type of capital model, and King, but you know, if you’re, if you’re a challenger bank, you know, it depends on the customer, if you’re a challenger bank, you know, you probably don’t want to keep a quote on your on your books full time to calculate operation was capital. And so so we, from our from our background, know how that is calculated and provided as a service. And the hope is that that well, that’s novel, perhaps we don’t we can reduce the cost of being a quant there and use you guys as as our as our team quants. With your with your model on that. So that’s reducing the cost base for a client. And and so then that’s, you know, that is just an example. And it’s similar with the contact stuff we’re doing is applying machine learning it was can we can we reduce the cost of contact, you know, the contact career reduce that? And the answer is, yes, we can and data, and machine learning helped us to do that we can. So I think I sometimes think that that is this. Let me just apply machine learning at all costs. But what is it we’re going to apply it to? Yeah, I think we’re still in that phase.

And so we talked a bit about as far as capital modelling so about capital, they talked a bit about you know, contact, contact propensity, talk a bit about vulnerability, financial valency. What do you think the next sort of outcome was like use cases where are we not looking that we should be looking to try and to try and leverage that to leverage those that kind of thinking, do you think other other new frontiers that are coming up because I do feel that vulnerability has been a frontier for a while, and it’s great to see that that’s now really being closed? Right? But it’s is that where do we get an extra thing?

I still I still, I still wonder, Are we closing vulnerability There’s a lot of people out there that’s doing vulnerability, but I, I don’t, you know, I don’t think I want until till the regulators can robustly challenge, you know, what really should happen is that is regulators should be able to robustly challenge and the claims that are being made, and they need data, they need data to do that, and stand up to, you know, so when a, when a when a bank decides to close a branch in an area that’s has a high propensity financial vulnerability, like stand up to to the bank and say, Look, you know, we are data assets, or, you know, our external suppliers, identified this area as a, as has a high financial vulnerability, you’ll want to close that branch, you know, it’s more likely that in that area, individuals are not going to be able to, to, to respond to branch closure, you know, they’re already close to, to, you know, asking them to travel 510 miles on the bus just to come and see you or asking them to afford the internet or, or buy an iPad, just to just interact with you. Some people are not going to do that. And those people, those people are going to become financially dispossessed in those. And so until the regulator starts to meaningfully challenge I don’t I mean, I don’t I don’t think we have solved it.

Yeah. So it sounds like there’s I mean, there’s still more to do, and it’s more granularity in the same topics. Right. So yeah,

I think that’s it. I think that’s it, there’s, you know, there’s more to do. You know, in terms of in terms of, yeah, the next steps, you know, you know, you would expect to see more kind of product type development, but I think, I think regulators in a certain amount, trying to question that in terms of the machine, the application of machine learning for all that.

Yes, was just touching on that, I suppose, in terms of like transparency on the mouse topic for a while, particularly around machine learning. I’m thinking on the lending side, right, which is a little bit different, maybe on the on the fraud or even the, the vulnerability side, I suppose, but to a certain extent, but it’s like, what, what’s, what’s the latest view on transparency? Mean? Is that, is that going to continue to be a topic? You’re making strides on that? Or is that the that concern sort of fading away a little bit? I think.

Like, the barometer for a lot of these things, is is one signals consultancy, for instance? Yeah. What are they doing? So they’re still I still see thought pieces coming out about? You know, the transparency? I think, I think, you know, like you say, when that is driving Indy, or when it’s driving capital, then then rightly, or even, even

financial negatives, as well as false negatives. Concerning?

Yeah, I think I think when it’s driving those types of models, I think, where it has a direct impact on the p&l, I think I saw audited and, and the capital that that’s required, I think there is a need for transparency on our site, on our end, we’re doing it as a, you know, we wouldn’t want to start to share details of how we do things, because it’s commercially sensitive and, and to a certain amount. What’s the transparency? Well, our transparency is, is the fact that we can deliver, you know, a benefit, and the benefit, if the benefit isn’t there, then guess what, you won’t you won’t bought it. So I think it depends on the application of the tool. It doesn’t need to be transparent across the whole industry or the whole application, it needs to be transparent on, on on, on those on those areas where people are, you know, making investment decisions, etc.

And I suppose it’s because that’s your heart and conversation, doesn’t it? So if it’s, if it’s value added insight, on top of what they would have before that prevents more harm than that’s probably you probably bring I be okay with that. But if it’s, if it’s lending, and you’re denying people, even if it’s, even if there are false negatives around vulnerability, and you get it, and it could result in harm, then yes, then then that’s, that’s where it’ll get more difficult, I think. Yeah, the other question that comes up is really around bias. Right. So there’s bias in the models and those kinds of things and just, you know, is the fact that they’re trained models, and if the trainings not been right, then Does that, does that introduce bias? I suppose they’ll probably come through at some point.

Yeah, I think so. And I think that’s something that particularly on our, our vulnerability, you know, on the on the Other things that we can demonstrate an uplift to certain amount. So for example, the contact, you know, to certain amount the uplift it, you know, we can demonstrate that uplift. So, you know, use use our product, and then you will get a game. And so it that those types of, you know, they are at consideration that those shortfalls are consideration, but perhaps not as immediate to say, the vulnerability piece. So, we wouldn’t, you know, we we strive to eliminate error and bias in vulnerability. And we do you know, we work with other with other firms to do comparisons. So, with the, with Helen, the VRS, to compare our data and their data, and broadly, we see the same, the same type of outcome. So that that gives us added comfort. So, again, we, you know, it is it is a consideration within within models. But again, the application of the model, was the model being useful.

Yeah, yeah. So it’s always nice, that human element, as I think it does, it does. And so just going to talk about about the debt sale market, I suppose. And pricing is a bit about capital there. I mean, we’ve had quite a bit of a shock, I suppose the last sort of 18 months, two years. And the data structure, the data over the last few years is not quite the same as the data it was the previous two years. And it might not be the same as the data that’s going forward, and you sort of see some of that change in some of your data. How much of a challenge Do you think that’s going to be for firms to get pricing right? To really understand what’s going on from a dynamics point of view?

You know, we’ve had payment holidays, we’ve had, you know, we highlighted the FT highlighted cases of less CRA data is going to be accuracy is going to be eroded by a wide pandemic. I certainly we’ve heard on on from various sources of kind of internal process failures that some of the some of the lenders that have mentioned that reporting back to CRA is is it practices accurate? Is it because of payment holidays? It could have been? And so I think there is a challenge for that type of, of modelling approach, depending on that that type of data, I think ours, you know, to certain limits, as long as we’re getting, you know, the foundation data we get is, is updated continuously. And then the outcome data, we were still getting outcomes through the pandemic. You know, there were points that were, you know, I think they’re in a very, we did see the drop off in outbound contact. But that that came back pretty pretty quickly. So, yeah, I think it depends on the data source. I think I think some are challenging. Some some probably going to

be and what about what about future things you’re kind of worried about, and we talked a bit about the economy, some of the economic impacts, I mean, anything that you think’s on the horizon that we’ve got a we got to look out for? I mean, obviously, you see changes in vulnerability or resilience. Right. I mean, that’s, that’s something that one aspect for sure, but is it what anything else? Or?

I think, I think, you know, we kind of touched on it earlier was the inflation. I know, it’s a supply, it’s a supply side type inflation. So in the news debate, whether it’s here to stay, whether it’s just a flash, but, you know, I think the assumption for firms is that, you know, this has been, and I think you’ve got to look at this and think and actually is now the time to invest, you know, we’re going to get through deflationary cycle, there may be now I should be investing in my, in my company, you know, we’ve had a little bit of stagnation, you know, around Brexit and through COVID Is now the time to, you know, to to increase the effectiveness, you know, invest in our operations, increased effectiveness of operations and an increased productivity, we’ve been through this period of reduced productivity, I think never perhaps is the time to, to invest and increase productivity, you know, ahead of

Korea. Could be inflationary increases. Yeah.

Yeah. Yeah. And I think that from the business side is the question is, is an avatar to do it? And I think I think I personally think it is,

I think, yeah, I suppose because there’s always a lag there between suppose inflation and prices going up. And then the cost if you sort of like invest now and the cost of capital is going to be longer so like doing it now is probably going to save you money in the longer term. Yeah, yeah. Well, John, thanks very much for the time. I really, really appreciate it definitely got some great insights on data and how to look at data as well. Right. So how to look at data and think about it and sort of build that into modelling. I mean, I think that’s, that’s, that’s fascinating. So, really appreciate the insight.

Alright, thanks for the opportunity. It’s good to chat.