How we reduced the loan to 2 clicks
Weeks, months - this period comes to mind when they talk about loan processing. Our product "Credit Online" had to radically change the situation, reducing the entire procedure to a few clicks. Impossible? The developers of Promsvyazbank managed to turn this “torture walk” into a convenient “magic wallet”.
The idea of ​​creating a product was born in 2016, it was then that someone had a bright mind that in the age of ubiquitous digitalization it was somehow not good when torturing loans to torment customers, it’s time to move the whole process online. They began to think over ideas on how to abandon documents, looked at risk models - they hooked up colleagues from several departments to the brainstorming session. As a result, we managed to find a combination of conditions under which we were ready to lend to our clients, whose turnovers are clear to us, and the scope of our activity is transparent.
When the project was just beginning, there were fears that this service would scare people away. The audience of small and medium-sized businesses has grown at a time when in the field of lending to everyone right and left cheated, everyone was afraid of hidden fees, enslaving conditions in small print in the contract on page 115. Therefore, we did not set any transcendental indicators for online lending. We thought that the product will be used by the most savvy enthusiastic customers, and we will hone modern digital approaches - this will be good.
')
The description of business requirements took place in November-December 2016. By March, the prototype was ready, which could be tested on the combat system. First, we conducted testing in a bank with colleagues: try, poke, convenient / inconvenient. Then gave the selected customers. Of course, at the first stage, everything was done manually; at that time there was no talk about Bigdat, I had to grope for a niche with which to work.
In April 2017, the product started, the first real customers went. We collected feedback from the first hundred, everyone was delighted, no one believed that it was possible so quickly: I went in, made two or three clicks, got a loan.
Getting a loan according to the classical scheme takes a month or two or three. This is very long and often unacceptable for customers. They suffer business, accountants prepare documents and spend time to take them to the bank, and more than once.
For standard lending, SMB client provides a large amount of financial and other information characterizing its activities. The application goes to the analyst, who can sit for a month and consider the accounting documents, accounting, collateral, security. As a result, the transaction is highly structured and, as a rule, inconvenient for customers.
We began to look for opportunities to simplify and speed up this procedure.
In the first versions of the product, we also requested documents from clients, but only in electronic form. Clients downloaded these files to us in a form, and even this variant was perceived with enthusiasm.
Having evaluated the risks, having seen a responsible attitude towards monthly payments from clients, we realized that we could go further. As a result, in the summer of 2017 it was decided to completely abandon the collection of documents.
As further sales funnel showed, we guessed with this decision, customers began to use it en masse. The response pleased, especially when we introduced a new approach, replacing the applicative model with machine learning tools (Machine Learning).
Working with big data allowed us to create a portrait of an ideal client and process behavioral information. We have learned to automatically find those customers who need lending, and whom we can also lend to. As a result, the time from the moment of clicking on the “Get a credit” button to the live money in the account was reduced to a few minutes.
We started with a small number of testers, now their number goes to thousands. The product shows a high percentage of those who have used the loan, repaid and retake. In general, everyone who tried such crediting noted its convenience: even if a loan is not yet needed, it's nice to know that you have a backup solution, a “magic wallet”, which you can use at any time.
At the moment, we have issued more than 4,300 loans, the total portfolio has reached several billion rubles. It is clear that the scale of the bank - a small amount. One transaction in a large corporate segment can immediately block everything. It is important that two years ago no one would have thought that the project would grow at such a rapid pace in terms of portfolio and number of transactions. At first, the plan was almost 500 deals in a year and a half. We immediately took this bar in the first six months after the start.
The most important component of a successful service was the use of machine learning tools. Especially for online lending, we have made a model based on the client’s demonstrated behavior from external and internal data sources. Priority for financing are companies with stable business relations, having no problems with government agencies, competently conducting their financial activities, actively using current accounts.
Machine learning tools select suitable clients, make a control sample to make sure that the results are correct, and the automatic sending of the proposal starts.
In addition to using machine learning in scoring, we use this tool to create targeted personalized offers for customers who need this product with high probability.
In parallel with the loan, a similar offer was launched - “Overdraft online”. This is a slightly different kind of credit product. Its main difference is that money can be taken in any amount within the limit, at any time and for any fixed period.
A contract is immediately concluded with the client for a period of 5 years, during this period we give him the opportunity to go into overdraft within the agreed amount. However, he manages the money as he wants. It took money - borrowed money. Funds have been credited to the account - the overdraft is automatically extinguished. In the case of a loan, there is a clear agreed schedule when and how much to pay.
We started with a million rubles, then we increased the limit to two. Gradually, they began to make clients more and more advantageous offers: we are lowering rates, increasing terms, amounts. Now we have reached 5 million rubles in both overdraft and loan.
There is no limit to perfection, we continue to improve this product. While we are working with the visual part, especially with the mobile and new Internet banking in terms of usability, it is even easier for the client, more convenient to use them. We will try to increase the maximum loan amount. Perhaps we invent new credit logic, new products. We will begin to ask questions to clients, for what purposes they need a loan, and based on the answers we will form a suitable offer. Pans a lot.
There are ideas for creating online products for the external market. In about a month, the PSB product “Everything is simple!” Will enter the wide market. The client may not even be serviced at our bank. He simply attaches information on his turnovers and consent to the processing of data on a special landing page, and the application goes to our work. In general, we think, what else is pleasant to surprise our customers.
From idea to product
The idea of ​​creating a product was born in 2016, it was then that someone had a bright mind that in the age of ubiquitous digitalization it was somehow not good when torturing loans to torment customers, it’s time to move the whole process online. They began to think over ideas on how to abandon documents, looked at risk models - they hooked up colleagues from several departments to the brainstorming session. As a result, we managed to find a combination of conditions under which we were ready to lend to our clients, whose turnovers are clear to us, and the scope of our activity is transparent.
When the project was just beginning, there were fears that this service would scare people away. The audience of small and medium-sized businesses has grown at a time when in the field of lending to everyone right and left cheated, everyone was afraid of hidden fees, enslaving conditions in small print in the contract on page 115. Therefore, we did not set any transcendental indicators for online lending. We thought that the product will be used by the most savvy enthusiastic customers, and we will hone modern digital approaches - this will be good.
')
The description of business requirements took place in November-December 2016. By March, the prototype was ready, which could be tested on the combat system. First, we conducted testing in a bank with colleagues: try, poke, convenient / inconvenient. Then gave the selected customers. Of course, at the first stage, everything was done manually; at that time there was no talk about Bigdat, I had to grope for a niche with which to work.
In April 2017, the product started, the first real customers went. We collected feedback from the first hundred, everyone was delighted, no one believed that it was possible so quickly: I went in, made two or three clicks, got a loan.
The Origins of the Going On
Getting a loan according to the classical scheme takes a month or two or three. This is very long and often unacceptable for customers. They suffer business, accountants prepare documents and spend time to take them to the bank, and more than once.
For standard lending, SMB client provides a large amount of financial and other information characterizing its activities. The application goes to the analyst, who can sit for a month and consider the accounting documents, accounting, collateral, security. As a result, the transaction is highly structured and, as a rule, inconvenient for customers.
We began to look for opportunities to simplify and speed up this procedure.
Documents are superfluous
In the first versions of the product, we also requested documents from clients, but only in electronic form. Clients downloaded these files to us in a form, and even this variant was perceived with enthusiasm.
Having evaluated the risks, having seen a responsible attitude towards monthly payments from clients, we realized that we could go further. As a result, in the summer of 2017 it was decided to completely abandon the collection of documents.
As further sales funnel showed, we guessed with this decision, customers began to use it en masse. The response pleased, especially when we introduced a new approach, replacing the applicative model with machine learning tools (Machine Learning).
Working with big data allowed us to create a portrait of an ideal client and process behavioral information. We have learned to automatically find those customers who need lending, and whom we can also lend to. As a result, the time from the moment of clicking on the “Get a credit” button to the live money in the account was reduced to a few minutes.
Path to 2 billion
We started with a small number of testers, now their number goes to thousands. The product shows a high percentage of those who have used the loan, repaid and retake. In general, everyone who tried such crediting noted its convenience: even if a loan is not yet needed, it's nice to know that you have a backup solution, a “magic wallet”, which you can use at any time.
At the moment, we have issued more than 4,300 loans, the total portfolio has reached several billion rubles. It is clear that the scale of the bank - a small amount. One transaction in a large corporate segment can immediately block everything. It is important that two years ago no one would have thought that the project would grow at such a rapid pace in terms of portfolio and number of transactions. At first, the plan was almost 500 deals in a year and a half. We immediately took this bar in the first six months after the start.
Smart scoring
The most important component of a successful service was the use of machine learning tools. Especially for online lending, we have made a model based on the client’s demonstrated behavior from external and internal data sources. Priority for financing are companies with stable business relations, having no problems with government agencies, competently conducting their financial activities, actively using current accounts.
Machine learning tools select suitable clients, make a control sample to make sure that the results are correct, and the automatic sending of the proposal starts.
In addition to using machine learning in scoring, we use this tool to create targeted personalized offers for customers who need this product with high probability.
Overdraft online
In parallel with the loan, a similar offer was launched - “Overdraft online”. This is a slightly different kind of credit product. Its main difference is that money can be taken in any amount within the limit, at any time and for any fixed period.
A contract is immediately concluded with the client for a period of 5 years, during this period we give him the opportunity to go into overdraft within the agreed amount. However, he manages the money as he wants. It took money - borrowed money. Funds have been credited to the account - the overdraft is automatically extinguished. In the case of a loan, there is a clear agreed schedule when and how much to pay.
We started with a million rubles, then we increased the limit to two. Gradually, they began to make clients more and more advantageous offers: we are lowering rates, increasing terms, amounts. Now we have reached 5 million rubles in both overdraft and loan.
Further development
There is no limit to perfection, we continue to improve this product. While we are working with the visual part, especially with the mobile and new Internet banking in terms of usability, it is even easier for the client, more convenient to use them. We will try to increase the maximum loan amount. Perhaps we invent new credit logic, new products. We will begin to ask questions to clients, for what purposes they need a loan, and based on the answers we will form a suitable offer. Pans a lot.
There are ideas for creating online products for the external market. In about a month, the PSB product “Everything is simple!” Will enter the wide market. The client may not even be serviced at our bank. He simply attaches information on his turnovers and consent to the processing of data on a special landing page, and the application goes to our work. In general, we think, what else is pleasant to surprise our customers.
Source: https://habr.com/ru/post/441208/
All Articles