This task was solved as part of the work on designing the interface of the loan application processing system for a large bank. However, this method can be useful in other applications. For example, where the user reviews his own credit history.

As a result of conducting interviews with users and studying the experience of using the existing interface, one of the bottlenecks was found.

It turned out that credit analysts spend a lot of time analyzing the applicant’s credit histories presented in the large tables and they make a lot of mistakes.

This increases the bank’s costs for reviewing loan applications. The team decided to improve the user experience of interacting with credit history information and reduce these costs. Also, this improvement could improve the automatic processing of applications, because with the reduction of hand processing errors, the learning indicators of automatic algorithms (machine learning) improves.

Task setting

In the process of collecting primary information, it was found out that it is important for the user to quickly evaluate the following parameters of the credit history:

1. The number and form of delays. Were they, and if so, when and of what type. For example, do they say that the borrower is unpunctual (not critical) or that he is insolvent (critical).
2. Estimate the size of the current debt and its ratio to the total amount of loans paid before. This gives an understanding of the probability of repayment of the loan.
3. Evaluate which part of the current loans the borrower has already successfully repaid and which is left.
4. Evaluate at what interest the borrower received loans in the past. Which indirectly indicates how other credit organizations assessed the risks when issuing loans. The higher the interest, the less reliable the borrower.
5. See if there were early repayments and to what extent. If there were too many, it increases the risks of profit losses for the bank.

Solution

During the design process, several approaches were considered, several prototypes of graphical display of information were tested, and an optimal solution was created that meets all the requirements.

The basic form is the display of data in tabular form (an improved version of the previous implementation), but if desired, you can switch to a graphical form. This approach allows you to safely test a new function, get feedback and improve the interface based on the results, and at the same time not interfere with the traditional analysis process based on viewing tables.

General view of the graph

Each loan on the chart is displayed as a set of monthly payments. Upcoming payments are shown in bright blue, repaid payments are shown in gray. If there were delays, the payment is highlighted in red, orange and yellow. Depending on the duration of the delay in days. The area of the graph (the sum of all columns) shows the total size of the loan — this analogy is taken from mathematical analysis, where the geometric meaning of the integral is the area under the graph.

Results

The introduction of this graphical display of credit history was recognized as successful. After a number of small improvements made following the results of the test operation, it was implemented for all users.

1. The time spent by credit analysts working on the application has been reduced by about 20%-30%.
2. The total number of errors has decreased. Managers have stopped making some types of common mistakes, and in general, the number of applications sent for revision has decreased by 10–30%.
3. As a separate bonus, some users noted that it became easier for them to distinguish the questionnaire of one borrower from the questionnaire of another. The fact is that analysts often opened many loan applications in different windows and made efforts to distinguish them from each other. Now it has become easy to do this by drawing a credit history that is unique for each client.

Further development

It was decided to implement the tool in a custom application. To enable the bank’s client to evaluate his credit history not only with the help of a table, but also in the form of the same graph. This should not only improve the user experience, but also add a unique tool for the market to the client application, which will be useful for highlighting the bank in the market.