How many rules are best to implement in the early stages of building a pre-lending strategy? Is it better to have more rules online at one time? This may be a headache for strategists in the early stage. They are not only afraid that too many invalid strategies will delay the launch time, but also miss the best opportunity for the business to enter the market; they are also afraid that there are too few online rules and some bad guys will be missed, resulting in early risk indicators. Hard to explain.

In fact, generally in the early HE Tuber stages of business launch, the fewer rules the better. It is best to have an MVP version, and it must be able to support rapid iteration.

The specific reasons are as follows:

In the early stage of business, it is necessary to quickly test the water and find out the natural risk situation of the target customer group, so that rapid evaluation and adjustment can be made at the product level; the absolute value of the risk generated by the early water test and the later water test are completely different in magnitude. The early stage is less, and the loss can be stopped quickly.

Iterative optimization of strategies after launch requires a sufficient sample size. More samples in the early stage can help quickly locate problems and fill loopholes. In this way, there will be fewer loopholes and greater confidence in business expansion; at the same time, iterations can be avoided The embarrassing situation of insufficient samples affects the iteration speed.

In the early stage, there is no Y value of its own customer group, and it is impossible to accurately quantify and define the effectiveness of early strategies, so some uncertain strategies can be observed offline.

In the early stages of building the strategy, some obvious risk strategy loopholes have been filled through rapid iteration, and the rules for the initial launch have been comprehensively reviewed and adjusted to basically ensure the efficiency of the strategy in online operations.

Under the premise of the same customer base, there is no obvious gap in the risk strategies of various financial institutions in the early stage. Unless some financial institutions have obvious advantages in their own data, the gap will mainly focus on the mid-term. .

The number of strategy rules launched in the mid-term will quickly reach a high number, like a volcanic eruption, and will continue for a period of time. The reason is that in the middle stage of building the pre-loan strategy, the business volume has reached a relatively considerable level, which can support quantitative analysis of customer segments and also has a sufficient sample size. Next, large-scale quantitative analysis and strategy iteration are required.

The middle stage of strategy construction is mainly divided into the following four stages:

Carry out large-scale and orderly online A/B testing of available rules. The main goal of this process is to find bad actors and convert the tested rules into correct ones and put them online;

Conduct internal cross-analysis on this batch of regularized rule sets to eliminate impurities. To put it simply, it is to see which rules can be replaced by other rules; then cross-analyze the set of corrected rules after removing impurities with the original set of rules to further remove impurities;

After the completion of the above 2-step system, the next step is to do the fishing. That is to find some good customers among the rejected customers. At this time, you also need to do a lot of A/B Test, and verify the correct rules and put them online;

The last step is to classify and summarize the refined strategies to form a complete strategy system.