Keywords

1 Instructions

1.1 Model Assumptions

In the era of big data, the general analysis of the basic information of members, to make a correct assessment of their consumer behavior can help managers make the right marketing decisions. In the retail industry, the purchasing power of members reflects the consumption level and consumption level. Understand the purchasing power of consumers, to do more accurate member marketing programs and improve sales. This paper studies the method and model of shopping mall members’ purchasing power evaluation in big data environment.

In the retail industry, the value of membership is reflected in the consistent generation of stable sales and profits for retail operators, as well as the provision of data to support the retail operators’ strategy development. The retail industry will adopt various methods to attract more people to become members and to increase the loyalty of members as much as possible. At present, the development of e-commerce has led to continuous loss of mall members, which brings serious losses to retail operators. At this point, operators need to implement targeted marketing strategies to strengthen good relationship with their members. For example, merchants take a series of promotional activities for members to maintain their loyalty. Some people think the cost of maintaining old members is too high. In fact, the cost of developing new members is much higher than the cost of taking certain measures to maintain existing members. Effective ways for the brick-and-mortar retail industry include improving the member portrait depiction, enhancing the refined management of existing members, pushing products and services to them regularly, and building stable relationships with members. The mathematical model was established to calculate the activation rate of non-active members in the life cycle of members, that is, the possibility of transforming from inactive members to active members. Based on the actual sales data, the relationship model between the activation rate and shopping mall promotion was determined. Generally speaking, the higher the commodity price is, the higher the profit will be. Joint consumption is the core of shopping center operation, if the business will plan a promotion, how to plan the promotion according to the preferences of members and the joint rate of goods. The Symbols Shows as Table 1.

1.2 Model Notations

Through the three fields of document number, cash register number and consumption time, we can uniquely identify an order (receipt), which may contain several different products of different brands.

Table 1. Symbols
Full size table

In other words, the model assumes that there are no two customers settling accounts at the same register at the same time, so there are no identical bill numbers in the system. Suppose there are only two forms of promotional activities in shopping malls. One is direct price reduction or discount, which is reflected in the difference between the amount paid by customers and the total amount of goods; the other is store points, which is reflected in the increase of membership points.

2 Problem Analysis

2.1 Problems to be Solved

Firstly, we determine the indicators for evaluating the promotional activities of shopping malls. According to the assumptions of the model, we will establish the evaluation indicators from the aspects of discount and points. Discount indicators provide a comprehensive measure of discount strength, such as monthly discount rates, total discounts, total number of discounted products purchased by members, percentage of discounted products out of total products sold, and average discount range for each brand. In terms of points, the total number of points issued in a month and the ratio of points to the total amount (i.e., the ratio of points) can measure the generosity of points issued by the shopping mall, which is also a method to motivate members. Then we research the correlation between the activation rate and the above indicators to determine whether promotional activities have an incentive effect on the activation rate. At the same time, we study the overall impact of indicators on the activation rate through the regression model due to the large number of indicators. In this process, considering the possible strong correlation between indicators, we screen variables through Lasso regression. In order to study the associated consumption of commodities, we can analyze the association rules by integrating the commodity records of each purchase, and find out the commodity combination that customers often buy at the same time to understand the associated consumption. Finally, we give marketing recommendations based on joint consumer preferences.

2.2 The Activation Rate of Non-active Members

The following indicators can be used to evaluate the strength of promotional activities, and the activation rate of inactive members and invalid members may be related to these indicators.

Discount rate: total sales for the current month/original selling price and price of all goods sold for the current month;

Discount number: the number of items purchased by members of the store in the current month for less than the original price;

Number of discounted items: number of discounted items/total number of items purchased by members in the current month;

Discount rate of discount brands: the collection of all discount brands purchased by members in the current month calculate the number of items sold to the store members and discount items for each brand participating in the discount, the average of the discount product ratio of each store is the discount variety ratio of the discount brand in the current month. This measure measures the degree to which stores of various brands participate in discount.

Discount merchant ratio: the ratio of the number of brands that sold discounted goods in the month to the total number of brands.

Total points distributed this month: the sum of points distributed to members of the store this month.

Ratio of bonus points issued this month: total bonus points issued this month/total sales amount of members of this month. The correlation is not significant, P0,2(5) Is negative correlation.

Discount Variables are shown in Table 2.

Table 2. Discount variables
Full size table

The analysis shows that there is a negative correlation between the rate of bonus point payment and other discount indicators. In other words, the discount is relatively low when the rate of bonus point payment is high. The incentive of points to the lost customers and inactive customers is far less than that of discount, so the high rate of point payment does not contribute to the improvement of the activation rate. On the contrary, in the months with high rate of point payment, the activation rate is low due to the low discount, which resulting in a negative correlation between the rate of point payment and the discount rate. At the same time, the positive correlation between If and activation rate is probably due to the higher discount rate at that time, which leads to higher sales volume and thus increases the total number of points issued, resulting in the above positive correlation. No matter which index is evaluated, the increase of discount will increase the activation rate. Among them, the discount rate is the most correlated with the activation rate of inactive members, while the number of discount pieces is the most correlated with the activation rate of invalid members, that is, the scale of discount products. Relevance matrix are shown in Table 3. We use Lasso regression (alpha = 0.1) to screen variables because of the strong correlation between variables. P0,2(5) Loasso Model Parameters are shown in Table 4. P1,2 Loasso Model Parameters are shown in Table 5.

Relevance matrix:

Table 3. Relevance matrix of activation rate and discount rate
Full size table

P0,2(5) Loasso Model Parameters:

Table4. Lasso model parameter table of churn customer activation rate
Full size table

P1,2 Loasso Model Parameters:

Table 5. Lasso model parameter table for inactive customer activation rate
Full size table

2.3 Conclusions

The total quantity of discount goods and the quantity of points released are significant variables selected by Lasso regression. In general, increased discount rates, increased brand coverage and size of discount events increase activation rates for inactive and inactive members. The increase in the rate of points may have a stronger incentive effect on active members, but as the rate of points increases, the discount intensity in the shopping mall is generally weak, so the rate of points has no incentive effect on inactive members and invalid members.

2.4 The Associated Consumption of Commodities

The associated consumption of commodities is an important phenomenon in the process of business operation. Paying attention to customers’ preference in the process of consumption is beneficial to the planning of promotional activities. Establishment of association rule model: In order to analyze the associated consumption of commodities, the following definitions are given: Commodity purchase data set T = {T1, T2, …, Ti, Tn}, A transaction that represents the purchase of an item by a customer, Ti = {I1, I2, …, Ii, In}, represents an item in the Ti consumption transaction. Commodity group: let I be the set of all items in the commodity purchase data set T, and any subset of I is called the commodity group in T.

Support count: the support count of item group X is the number of times item group X appears in item purchase data set T.

Support degree: the support degree of commodity group X is the percentage of commodity group X in the commodity purchase data set T, which describes the probability of a commodity combination appearing in all commodity consumption records. The support degree of commodity group X is expressed as

$$\mathrm{support}(X)=|\{occurency(X)|X\subseteq T\} |/occurency(T)$$

Frequent commodity group: the commodity group whose support degree is not less than the given minimum support degree is regarded as frequent commodity group. Confidence is the percentage of the goods purchase data set T that contains both goods group X and goods group Y. Write rules of \(\mathrm{X}\to \mathrm{Y}\) confidence for the \({\text{conf}}\left( {{\text{x}} \Rightarrow {\text{y}}} \right)\)

$$\mathrm{conf}(\mathrm{x}\Rightarrow \mathrm{y})=(support(X\cup Y))/(support\left(X\right))$$

Confidence means that for the association rule X Y, the higher the confidence is, the greater the probability that both X and Y of commodity group appear in the consumption. In order to mine related commodity groups that meet the minimum support degree and the minimum confidence degree, it can be divided into the following two steps:

  • Step 1: find out the frequent commodity group set that meets all conditions in all data of commodity purchase data set T.

  • Step 2: generate association rules with frequent commodity groups, that is, find the rules satisfying the minimum confidence degree from frequent commodity groups obtained in the previous step. No less than the minimum confidence, the association rules.

3 Examples and Illustration

The tables and data above shows that solution of association rules: the consumption records of members are processed, the records belonging to the same consumption are identified through the document number and time, and the commodities purchased at the same time in each consumption process are summarized and recorded in the form of code, forming the data set of commodity purchase. We found that these sets of all goods are cosmetics by observing the commodity group, which represent this category is more suitable for joint consumption. At the same time, the cosmetics is also the main item sold by the store, because the volume and sales of cosmetics are more than the volume and sales of any other category. Thus we speculate that cosmetics sales should be the main business of the store. Secondly, we found that all associated commodity combinations belong to the same brand, and customers tend to purchase multiple commodity combinations of the same brand at the same time when purchasing commodities. A common pattern is to buy sets of skincare products (for examples, a day cream with a night cream, a moisturiser with a cream, a softener with a lotion) or sets of bottom makeup products at the same time.

When only the minimum confidence in the model is changed, the generated frequent commodity portfolio and its support degree will not change. Association rules and confidence are not changed, but quantitative filtering is performed. When only the minimum support degree in the model is changed, the generated frequent commodities and their support degree will not change, but will be screened quantitatively. Association rules and confidence will change greatly. Therefore, it can be explained that the model has good robustness, and the changing of minimum support and minimum confidence will lead to frequent commodity combination and quantitative screening of association rules, while the change of the content of relatively important association rules is less. The programs explanations are shown as following:

figure a

4 Conclusions

Conclusions for shopping mall promotions: Based on the above conclusions, we give Suggestions for shopping mall promotional activities.

  • Conclusion 1: the main target of promotional activities should be cosmetics, and it is better to launch promotional activities for products of the same brand.

  • Conclusion 2: with reference to the groups of commodity combinations with associated consumption relationships, preferential package can be launched to stimulate purchase, or the sales volume of associated consumption commodities can be increased by offering discounts to actively purchased commodities.