Joffre Toerien discusses scoring for microfinance, and Georgia

We like to talk about algorithms, and big data, and machine learning, but sometimes we’d do well to take a step back and ask about what the business really needs. This is especially true in developing markets, where consultants can forget to bring the organisation along with them when they implement the latest and greatest scoring tech.

In this episode 5 of How to Lend Money to Strangers I speak to Joffre Toerien of Credit Insight Analytics about the lessons he has learned building scorecards for microfinance lenders in Africa and Europe, how he manages the balances between scorecard predictiveness and organisational inertia, and a little bit about lending in Georgia, a country of 4 million nestled between Eastern Europe and Western Asia.

If you have any feedback, questions, or if you would like to participate in the show, please feel free to reach out to me via the contact page on this site.

Regards,

Brendan

The full transcript with timestamps is below:

.Joffre Toerien 0:00

To put it in perspective, right, Georgia, the country, I would say is half the size of the state in America, in population and the size.

Brendan Le Grange 0:28

Welcome back to How to Lend Money to Strangers, the podcasts about lending strategies across the credit lifecycle and around the world. Last week, we were in China, a country with 1 billion people. This week, we're in Georgia, a country of just 4 million people, nestled between Eastern Europe and Western Asia.

I'm joined by Joffre Toerien, Joffre started his credit career with a furniture retailer in South Africa, before moving through Uganda and Tanzania to Georgia, where he's now the director at Credit Insight Analytics. Along the way, Joffrey has developed some impressive experience building models for microlenders in emerging markets. If you would like to work with him, you can find more at https://www.credit-insight.com/. Georgia... it's obviously a country that's not top of people's minds. I know it, strangely enough from a wine tasting in Hong Kong when there was a Georgian wine stand. But what is Georgia, like when you are lending?

Joffre Toerien 1:30

Yeah, so Georgia is a very old country. You know, countries that if you had something in the 18th Century would be really old, but here I mean, you have buildings that were built in the 5th Century.

But yeah, maybe just quickly, how I got here. So I joined a microfinance network of 20 countries, and I was responsible for scoring analytics in Africa, we stayed in Uganda and Tanzania and then there came a role as the global lead for credit risk - and that was between the UK and Georgia. And luckily for us, I wanted to do Georgia, and it kind of worked out, you know. Georgia is just very friendly with visas and work permits and that, so we came to Georgia and I would recommend that to anyone - put that on your holiday destination, it's very affordable, easy to travel and extremely safe as well.

But yeah, the 4 million really says a lot. And it's not like the population is growing. Think of DRC, you know, 80 plus million or, you know, Pakistan, 200 plus million population. So, really, people are a lot more excited, obviously, to move into those markets. It's a saturated market, there's 15 commercial banks, two of which control most of the market share. And, you know, there's 30 plus microfinance institutions registered. So if you think about it, 4 million people with 50 credit providers. So it's really tough market. And I think more than 50% of the workforce is related to agricultural industry. I mean, they're pretty sophisticated. I mean, all types of lending. Usually, it's loans for businesses, there's, you know, for cars, for home loans, or you know, everything's here. There's one credit bureau, but it's just a tough market. And since it's not growing, it's not a you know, it's kind of just competing for each other's share really.

Yeah. And I would have thought that it would really be quiet Russian in influence, but actually, when I looked at it, it looks like quite a mix of cultures, but also quite Georgian, I guess coming from being an old country. So when you look at the brands, and the companies operating there, are these regional players or homegrown or is there a mix?

It's very much home grown. I mean, the banks that flourish are usually the Georgian ones. And they're not in other markets, I would say. But there are, there are companies, like FINCA, ProCredit, that are, I would say, worldwide and also in Georgia. But even like, like FINCA, they've sold FINCA Georgia. And not that it's not profitable, or... it was just, it's just difficult to gain market share here.

Brendan Le Grange 4:04

So I worked in Denmark, which is 5.5million people ,richer, so more incentive to go., but even in a market like that, if you're there, it was profitable., but the cost of entering meant that it was very difficult. And you know FinTech there's always gonna be someone who's finding a new way around it. But I think there's like a goods and services ,or goods ,where you could just start shipping. You start out financial services, there's new regulations, new customer expectations. And if you've got to say, Well, now I'm going to go to a new country, 4 million people, and I have to pay all that money to get there ,to build trust. I can see why it would be tricky.

In terms of the risk profile, you're saying it's not growing very fast - is it kind of a stable market, under control risk, or in terms of environment - you've obviously worked in it in a few developing markets, and I'll pop back to that point in a minute - but what's the risk like, is it a high risk, low risk sort of market?

Joffre Toerien 5:01

No, I think it's pretty low-risk. People, and all the companies, I think are very risk averse. That's kind of what I've seen is, most of the banks, they follow the same approach. So I do think there's opportunity, you know, specifically, if you think of COVID, right, COVID would have impacted a lot of institutions, it impacts the credit bureau, right. So there's nothing to do about that. Right. So there were instances people couldn't pay it negatively affects your credit bureau. And I just see opportunity, for example, speaking to, you know, our company and just saying, you know, if everyone just goes with the bureau score, that would have been affected, you know, so we... that idea of finding some of these false bads, but I kind of, I kind of got the feeling, they're not keen to change the process. So I think it's a very 'everyone follows the same process' which give some opportunity, if you think a bit differently or outside of the box.

But it's low risk. And even in COVID, people want to repay, unemployment is hovers around 20%. Inflation is a bit high, is it just above 9%, but overall, the portfolio's they look good. You know, low par rates. So I don't think so high risk in that regard. And Georgia, interesting on the Russia influence, they're very friendly with all the neighbours and highly dependent on tourism from Russia and neighbouring countries. So they they keep a good relationship. So I think it's low risk. It's just yeah, I think the cost to enter is very high. Yeah. And also the, just the geography of the of the country, it's, there's like two main cities, I would say, I guess, 30% of the population is in Tbilisi, maybe Batumi the rest, but then the rest of the country it's spread.

And we spent the last week just travelling across the country, and you would see, you know, it's almost like summer houses or these villages, and in the summer time people would be there, but there's not much in the in the countries, but it's, it's an amazing experience such an old country that on the one hand, I mean, the city is quite developed, right? You see, you know, electric cars, the whole place, you know, you see so many Tesla's and all these things. So it's developed in that sense, you can pay for your parking on app, you can order food on an app, you know, Airbnb, all those things are here. But if you go to the, to the rural part, I mean, it's proper rural. So it's an interesting, interesting, dynamic. So even even us, we stay just outside of the city also in a village, but it's just outside of the city and our neighbours have cows, right, and everyone makes their own wine. And it's like, an old style of living. I dunno, I like it. It's it's just a combination of really old history. And... I don't know how to explain it, but it's definitely something to see. And I think it's a interesting place to visit. So just just from my perspective, it's difficult to place. It's something that, you know, well, I wasn't used to, and I think it applies a bit to banking as well. And the way they do things, it's, it's a bit different, right. And you can see, working with Georgians that have worked in the UK or Germany, you know, you could see that they have not the traditional approach.

And I think it's nice that they still in the world, these pockets where it is still more unique, you know, 1000 year old country is still keeping its own...

Brendan Le Grange 8:22

And I mean, you tasted the wine, right? There's two ways of making wine: one is, you know, the European way, which is just the way that everybody knows, and then there's the Georgian way, which is, you know, it takes it takes longer, takes like nine months and creates very different wines. So usually that would be our gift travelling, we would take Georgian wine. Yeah, they, they have their way of doing it, they do it like that, then it works. You know, obviously, they they love to say that they were the first country that made wine and you'd always think, no, that must come from France or somewhere, but apparently, it's proven.

Joffre Toerien 8:56

Yeah, I didn't realise that, as I said, we sort of stumbled upon it in, in Hong Kong of all places, one of these world food and wine fairs, and there was a tent of Georgian wine. And I was there with my friend from Moldova, and there was some Moldovan wine as well, so we were just kind of tasting them. And they hadn't quite explained that story. But yeah, I think it's really nice to to hear those stories that don't make the headlines.

Brendan Le Grange 9:20

If we go back a step, then. You started your career in South Africa, which is, yeah, we've obviously also got a lot of developing to do and there's there's big sectors of the market that are left out in many ways, but for all intents and purposes, a pretty developed credit world, at least within the big banks. But from there, you took quite an interesting path and you got into microlending.

After university... so my background's actually actuarial science, and I had a friend that worked in in credit and I just, I just didn't want to do the insurance, the normal, you know, route, so I joined a furniture retailer in South Africa and they paid my study debts. So I had to work for them for three years, I thought it was a great experience. And, and I ended up working eight, nine years for them. And I learned some valuable lessons just because they were an operationally run business, right? So I would come with my statistics and scoring stuff, and, and they would just ask practical questions. But how does this work? How does that work? And oftentimes, I didn't have the answer, right, or just like, I mean, this is what it is, you know, and usually, if you speak to, let's say, credit, risk director, you know, people with your background, they would kind of, you know, they would just go with it, right, whatever you say, gini and, you know, and I realised just the importance of 'you have to do sales'. It's like the branches exist to serve the clients, and the head office exists to serve the branches.

Joffre Toerien 10:51

So that mentality helped me extremely a lot. After working for retail creditor, I was a consultant, I was in unfortunate, fortunate position that, you know, the Scorex guys came to South Africa, started scoring and these guys obviously worked with, almost want to say developed it with, you know, not laptops and all these fancy programmes... I mean, a lot of it was done, I almost want to say, by hand, but really understanding it. And I would say that generation, so, that generation then taught us how to do it. And then obviously, all these software came in, but also having a good understanding of, you know, how it actually works, you know, would you be able to do it by hand. And I think that was also a great experience. I worked as a consultant then and worked for most of the banks, and then retailers in in South Africa.

And then there was opportunity to do some work for a company in Uganda. Yeah, I was just, we were just married. So me and my wife, we just said, yes, we want to go. And so I joined FINCA Africa. And it was an amazing experience. What I learned there was, you know, joining FINCA, you know, usually as an analyst, you would come in and just really connect to the data and, you know, do your thing. But joining FINCA I realised a lot of these databases and things. They're not in place, right. So now it's more a process of how do we make sure we get data in place to be able to do it. And so that's how I joined Finca with six countries in Africa. And it was also amazing experience and what I really liked, and later, you know, FINCA's got 20 countries, and just the diversity. So some countries would be NGOs, otherwise, others would be a microfinance bank that's not allowed to take deposits, other could take deposits, here in Georgia and Kyrgyzstan, its commercial bank. So I really liked that, it was just, you know, just extremely different scenarios.

Yeah, it's under one umbrella, but you are basically a consultant in each country starting afresh. But usually, as a consultant, you have the advantage that there's a client who's there to make things happen, you know, try to find.. has to find how the regulations work or how operationally a branch works, you can hand that off, but because you're inside, yeah, that's your job as well. So you don't just have to teach someone how to build a scorecard or tell somebody fetch me this data and I'll build you a scorecard. You've got to know how to build a scorecard and get the data, and I guess, yeah, each country as you say a different, not just different countries, you've got a whole different organisational structures, different banks, different rules. So yeah, I can see why it would be something that would keep you interested for for a while.

Brendan Le Grange 13:34

I almost want to say for a lot of companies, the data available is more than the skills, they have to use it. So we would support the subsidies, and you can obviously influence global policy, you know, but these things take long. But yeah, it is kind of working as consultants, so it's also going to subsidies and, you know, not having the authority to just tell them what to do, but also obviously influenced them and you know, sell the idea, you know, we found some subsidies that were comfortable, and, you know, you could just get a lot of more, you know, a lot more done there than others.

Joffre Toerien 14:06

But yes, they that also gave me a lot of experience as consulting working with different different organisations. But really, also something I learned with that is to understand the data you have, and also learning what can be done and, you know, setting realistic expectations. So for example, some projects it would be you know, scoring for new clients, but certain data weren't available and you know, just knowing what would work and then understanding what's the requirement and then just having confidence in saying it can't be done. So if you want to pay these other consultants to do it, by all means, do it but in six months time, you know, when things don't go as expected, you're gonna call me again and, and that happened a couple of times, right? Which just gave me confidence to understand basic principles in scoring on credit granting, that, it's all like gravity, right? It's just these rules apply. And if you don't have these things in place, it's not going to work.

So oftentimes it would be going to a country and say, Well, you can't do that now, we need to put these things in place and that's going to take, let's say, 15 months, but, you know, let's do it. And then, you know, in 15 months, we can we can get going with what you wanted to do now. And oftentimes, you know, would be, you know, people want a solution now, but then 18 months later, you know, you think, if we just did the basics, right, we could have been in a position to do that. So that was an interesting experience. And, I mean, I met remarkable people all over the world. And I mean, it was really amazing experience. So I'm very grateful for that.

Very broad strokes, and perhaps a bit unfair. But if you think about lending in some of these markets, it might be easy for somebody in a developed market to think the closest business in our market to that would be something like the payday lending: you've got payday lenders that are doing small value, high risk loans, sometimes to populations with limited or patchy data. And I don't want to besmirch the whole industry, but often fairly predatory. At least their history has been. But in fact, you take a very different approach, even though, yes, it's more value, yes, it could be quite high risk, yes, the data might share some similarities. This is very much not payday lending. I mean, you've talked about prioritising these loans being beneficial to customers. So can you talk a bit about how you think about lending in these markets, that might be different to what you would do to a high risk population in a big developed market, like the US or even South Africa?

Brendan Le Grange 16:45

Yeah, so well, you know, working with, with companies that, you know, have this mission driven, oftentimes would say, a double bottom line, right, looking at financial inclusion, gender diversity, that puts a different spin on it, you know, and oftentimes, there are donors and funders that would contribute either technical assistance or funds towards it. So I mean, that that's very helpful. Maybe just a quick example, you know, some type of lending, it's not profitable, specifically, something like group lending, usually those type of lending, it's not, it's not profitable, and it's very difficult to manage and to scale. So what I like about that is, if you can get funding or support with, let's say, a product like that, it's not profitable, but you see it as a cost of acquisition, right, where you join a group, these group needs to meet, you know, weekly, or bi weekly, and it's part of education, and they pay off the loan. But once that's done, you need to migrate them to an individual, right, and that that loan would be, you know, profitable for for the company.

Joffre Toerien 17:46

And, you know, if you think of Responsible Lending, it just needs to be win-win, right? So it would be it should be a win for the company first, and then one for the client. I mean, if the company's not profitable, that's also not responsible, right, to keep deposits, etc. So yeah, the, the Responsible Lending part, you know, I thought about that. So it's really not that the scoring or anything would be different, right? It's not that you use different variables, or you you exclude other variables or anything like that, when it comes to the scoring, you're trying to develop the most predictive model to predict risk. And I think, when it comes to the Responsible Lending part, it's really the whole credit underwriting process to have that mindset. So, you know, for example, you develop the scoring to measure risk. But oftentimes, it's different calculations, you determine risk, and then you need to have a separate calculation that looks at affordability, you know, capacity to repay, you know, let's say there is a credit bureau. But you know, a lot of times, depending on the quality of the Bureau, there's so many informal lending going on that you need to take that into account. So that idea of if you get credit, bureau information that needs to be taken account, but also living expenses, you know, you could have an average percentage apply to school fees, and rent and these things aren't on the credit bureau. And then maybe just taking into account informal lending as well, those kinds of things. And then there's a third calculation where you calculate the loan limit, right, based on the risk based on the affordability, you come up with a what amount, can we give this client that he would be able to repay?

Because I mean, Responsible Lending is really just you need to take all the reasonable steps to ensure you have a pretty good picture of this guy's financial situation. So I would say it's the same. The scoring I just think the company with that double bottom line, it's easier to keep everyone honest and just say listen, but we don't just have a goal of making profit, obviously, it's one of them, but let's let's look at the impact of that. And also thinking of not just one product, right? So a lot of times if clients have had 10 of these loans, are they in a better position or not, you know, with with a company like FINCA you have luxury to sit and discuss those things.

The sort of person that's going to lend in the highest risk market traditionally would be funded by somebody who wants a big return, and would have this profit pressure, and possibly also just from the ownership team, a lot of drive to grow, to be the next unicorn, you know, to get the headlines. If you think about a bank, it has really cheap capital, it doesn't necessarily need that same return, but the bank is also probably more conservative and going to stay away from markets. And what you're saying is, it's kind of the ownership of these companies, the way they raise funds is a bit different, and so that mindset's there, the work you're doing is the same but that pressure to say we need 50% return on equity every, every year, that's gone away.

Brendan Le Grange 20:42

Yeah, Brendan, maybe, something to mention just on developing markets, I think it's, and it might be surprising to, you know, some people listening that, you know, might be, in the UK or in these markets, but something I realised and it's, it's something that's worth noting, is usually, I mean, your your show is Hot to Lend to Strangers, right? You wouldn't consider somebody that's existing client or previous client a stranger, because you've dealt with them. But that's oftentimes how it works, right? If your new client or existing client come in for a loan, you follow the same process, let's say if there's some scoring, it's the same, it's the same thing, right? So that's something I realised with just all clients that are given loans, any company that's got a core banking system, right, they have to have something just for regulatory purposes and finances, that data is very valuable.

Joffre Toerien 21:34

So that was that was my focus point is, if you've got nothing, that's where we start. And the type of data required, it's really a monthly snapshot just of the portfolio. And that, in any company that is available, right? It's used by finance. And then if you can get these summary snapshot of all your clients for the last 18 months, that is extremely useful. I mean, you've got experience working with bureaus, right, and the Bureau's get the same data, but they create different variables, right. And that's usually the differentiating factor between the quality of the bureau's and the same thing there. So if you've, let's say, the last year or year and a half of these files, it's extremely predictive, right? So you can see the trends, it's very easy to test that as well, usually, you know, you just need key variables to work balance the rears, you know, and usually we'll see the type of products, and that that data can be used to create variables, we call it behavioural scorecards. And it's, it's very intuitive, it's easy to test. And it's difficult to test new clients, for existing client, you can just go with, you know, the the Chief Operating Officer go to a branch, have your scoring, talk to the loan officers about the clients, they know them, right, they you'd surprising how many they have, but they know them by name, and test the scoring. So from my perspective, I can test the stability and the performance of the model, back test it back and forth. But usually, we test it in the subsidiary, in the branches, more to convince staff and to convince management.

And that's usually the place to start, right, that data. And it might seem traditional, but oftentimes I deal with, let's say, consultants coming in. And those basic things, they don't know that, for example, they would try and develop scoring for all clients, new and existed, right. So that's a key segment where you would always handle existing clients differently, just because you've got much more predictive data. It's also useful to educate, you know, upwards even, management teams, get them used to scoring, so that they also understand the benefits, but also the limitations, right, the limitations of scoring, once you get that going, got the customer data, you can score the entire base end of the month, first of the month, you can use it for you know, retaining the best clients, pre-approvals even. And while you do that, obviously, as a developer, you will get to know the market, the segments, the branches, then you can start looking at scoring new clients like perfect strangers...

As you said, you've got all this data on existing customers, you know how long they were with you, you know how they spend when you're with them, depending, yeah, you might have deposits/ you might have investments, you've got this big pool of customers, you've got so much data on, they must like you in some way, because they were a customer of yours - maybe you need to update your brand, the colours might be wrong because they're 20 years old? You can't look at it and just say, okay, here's a scorecard and we fixed your problem. Or we're going to bring in some consultants or build a new scorecard and the problem will be solved. You do need that person to sit down and say, well, this is what a scorecard can do, this is what the data can do in this bigger context.

Brendan Le Grange 24:38

People wouldn't even think of it as scoring it's just, practically, if somebody repays you give them a loan. I mean, it's it's almost as simple as that. But it's it's very useful to show a company, how many clients you give loans and they settle and they, you know, then they never come back. And a lot of times there's not a focus on that - to make sure we retain the low risk clients. And it's much cheaper than finding a new one.

Joffre Toerien 24:59

But yeah, so If you can talk a little bit about some of those approaches you've tried, and maybe some of the data you've looked at that didn't didn't work in the sort of market be really interesting.

Brendan Le Grange 25:09

Maybe I'll start with just for new clients. So, obviously, credit bureau data, if it's predictive, that's, that's the best way to go. If you think of lending, right, just break it down to first principles, you have to ask the client some questions or gather data, right? Either you have to get it from the client, tells you, you know, things, which obviously, you're gonna have to verify some of them. Or you can get the data from external sources, like a national ID database, or fraud database, or blacklist database or any, but if you have nothing like that, then you really just depend on what the clients tell you - or you're gonna have to do a manual assessment, like a business analysis, which is also costly.

Joffre Toerien 25:48

But I think in some of these markets, you'll always require something like that if you don't have the credit bureau data, you're gonna have to do this type of analysis, you know. I see, and especially micro finance, big portion of the loans will always have to be reviewed. So I think it's important to have an idea of 'central underwriting' or having an underwriter team. And the ideal situation would be whatever information they need to make a decision should be in front of them in a digital format. And then you can track that, how many deals do they see? And what's the decision they make? And then we can track their performance. So I mean, that's, that's key to the to, I think, to the process.

Now, obviously, people want to automate it. What I've seen, when people say we use, you know, cell phone data, or social media data, to give out loans, you know, I haven't seen that being done for $1,000 loan. So it's usually, you know, it's usually smaller. So if if somebody comes with a success story, saying, we've given our $200 million of loans, you know, based off of these 2,000 variables, it's good to ask some of these questions, right to understand what's the product? How big are the loans, how does itwork? And especially the collections on it, right? So for example, if it is a mobile money, and it's based on transactions, and you just give everyone $5 loan, if they repay you give them $7, then, and if it is that it's a mobile money operator, that, let's say it's the main one in the country, right, so there's not five others, then there's some stickiness to it, right? You can't just drop your sime. And if if they have automatic deductions, that as it's your due date transactions come through the wallet, they subtract it, that product will work, doesn't matter what the scoring is, right? If the scoring were good or bad, that product will work. So that's important as well, with sometimes with the success stories.

Yeah, and I'm not against alternative data sources. I'm keen to use check them all. It just sometimes some of the companies, they get burned by, you know, people promising things. Or even if somebody says, you know, we have 2,000 variables, and we use this and that, you know, it's it's oftentimes frustrating, but also just disappointing, right? When I ask questions, okay but explain to me... because usually scoring models would have 8 to 15 variables. I mean, I can go with if somebody says it's got 30 variables, if they can show me the weightings and correlation and some things, but there is no way you're going to develop a model with 200 variables, 2,000 variables - I mean, monitoring would be impossible, implementation would be impossible. So oftentimes, when I ask the questions to get more down to it, it's, it's really not what they say it is.

But I am interested. And maybe the other option is if you've got your own app, working with a mobile network operator, and having our own app where we can see transactions. And I almost want to say applying the same methodology to both, but just the one is clearly successful, and other not. So usually, with the mobile network operators, they're not always willing to just cancel everything, right. So usually, they work with banking partners, right? So it's mobile-led but there's a banking partner behind. And usually the relationships not that 'if you don't pay your loan, we're going to switch off your...', you know, we're not going to cut our own throat. But that makes it difficult, where what I've seen with with mobile loans that work well is if you if you have something transactional base, and again, if you can base the risk on the transaction, so almost like a salary, right? Think of if they want your three months bank statement, they want to see your income, right? So similar approach, you can do that. So you can calculate like the average inflow into a wallet as income. And even use that to calculate risk, meaning if it's a stable income, increasing, decreasing, and if you can set it that deduction is done from the inflow. I mean, that's very low risk. And we've seen that that works very well. So that's one way but obviously, you need the customer to join you, to at least transact for three months so you've got some history, so it's not scoring a brand new client, but that principle is useful with agent networks based on the transactions. Those type of scoring based on transactional, it's it's very predictive and unusually low risk. So those would be, you know, ones that definitely work. And yeah, then just doing the doing the financial analysis with underwriters, but having the controlled underwriter almost like a call centre, but you have a team, you can educate that team, you can monitor their performance, I think that's the best way to go with bigger loans.

It's just the lack of solutions in that technology, right? So I would find fintechs, or companies would come in with, with some solution, but it's really just addressing one one of the capabilities that require, let's say, you know, taking it from paper base, you know, like digital applications, but then usually the software is not equipped to, you know, do some basic rules, or you can't implement the scoring in there. So oftentimes, it's, it's not one solution fits all, it's like you need the processes in the, in the company, right, those workflow processes, there's always going to be a need for the branch to go and verify something, go and check something, right. But then they need to have the ability to, you know, come back to the browser or have a tablet to say, I verified it, yes, and then the process continues. So we know the Bureau's they have decision engines. And then if you go to the developing markets, they don't even know what it is or why they need it, and you know, it's got a hefty price tag, right, 100,000 plus licence a year type of thing. Because the challenge I found was, you would go to companies. And if you, if you look hard enough, you'll find data, right? So you find data, especially on existing clients. So you develop a really predictive behavioural model but how do you implement it? Right? That that was my challenge a lot of times, and then you end up, you know, doing an Excel thing, right. So you have a Excel with all the customers, their score for the month, their loan limit, pre approval, but then it's like an Excel thing, you know, and it's easy to understand that Excel can just be in a database, and you can have systems that reference it so it doesn't have to be Excels, but just to prove that it works, you can just run the score, give an Excel to the head office, give an Excel to each branch, and let's test it and usually that, that that works, but it's not sustainable, right, you want to have control over over it in a, in a better way than Excellent.

So that's something I'm focusing on a bit in the coming months, partnered with a with a tech company to see if we can develop some credit decisioning that's just affordable. That's the one thing.

In my first episode, I spoke to Raymond Anderson, and he did a project in Kenya, then we said, yeah, they built the scorecard. And then the company realised they couldn't implement it. So I'm not sure with your experience, but the brief experience I had in Ghana, where we worked with this commercial bank the first time they were going to consumer level loans, they had, you know, each person in the branch could approve up to a certain amount, if it was above that it had to go to the branch committee, and above that it had to go to the regional committee, and above that to head office committee. And it was like, if we could just put the rules that you've got now, forget about a scorecard, because that's a scary word to some people, we're gonna put the rules, there's we're gonna write them with you in here, we can look at the paper, you can check your comfortable to, like you mentioned, like this will give ourselves 18 months in the future, we'll build a statistical model but for now, the benefit comes purely from the fact we can see what's happening. And you can automatically escalate. And the same people can do the same job. But it's controlled and we're getting data. And sometimes that's going to be a better approach than having a million dollar scoring project that gathers dust.

Brendan Le Grange 33:51

That would be a big one. But it's not easy to take these branch trade committees, right. So what I learned is you have to take them centrally, but what you do is you have to take experienced guys in the branches, and you take take them to the head office. So at least your branches know, okay, they know these guys, so just instead of having the committee here in the branch, they're having it somewhere else. And and we have control over that. So that's the first thing and then just understanding what they are doing. And a lot of times it would be going with the attitude of, you know, especially if the company is doing well, you say you guys know what you're doing so we want to understand what you're doing. And usually it's rules based, and that's the first thing that you try and do, is just track what happens, right? So oftentimes, a lot of these companies, they wouldn't be able to produce a report of how many applications they had. It sounds funny, but you would see everything that's disbursed, but you don't always know who came through the door. So you don't really know the true accept rate, decline rate. And even talking about you know, marketing, you don't have ideas and you're trying to link marketing and disbursements - we need to link marketing and applications etc.

Joffre Toerien 35:01

What I try to explain sometimes or get this idea of let's let 100% of decisions be made by the underwriters, right, 100%, or these committees, usually then call them underwriter, and then slowly, we figure out what they do. And then the bottom 5%, you know, based on just rules or scoring, whatever, we just decline them, and we make sure the guys are comfortable, you know, usually they would have declined them as well. And then you start on the top end, let's prove this 5% that just the best of the best. And usually, you know, then the underwriter would say, Yeah, okay, and usually they're 100% aligned. So then you have 10%, that's automated, 90% is still being done. And then you just over time, try and get that gap closer. And when it comes to new clients in developing markets, I don't think you can get to to less than 40%, I think 40% will have to be done by these underwriter. But at least you can have efficiency gains of the, you know, let's not spend time on the worst ones and the best ones, because if you just go from the manual approach to automation, that never works.

I will also tell people that the reason we can use the behavioural scoring on the existing clients and even give them a another loan without further analysis is just the fact that we know the initial analysis was done. Right. So I think there's value in doing a proper analysis the first time and then based on that you can just continue, you know, maybe check the bureau if you can, or just do the affordability assessment.

Brendan Le Grange 36:34

But yeah, thank you very much. I think we've already run over a bit, so is there anything I've missed?

Joffre Toerien 36:39

Just something I'm interested in? You know, obviously, you see all the blockchain and the crypto stuff, which is sometimes irritating, because they just think of, you know, price increasing and some crazy people in that field. But something that I do find interesting is some of the projects talk about actually real world problems, right, creating a digital identity with blockchain. And that's something I can get behind. If that happens, specific project is Cardano that I follow it. No, it's that could happen, right? That would be very valuable data right. Now, I don't know if that's 5 or 10 years away. But that's something I like, and those guys are focusing on Africa. They've got something in Uganda, and you know, these markets they used to mobile wallets. Like in South Africa nobody understands mobile wallets, right. I mean, it's just it's more useful than a bank account, right, or a mobile banking app.

Brendan Le Grange 37:32

Thank you, Joffre. And thank you for listening. This has been How to Lend Money to Strangers, and we will see you next Thursday.

Joffre Toerien 37:39

Cool. Thank you, Brendan for asking me.

I'm impressed that you had Raymond Anderson - when I started working, I found his first book very, very helpful. And oftentimes, you know, I recommend that book, just because it's so thorough.