Operation DEMO: The principle and application of RFM user value model

Among the many user value analysis models, the RFM model is widely used; the RFM model is an important tool and means to measure customer value and customer profitability. In the RFM model, R (Recency) indicates how long ago the customer purchased, F (Frequency) indicates the number of times the customer purchased within a period of time, and M (Monetary) indicates the amount of money the customer purchased within a period of time.

General analytical CRM focuses on the analysis of customer contribution, while RFM emphasizes differentiating customers based on their behavior. In actual applications, there will be many variations of reference data dimensions depending on the business.

The significance of RFM model

The RFM model shows the entire profile of a customer in a relatively dynamic manner, which provides a basis for personalized communication and services. At the same time, if the time spent dealing with the customer is long enough, it can also more accurately judge the long-term value of the customer, and by improving the status of the three indicators, provide support for more marketing decisions.

The application of the RFM model aims to establish a user behavior report, which will become an important indicator for retaining customers.

Therefore, the RFM model is widely used in marketing to stimulate continued consumption and retention of new users. It can also serve as an important indicator for monitoring the health of business users. If the report shows that the number of customers whose last purchase was very recent is increasing, it means that the company is a steadily growing company. On the contrary, if the number of customers whose last purchase was one month ago is decreasing, it is a sign that the company is heading towards an unhealthy path.

I would like to add one more thing here. Vertical businesses with single functional consumption often fall into the trap of pursuing short-term goals only for transaction volume, while neglecting to establish long-term relationships with users. Maintaining continuous interactions with users and winning user loyalty are more long-term goals for the business.

Using the absolute contribution amount of a customer to analyze whether a customer has churned can sometimes misinterpret customer behavior. Because the price of each product may be different and there are different discounts for promotions of different products, using relative grading (for example, R, F, and M are each divided into five levels) to compare the changes of consumers in the level range can better show relative behavior.

Companies can use changes in R and F to infer changes in customer consumption, list customers based on the possibility of churn, and then analyze from the perspective of M (spending amount). They can then focus on customers who have high contributions and high chances of churn, and intensively visit or contact them to win back more users in the most effective way.

method

The purpose of using the RFM model is generally to segment the most valuable users, focus investment with limited marketing resources, and maximize value output. For this purpose, we usually need to sort the segmented users according to the grades given by the RFM model, from best to worst, and then keep everything normal. If the business results are consistent with your ranking over a period of time, then it means that your segmentation can relatively accurately separate valuable users from the overall users.

Usually the operations team will continuously verify and revise dimensions to improve the efficiency of member operations and spend the budget where it should be spent.

User behavior is constantly changing. When a month has passed since the last purchase by a user, he or she will be recorded in the database as a customer with a two-month purchase history. Conversely, if a customer with a purchase history of three months ago makes his or her next purchase on the same day, he or she will become a customer with a purchase history of one day ago and may receive new marketing information in a very short period of time. We constantly adjust the model to segment the population and then compare business results to verify the accuracy of the model. The model application process involves an algorithm and a decision tree model.

The K-means algorithm is a hard clustering algorithm and a typical representative of prototype-based objective function clustering methods. It uses a certain distance from the data point to the prototype as the optimized objective function and uses the method of finding the extreme value of the function to obtain the adjustment rules of iterative operations.

The K-means algorithm uses Euclidean distance as the similarity measure. It seeks the optimal classification corresponding to a certain initial cluster center vector V so that the evaluation index J is minimized. The algorithm uses the error square sum criterion function as the clustering criterion function. Okay, I admit that this thing is too tongue-twisting. I suggest that those who are interested go and check out the encyclopedia of the original word.

Decision tree models are very basic mathematical laws that determine the process of logical reasoning and, based on the available information, determine the confidence of various possible alternative options.

Many business decisions involve uncertainty, and decision trees can help choose the best course of action when faced with uncertainty. Although the decision maker does not know what the unknown consequences will be, he generally has some understanding of the possible consequences and the probability of their occurrence. This information can be used to select the option that will yield the greatest benefit, that is, whether the user will continue to consume next time.

Case

The budget of the membership operation team is usually stretched compared to the huge user scale, so when our budget is limited and we can only provide service information to a small number of key users, the RFM model comes in handy.

The cost savings from such marketing can be considerable. Combining these three core indicators, we divide customers into multiple categories, conduct data analysis , and then formulate our marketing strategy.

Taking a vertical functional Internet business as an example, a sample of 1 million users was randomly selected for data analysis;

• Time interval: Divide the time point when the sample is extracted by the time of the user's last consumption to see how many days the average time interval is.
• Consumption frequency: the average number of times a user consumes, the maximum number of times, and the minimum number of times.
• Consumption amount: What is the average consumption amount of users, and what are the maximum and minimum amounts respectively.

Using the K-means method, the sample set is clustered and usually divided into 8 categories.

The clustering results are divided into a training set (30%) and a test set (70%), and a decision tree model is generated based on the training set.

Through RFM analysis, the customer base is divided into eight levels: general retention customers, general development customers, general value customers, general retention customers, important retention customers, important development customers, important value customers, and customers to be retained;

While continuously revising and improving the model by comparing business results, marketing methods are used to further expand the important value user group (considering short time, high frequency, and high consumption).

Conclusion

Consumption, consumption frequency, and consumption amount are the most important and easiest ways to measure consumer value, which fully demonstrates the guiding significance of these three indicators for marketing activities, but it does not mean that these three indicators are unbreakable. For example, Didi (express, private car, ride-sharing and multiple business types) and Alipay (multi-functional scenarios), in addition to consumption amount and frequency, the business side will also consider users' cross-scenario usage when formulating subsidy strategies. The more functional business scenarios are used, the higher the user loyalty. At this time, adding or replacing the core indicators of the model can be actually applied to the auxiliary subsidy strategy. This is why others can get 5 yuan coupons, but you can only get 1 yuan, and why others can get express coupons, but you can only get airport pick-up car coupons.

Mobile application product promotion service: APP promotion service Qinggua Media advertising

The author of this article @戴某DEMO is compiled and published by (APP Top Promotion). Please indicate the author information and source when reprinting!