For data analysis on credit card consumption, if we can get everyone’s credit card consumption data (one person may have multiple credit cards), then how should we carry out analysis after getting these credit card consumption data.

The idea that has been talked about a lot about user consumption behavior analysis is still that we need to first figure out the goal of the analysis, and then collect and process the required data information based on the analysis of the goal. That is to say, data analysis itself is KPI-driven. So if we start from the most original data details, how should we expand and expand the data dimensions?

For those who have credit cards, the credit card bills we receive often have the simplest consumption details, as follows:

Consumption list (card holder’s card number, name, consumption Merchant, consumption time, consumption amount)

It can be seen that the detailed consumption data itself is relatively simple. If it is not combined with other data dimensions, simply doing statistical analysis will not produce too much. significance. Any data analysis needs to be combined with the dimension expansion of the original data. After the dimension expansion, the entire data model will be richer, and multi-dimensional analysis and data aggregation can be generated.

From the above consumption detailed list data, the following simple expansion can be made

Personnel information (person name, ID number, age, name, occupation type, residential address, family Information)

Business information (business name, business address, business type)

With personnel information, there is a first layer of expansion, that is, our data aggregation can be based on personnel The attribute dimension, that is, the detailed consumption data we obtain, can be aggregated according to consumer gender, age group, occupation type, etc. The unique code for identifying a person is not the name, but the person's ID number. That is, through the ID number, we can aggregate the consumption data of multiple credit cards for one person.

With merchant information, we can aggregate different types of consumption data according to the business type of the merchant. At the same time, you can see that the detailed address information of the merchant itself cannot be aggregated. Then we must consider the hierarchical expansion of the single attribute itself in the attributes of the subject object, that is, we can expand the address information, that is, city-》district-》region-》consumption area-》business district-》big shopping mall-》specific address .

If the address has this extension, you can see that the final consumption data can be aggregated by consumption area. We can analyze the consumption summary data of a certain business district or shopping mall, and the data itself is The model is extended from the original consumption detail data.

To achieve this, you can see that any dynamic consumption details must be matched with a large amount of basic master data. These basic master data may have a table structure or a dimensional structure. These data must be Organize and map detailed consumption details. In this way, the final consumption data can be easily analyzed in multiple dimensions and based on dimension aggregation.

Consumption time itself is also an important dimension. Through time, we can summarize data according to time periods. At the same time, time itself can be expanded layer by layer by year, quarter, and month. It is also a method that can be expanded hierarchically. structure. At the same time, you can also note that time itself can also be used to analyze consumption frequency, that is, the number of card swipes in a certain time period can be inferred to the popularity information of a certain area itself in certain time periods based on the consumption frequency.

If it is just the credit card consumption list data, it is difficult for us to locate the specific product SKU information. If it is a large supermarket, the detailed user consumption purchase data can also be broken down to specific items. For products, the expansion of the dimensional attributes of the product itself is content that can be expanded for analysis and aggregation.

The data itself may be relevant. Card consumption data can often be directly correlated with other data, such as major events in a region itself, marketing activities held in a region, which we obtained from the transportation department. Traffic flow data for a certain area. These may all be related to the final consumption data in some sense.

If you only look at the card swiping data itself, as mentioned earlier, you can locate the business scope of the merchant based on the merchant, whether it is catering or selling clothes. Then we can count card consumption data according to different business types, and then we can analyze whether the consumption of clothing will increase when the amount of food and beverage consumption increases, that is, whether the food and beverage merchants have any effect on the sales of other supplies in a shopping mall. Leading role, etc.?

For the same reason as for people, can we analyze whether there is a certain correlation between the consumption data of people of different age groups? What types of merchandise sales do these correlations exist for? These analyzes will facilitate us to develop more effective targeted marketing strategies.

Reference materials:

/archives/34264

/archives/38227