Sberbank System of assessment of clients (scoring)

Main article: Credit scoring

2020: The patent application for the system of assessment of clients

On September 21, 2020 it became known that Sberbank decided to patent the system of assessment of clients. The feature of a system is that for issuance of credit the State Bank does not need information from bureau of credit histories and biographical particulars of the client any more. That the client got necessary scoring points, it is enough to it to use any map of bank.

Proceeding from this information the bank gives to each client a certain scoring point on which makes the decision, the credit will be issued or not. The portrait of the ideal borrower is approximately identical to all banks: women, family, people are more senior than 40 years, clients with the higher education and with good length of service pay better than other borrowers.

This system which Sberbank not only uses but also decided to patent, differs from traditional approach of banks. She stakes on studying of bank transactions, but not credit history of the client.

According to bankers, information on, where, when and what the client pays with the bank card more for speaks about his reliability, than the credit history and other factors used for traditional scoring.

Besides, such method does not demand from the client of input of any optional data, so the decision on the credit is made quicker (all process will be automated), and it is very difficult to forge data.

Therefore, there is no need to check correctness of information, unlike the questionnaire for receiving the credit and some other sources used for assessment. And the main thing, this method works even if the client has no credit history.

As affirms as the request to the patent, the bank constructed this system, having studied more than 200 million transactions of 740 thousand clients. As a target variable the event of default for a consumer loan within a year after its issue was used. As a result profiles of clients with the highest risk of nonpayment of the credit and profiles of the most reliable borrowers were received. For September 21, 2020, according to the request, as initial data the transaction level (a time stamp, the country, the amount, type of the seller) and card level (branch of issue, card type) are used.

As a result the portrait of the ideal borrower turns out a bit different, than at traditional scoring. As noted in documents, for the complete analysis it is enough to client to make about 350 transactions.

Real transactions on the debit or credit card really can tell about the client more, than dry digits from 2-NDFL or biographical particulars, – the leading expert of academy of financial literacy Alexander Nikolaev notes. – First, the amount of expenditure of the client according to cards gives the chance to bank to estimate its real income. Many to receive a cashback according to cards, deposit a "grey" part of salary to bank accounts. And the cost structure shows habits of the client and whether there is enough for him gained income for life.

Besides, on transactions of the client it is visible at once what is owned by him: whether there is a car whether there are an apartment or the house that also influences the final solution^[1].

Notes