Shulou(Shulou.com)06/03 Report--

Before the editor introduced the cautious use of "data-driven" risk control model, the need for "high-quality data + prudent and rigorous decision-making model + real-time comprehensive monitoring analysis" dynamic closed loop. Dynamic closed loop can be "dynamic", which needs the last step of "monitoring" to check, comprehensively and carefully analyze and evaluate the effect of decision-making, so as to make the decision-making system evolve continuously.

To do a good job of monitoring is a vital measure of the "cautious" principle, which requires both macro and micro, and data visualization. Macroscopically, a good definition of business indicators can provide insight into business trends, predict overall profitability, and distinguish the rate of return on risky assets of different risk control decisions (income excluding bad debts / the total amount of maturing assets under this decision); microscopically, you can see the extension and distribution of business indicators in time or space, achieve perspective and drill-down analysis in different dimensions, find business anomalies, potential risks or loopholes, and summarize new rules. The ultimate goal of monitoring is to strengthen the company's risk control ability and help management to make wise and timely strategic decisions.

The primary goal of monitoring is to select appropriate business indicators, purely from the perspective of risk control, overdue rate, return rate, pass rate, rate of return on risky assets and other indicators are more concerned. However, as a CRO, other indicators also need to be paid attention to, such as business volume (including registration, application, loan, re-loan and other levels, abnormal business volume is often more worthy of attention from the risk control level), system operation indicators (not only the technical team needs to repair the abnormal operation of the system, CRO needs to be in charge of the potential risks caused by the abnormal operation of the system) Data quality index (data quality determines the accuracy and credibility of the risk control model), and so on.

Business personnel at different levels and departments have different concerns, so business metrics need to support different "granularity". Take the "pass rate" as an example, CRO will spend some energy on the conversion rate of each product from application to loan, while the risk control strategy manager needs to know the normal pass rate range of each risk control sub-decision and rule like the back of his hand, and any abnormal fluctuations may be a hidden risk. In order to achieve the "drill down" of business indicators at different levels, we must first do one thing: unify the standardized definition, caliber and calculation method of indicators, establish a change mechanism, use data quality management tools to manage and monitor well, and land in strict accordance with the standard definition (data warehouse, BI application, ETL, etc.), so as to avoid managers' mistakes or inconsistent cognition leading to wrong decisions.

Data visualization is another "data-driven" management tool that management needs. Visualization is done well, which can help management extract the most useful information from complex business data and find business problems more quickly and accurately. Combined with the risk control scenario, here are some examples of visualization:

Trend analysis: focus on the short-term and long-term trends of business volume, overdue rate (1st, 7th, 30th, 90th, etc.), pass rate, gross margin and other indicators, evaluate the effectiveness of risk control decisions, gain insight into potential risks, and control the profitability of the company.

Funnel analysis: each combination of risk control approval, every step and even every rule should be put into the funnel for analysis and monitoring, and observe the volatility of the pass rate of each decision, step and rule, as well as the comparison between decisions. identify possible problems.

Comparative analysis: select a dimension (such as time, age, region, user score, etc.) and compare the difference of an indicator (such as overdue rate) in that dimension.

Finally, we review and summarize the contents of the three articles: it is not a day's work to be "cautious" in the use of "data-driven" risk control models, and it is necessary to manage and use "data" well, make prudent and flexible "decisions", and comprehensively and finely "monitor". In this process, in addition to the need for advanced ideas, the use of advanced management tools to achieve automation can greatly improve operational efficiency and reduce many detours. Some mature software include unified data management platform, data quality management tools, data warehouse, intelligent decision engine, BI application (management cockpit) and so on.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.