Agile decision systems for modern lending needs, with Dmitriy Wolkenstein

You can't afford rigid systems in a world that moves fast, but also, for a long time at least, you couldn't afford agile systems in a more dollars and cents sort of way. But these days you can.

Timvero's solutions are all laid out at https://timvero.com/ or head towards a demo by jumping straight to https://timvero.com/request-a-demo/

You're welcome to start a discussion with Dmitriy over on LinkedIn https://www.linkedin.com/in/dmitriy-wolkenstein-7a19b79a/

And while you're there, come and find and connect at https://www.linkedin.com/in/brendanlegrange

As mentioned more than once in this episode, my action-adventure novels are on Amazon, some versions even for free, and my work with ConfirmU and our gamified psychometric scores is at https://confirmu.com/ and on episode 24 of this very show https://www.howtolendmoneytostrangers.show/episodes/episode-24

If you have any feedback or 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.

Keep well, Brendan

The full written transcript, with timestamps, is below:

Dmitriy Wolkenstein 0:00

First of all, most of the brick and mortar banks just don't understand, at a granular view, which products and which segments are really making them money.

In general, they can tell us, sure, but they have a lot of different customer segments, right, different products and sometimes they struggle to understand where they need to adjust.

So in this sense, our advanced analytics is helping banks to understand how does existing products work, and then give a different insight and suggestions to risk, product, and marketing departments.

And also to financial departments and to their stakeholders, in terms of what's your business look like if, for example, you would say we'll use this newly proposed risk model. And basically all of these just allow them to be more agile, and to run business in the more well controlled and data driven way.

Brendan Le Grange 0:55

I've never been to Dubai, which on the one hand is not such a problem. It's not really my vibe: I don't like luxury, I don't like shopping, I don't like extreme heat. But in my history, it's still a little bit of a missed opportunity.

You see, very early in my career before I'd done much travelling at all, I was put on a project that would have taken me there, it would have been my first time flying in business class, it wouldn't be my first time staying at a luxury hotel. I think the one they used had an indoor ski ramp. But unfortunately in the week, running up to the trip suddenly happened in one of the other markets and I think I found myself in Gaborone putting out fires.

It was not to be.

But more importantly, it really should never have been. The reason behind that trip was that the bank I worked for at the time had just bought a state of the art collections decisioning system, all the bells and whistles, from a name we all know and trust to this day. However, to get it to operational condition. It needed collection strategy managers from multiple regions to meet together in person for a week to discuss the parameters we would insert into it and what we could and couldn't live with when we went home. In those days, systems were rigid, and it was our job to work around them.

Luckily, that's not the case anymore. Welcome to How to Lend Money to Strangers with Brendan le Grange.

Dmitriy Wolkenstein, CEO and co-founder at Timvero, welcome to the show.

Dmitriy Wolkenstein 2:37

Thanks for having me.

Brendan Le Grange 2:38

It's a pleasure. And Timvero OS is an advanced lending operating system that includes originations analytics and servicing technologies. So a real meaty discussion for us for you know what this show really covers. But before we start talking about the lending aspects, let's whet our appetites with a bit of a starter. What was your background? What were you doing before you founded Timvero

Dmitriy Wolkenstein 3:02

So, before the founding of Timvero, me the same as most of my colleagues and also on the two co founders, Anton and Anton, I personally have been working in investment banking, I didn't really like this job. And I believe that it's also impacted the current state and the current unique selling propositions of tomorrow as well, you're working with stocks with factories, and you have your Bloomberg terminal, you better understand the impact of all of the available data in a specific moment in the market, more or less, of course, and you can utilise this data to properly assess their specific values and stocks and company and so on. So I did like this than when we started this company to remember company.

And when we started to dive into the lending business, were really wondering why the banks in terms of the lending business, utilise all of the available data not so efficiently and really ramus low.

Brendan Le Grange 4:07

Yeah, it's quite interesting, because, you know, I came from the other side, starting in consumer lending, and we always thought of ourselves as being much more data efficient and much more database, then corporate lenders, and we compare ourselves to them using, you know, financial records from years ago.

But you're right, you know, when you think about stock trading, it's literally up to the minute. So it really is a reset of that baseline, you know, if we move towards the Tim vero story, it was about five years ago, almost to the day going on my quick LinkedIn search that you launched in Vero as a modern day lending system. Now, I've recently spoken to a few different teams that are building yet more agile decisioning systems. But in 2018, you must have been quite ahead of that curve.

So what was that gap that you saw, or is the motivation that you felt?

Dmitriy Wolkenstein 4:59

Yeah, that's a great question. Back in 2013, we started this company and like the main reason of starting this business and being extremely honest with you, it was not to change the world, the only one reason that's behind this, is just that we were about to make some money.

And I don't know how but just because my personal background towards really related to banking industry, the same as my co founders, and coming to the different digital players fintax Digital banks is when we work with, with her to to dive deeper into the problems that they are facing. And even like five years later, this problem still exist, that a lot of digital players did not have sustainable renting businesses.

And that legacy players, that's half their opinion, but they also did not have any any ability to evolve with this lending portals and with this lending businesses. So basically, it was a little bit more organic in terms of how we got to this business. Because we started from a career development, we've worked with some folks in the banking industry, and then we just started to dive deeper into their business. They all basically tell if that's that's a huge problem when you want it to lounge because first of all, you don't have any data, you don't have any understanding from where to start. And then you don't have really good tools really like working perfectly tools, which is going to help you to create this, as we call it, like continuous improvement protests.

Once you launch something, you receive the feedback from the market. And then you need to use this feedback as a market response on these products. Risk modelling, which we expect with the GATT and so on, so on. And perhaps in most of the chances you also need to make some pivots, it's very slightly chances that you will get this right from the very beginning. And unfortunately, they just don't have the skills, you need to buy some really great origination platform from one vendor, you need also to buy a lot of really great tools for data information and data collection and data warehousing when bound or endorsed, you also need to set up a separate team, which is going to be making all of the integrations and trying to make it work smoothly. So yeah, it was like a lot of efforts. And unfortunately, banks, in our opinion wasn't ready to go through this.

Not the same, but a little bit like a comparable story was with the brick and mortar banks. They do have lending businesses, but unfortunately, they ain't struggled with this continuous have proven robustness and continuous improvements of what they offer to market. Because first of all, most of their brick and mortar banks just don't understand on a granular view on a granular way, which products and which segments does really make money. In general, they can tell us like we have one bill in in deposits, we lent one bill 10% return, but they have like a lot of different customer segments, right or different products. Sometimes they struggle to understand in which produce segments, they need to be focused more and where they need to adjust.

So get one diktats the same view on analytics on the financial analytics and financial performance. As we do have right now, we really learned a lot while still understanding how they calculated this calculation of cash flow for each product. And for each segment. It sounds like very basic, but it's extremely useful. So yeah, right now, we wanted to offer this to a lot of banks, which is struggling, like even training to 83 with understanding the specific cash flow specific profitability, which segmented product.

Brendan Le Grange 8:46

And Dimitri, what was it like when you try to start selling these systems because I always think back to somebody, one of my MBA lecturers used to say that nobody ever got fired for buying IBM.

And the world of decisioning systems is somewhat like that, where we have one or two of these household names, often, you know, the head of decisioning grew up learning to code in their particular system. They've seen two or three generations of those systems at the banks and lenders I've worked at, and there's this heavy bias towards one or two names we all know so well, when you were trying to get with the procurement teams or with the buyers, and you're a new startup, you've got this great promise of more flexibility of more insight, but you don't have that 200 year history. How did you find that to actually get this launched and get people to buy in and trust you?

Dmitriy Wolkenstein 9:38

So that's the biggest struggle in the case of Timvero 'decisioning' is broader than just risk underwriting or product underwriting. So in our case, this is primarily related to executive decisioning.

What does this mean? This means like decisions, which is making by business departments and state equalled some business owners in terms of which products to Lounge, which risk models we should to enrol, and so on. So in this sense, our advanced analytics is the more like the framework, which is just stamped on top of the origination and servicing and helping beings to understand how does existing products work, and then give it different insights and then give a different suggestions to risk product marketing departments and also to financial departments and to their stakeholders in terms of what's your financial centre, your business look like, if you would say use this newly proposed risk models which our AI engines are created.

But we're not replacing big players like FICO experience, we still utilise this data. So beings can build the existing risk models in our tools with these classical data sources. We also help banks to utilise extra data sources such as payment information like tin plate blades, like Yelp, elite, and some solutions like for SMB, let's say like some extra information from a QuickBooks or from Amazon, and so on.

We just allow banks to connect all of these data sources, implement their existing risk models, and then work on their existing practices.

And then this analytics and these frameworks, just helping to gain new models, which has got to be better in terms of performance in terms of the quality and so on. Anyway, Regarding your question on this Rygel, sort of selling this technology, yeah, step by step. Most of the big names, like modern big names, like let's say men who also struggled with this, of selling to be guys, but their playbook was really, really useful for us. So you're trying to sell to small midsize banks, fintechs and your delivery platform to deliver your solutions. And these being started to gain some profits, and started to feel all of these impacts our solution can give them, you have the numbers, you have the real numbers, you have the real use cases, and then the big guys will follow 100%

Brendan Le Grange 12:08

It is obviously resonated. Because I see you are delivering solutions in 45 countries around the world with big players, small players, medium sized players, people in all different niches at the right time as well, I guess that there's a bit more appetite and understanding of the value of not just having data but being able to react as you say, like actually use this for executive decisions rather than some statistician hidden away in the back corner building a model.

Dmitriy Wolkenstein 12:36

Yeah, that's exactly the point, just not having data. But like having ability to analyse this data and react. First you need to have this data in a consistent and normal form. And then you should be ready in your system should be ready to implement these innovations.

So why we also are can be used in like so many countries, just we create our technology more like framework style.

Because we see a couple of trends. First of all, there are more and more banks that are trying to also have their labour departments to try to make a lot of customizations and fine tune the solutions for their specific needs. So when we have our solution using very, very popular technologies, such as Spring framework of Java react Python, not like some of the players also on the market does use Salesforce, and these are great, but unfortunately, the price to customise the Salesforce solution is just like way more higher than for Java developers. So they can do all of these customizations internally. And they can customise this framework solution for their custom niche needs, and also protect some of the local needs.

Brendan Le Grange 13:44

Yeah, and that framework is made up of, of different integratable segments or modules, I guess, for loan originations for loan servicing and for analytics. So before we get into those nuts and bolts, I just wanted to say that one of the things I loved and I think it underlines this point you've been talking about is that when you go to Timvero.com, you immediately met with some business numbers associated with these different modules.

You see 100 seconds time to money in originations 80% cost reductions through better loan servicing 20% increase in profits with better analytics. This is not a system that's been sold because it can process X many million transactions a second or talking about Gini coefficients going up from 60 to 61 or something. This is talking business numbers and I love that.

Dmitriy Wolkenstein 14:33

Thanks for noticing this because yeah, that's exactly the problem I believe a lot of business leaders facing because a lot of business leaders in banking they do not have this technical background.

So once there guys from the risk department's bring them the new model, your model, unfortunately, for a lot of guys, the information about the improve gini co-efficient got to be really useful. So anyway, they will be happy to see some financial projections, they can easily do this, and they can start their business lending business really quickly. And yet, since we allow beings to customise all of this stuff, that's just that average.

But you're not only analysing, you're taking this data, and machine learning powered tools, which is right now working on top of Amazon sage maker, help refund departments, product departments, marketing departments, create new models, and analyse new models and compare them to the previous ones to the existing ones. And then the second layer of analytics, financial engineering tool, as we call it, it helps to push these projections from these departments into the understandable for stakeholders view, and they can understand what's going to be the potential business impact of all of these innovations.

And all of this can be done just in a matter of one week.

Brendan Le Grange 15:55

I think one of the big things there is that feedback loop between analytical data and financial data, if I think to my career, we would have analytical models, on the one hand that might be relatively responsive in terms of certain bad rates are associated with certain profitability for a customer. But those were from a modelling point of view, they were quite different to the actual accounting numbers of the business, which was done on this annual budget cycle. And there was only once or twice you could make adjustments to that.

And that feedback loop. I mean, you said six months, I think often it was 12 months between when you could really see how the models are operating and how it's impacting the actual business profitability to have them brought together as a massive leap there. But first, let's maybe do it in a more sensible order and go back to originations, talk to me in kind of nuts and bolts terms. What are you doing within originations to help speed this process up insurance

Dmitriy Wolkenstein 16:53

And as for originations, we are covering end to end origination automation for consumer lending business and for SMB lending business starting from when the customer can apply. We provide all of the modern channels, of course, this is online channels, Mail app integration with web applications, Ultra, we've got half a vendor portal. So each of the banks which is using our technology might lounge, they're owned by an open later instalment business, because they can give these API's to the of the central portal their partners. So the application of automated underwriting, automated scoring, and our analytics basically starts from the origination side.

Because we are we are primarily focusing on cleaning the data from the first step, we really believe in in doing big things right from the first time and not spending month for years, resolving this data issues later, it cleans, makes consistence and transforms in real time, all of the data that comes and associated with specific borrower, so then it also stored in a very, in a very consistent and tabular form. And that all of the processes built on top of this. So with this ability to transfer data on the very beginning, you have ability to build all of their business proves it's really simple, because you can easily digest the data that you do need. And then you can build through to assess, which is tightly connected to all of this pieces of data, like your underwriting your let's say product documentation, so on very easily and very fast.

By the way, we also utilise some open source solutions for our origination automation. Because we truly believe that there was some solutions on the market, of course, open source, which handles some things way better. Let's say it's a great open source solutions for transforming HTML documents into the PDFs.

Brendan Le Grange 18:53

And then that carries on, I guess, if people want to have loan servicing, what sort of products are you able to manage there? And what what are you doing differently,

Dmitriy Wolkenstein 19:01

the outstanding functionality or the servicing part is that also utilising these transformed and consistent data, it allows us to provide a lot of reactiveness, and also a lot of automation. I'll give you an example. Let's say we have a bank, and we have a lot of customers, we sell, let's say, credit cards. And we also have some micro economic extra factors, let's say like an extremely increasing conflation.

So we'd like to understand on let's say be weekly or monthly basis, how much customers in real time by default, and what should we do with this? So with our solution, you can easily run these proactiveness reassess them utilising all of this data in a fully automatic way. And let's even make some actions. If we do see that the customer is defaulting on some other loans. So we may let's say cut his limit on our credit card.

So this proactiveness and the easiness of how you can build all of these practices automations, on top of the green state, and on top of these processes is really outstanding.

Brendan Le Grange 20:10

It's great to have it all there. But to have that data at people's fingertips with fast feedback, without that dreaded six month one year delay between I've come up with a great idea, and manually put it in a queue for it to build. To have it there is Yeah, from start to finish, taking in your data, sure, but also data from other sources. That is what underpins the right sort of culture to make these decisions, these feedback loops. And that incorporates finance, I guess now's a good time to revisit that last one with analytics data we can understand. And it's not just the simple chart, you've brought in AI and machine learning, you know, words we hear about and we're all interested in.

But really, what does a modern analytics engine look like? What are you able to do?

Dmitriy Wolkenstein 20:58

With our analytics, we utilise our and our principal in the first place. And also things like your famous same garbage in garbage out. So that's why we're focusing on the cleaning data from the very beginning. So we wanted to give banking cleaners ability or on our sounds, in which direction they need to move in terms of the strategy fast and efficient enough.

So that's where we're focusing. Because as I just mentioned, we have all of the data about each customer about each application performance. From the origination site, this servicing site, we have a lot of data points, which is our way consistent and tolerated transport routes, departments, product departments and marketing departments can easily take this data in our analytics framework really easily create a specific data sets for them with the different integers with the different customer segments with the different track records of this specific customer segment, or application and so on.

It's only takes just in a matter of minutes to create a specific dataset you wanted to build model on top off, and then push this push this dataset into the machine learning tools. In our case, this is Amazon sage maker. But we also can try to work with some other tools. Basically, we utilise Jupiter on them with on cloud, and then there you receive your modelling results.

First of all, you can compare them with the previous models, let's say you do have one customer segments within one product. But let's say you want to compare how does the specific vintages perform previous quarter and this quarter? So we can easily compare all this models inside Amazon and see what what is the difference.

The next step, you push all of these projections and all of these newly created model into the one cash flow financial engineering view where a lot of financial people and stakeholders see the expected situation. So they see all of the products that have been created within the taken periods of time. And what was the expected amount of let's say products to be distributed, the average balances expected write offs and so on.

Then the second thing they see is the actual one, so they see actual results. And then once this new model created, they see the third view, how does your bank will perform, let's say, if you did to use this newly proposed model, let's say in the previous in the previous quarter, or you can try to just project how just your bank will perform in the next quarter. If you let's say utilise this new model from now. And all these gifts, stakeholders and financial and business people understanding which direction to move.

And basically all of these just allow them to be more agile, and to run business in the more well controlled and data driven way.

Brendan Le Grange 23:47

And yeah, since we spoke about feedback loops a few times now talk to me about champion challenger A B testing is that something that the engine can also incorporate and speed up? .

Dmitriy Wolkenstein 23:58

So once you do have the spare post models, this is this is Explainable AI in our case, because you can see the entire path from the importance of the features and to foundation of all of these features from which raw data sources and how the speaker's has been created. That's where this value loop for to to be real loop.

Once you do understand all of this, what you need to test.

And you should get back from an Analytics site to origination site and create some Justin campaigns with basically the copy of the existing products over the existing customer segments like customer profiles, but fine tune for them something click is that it's got to be the testing one, and the system will be randomly or of course you can also customise this internally and track. So to understand if this model was really right in terms of her projections, or if it's going to be not enough data, the system also will tell you that so unfortunately, it's not an update.

So we need to let's say have this test another 2/ 3/ 4/ 6 months. So yeah It's all about the testing. It's all about this agility.

And then having these three way view of expected, current, and proposed.

Brendan Le Grange 25:08

And Dmitriy, somebody might be listening and saying, this all sounds good, but too good to be true for me because my lending organisation has some quirks or some things that make it special. But this is not just a case of kind of listen, and imagine you offer the ability to do demos with the clients real data.

So talk to me about how somebody who is interested in Timvero as a solution for them can go in and actually see how it works.

Dmitriy Wolkenstein 25:34

First of all, we does offer real demos. And we also offer so called pilots. So you can easily come to us, we will dive deeper into your specific pains, specific practices. And we also be able in a matter of like, maybe two weeks, maybe one month to create a pilot for you, is also where we prove the easiness with which you can customise origination servicing flows in our system.

Of course, it's got to be done, like the minimal cost or sometimes even with no cost.

Brendan Le Grange 26:07

And Timvero.com is where they can go to learn a lot more and see some of this.

Dmitriy Wolkenstein 26:15

Yeah, so you can easily visit visit our website and bring the request demo, most of the demos and most of their emails. I'm still I'm still taking personally and replying personally, I will be happy to have first of all this quick introduction, to learn more about your objectives.

And then yeah, connect you with their related product owners so they can show you a demo and dig deeper into this.

Brendan Le Grange 26:43

Well, Dmitriy, I think it's been really exciting to hear how decisioning has changed and what's possible today. So thank you for your time. But before I let you go completely we are recording this at the end of the year, the perfect time to think about the future and what's next.

So as you look ahead to 2024 what's on the horizon or the on the agenda for Timvero?

Dmitriy Wolkenstein 27:07

The main idea is just to grow, test new innovations because we have multiple of them, which we need to test with existing customers. Next year more aggressively go after the US and Asia Pacific, write more stuff for our blog. Yeha, things like this.

Brendan Le Grange 27:24

Perfect. Well, I wish you the best of luck for that. And yeah, thank you again for your time today.

Dmitriy Wolkenstein 27:29

Thank you.

Brendan Le Grange 27:30

And thank you all for listening.

Please do look for and follow the show on your favourite podcast platform and share the updates widely on LinkedIn where lending nerds are found in our largest concentration. Plus, send me a connection request while you're there.

This show is written and recorded by myself Brendan le Grange in Brighton, England and edited by Fina Charleson of FC Productions.

Show music is by Iam_wake, and you can find show notes and written transcripts at www.HowtoLendMoneytoStrangers.show and I'll see you again next Thursday.