User value Analysis

Source: Internet
Author: User
Tags radar

Who is using my site-user loyalty and value analysis

Described above are some of the user's behavior indicators and user segmentation, here is to introduce a comprehensive analysis and assessment based on the behavior of each user, mainly including the user's loyalty and user value. The "user-centric" theory requires the website to constantly optimize the user's experience, and thus improve user satisfaction, when the user's expectations are constantly being met, users will like this site, and thus develop into a website loyal users, while continuously for the site output value. Loyal users not only themselves for the site to create value, but also for the site to bring many hidden benefits, such as the promotion of brand and word-of-mouth, driving the entry and growth of other users. So the site's loyal users are the cornerstone of the survival and sustainable development of the site, we need to master the loyalty of each user, but also need to understand the value of each user embodiment.

This data analysis needs to come from the marketing department of the website, the marketing department's colleagues need to follow up some of the site's paid users and potential paid users, in order to better promote the site's products, to provide better services for customers, to guide the consumption of new users and the continued consumption of old customers. Because of the limited resources, the marketing department in the face of the growing customer base, they do not have the energy to follow up and service for each user, so they ask data analysts help, help them find targeted customers, in order to improve productivity. The sales department sent a request for data analysis to the mail.


It seems that this problem really bothers the marketing department's colleagues, if they do the marketing work of most users do not have any response, this is a very frustrating thing. Their goal is to narrow down the target group and position the valued clients with the potential to reduce daily ineffective work and improve efficiency. What they need is the analysis of user loyalty, the evaluation of user value and the continuous development of user value. We use the data analysis method to solve these problems by one by one.

Loyalty analysis based on user behavior

User Loyalty (Loyalty ) is the degree to which users frequently repeat purchases for business or brand preference. For Web sites, user loyalty is the behavior of users who visit the site frequently for the purpose of the site's functionality or service preferences. According to customer loyalty theory, the loyalty of the traditional sales industry can be measured by the following 4 indicators:

L Repeat Purchase intent (repurchase Intention): Willingness to purchase previously purchased types of products;

L Cross-purchase intent (cross-buying Intention): The willingness to purchase previously purchased product types or extended services;

L Client Referral intention (Customer Reference Intention): Recommend to other prospective customers, pass the brand word-of-mouth willingness;

tolerance Price Endurance: The highest price the customer is willing to pay.

The above 4 indicators for e-commerce sites, may also have applicability, but for most of the site is not appropriate, so in order to make the analysis of universal applicability, and in order to meet all the indicators can be quantified (above the customer referral intent is more difficult to quantify), in order to carry out quantitative analysis requirements, We select access-based user behavior metrics for all sites: user access frequency , last access interval , average length of stay , and average number of pages viewed , which are also Google 4 Indicators under User loyalty module in the original version of analytics.

These 4 indicators have been mentioned many times above, and the definitions are not repeated. Statistical data of the time interval is based on the characteristics of the site, if the site information update faster, more frequent user access, then the appropriate selection of a shorter time period, so that the sensitivity of the data changes will be higher; Conversely, select a slightly longer period of time, so that the user's data richer, The results of the analysis of the indicators will be more accurate and effective. After the statistical results of the 4 indicators, the value of the indicator is still unable to get user loyalty, need to standardize the indicators to get a corresponding score, by scoring can distinguish the user's loyalty in the overall level.

Here, using the Min-max normalization method, the 4 indicators are normalized and scaled to 10 points (0~10 points) of the scoring range. It should be noted here that the Min-max normalization will be affected by outliers, such as the number of users to browse the page has a 50 of the unusually large value, then normalized most of the values are concentrated in the smaller area of the score, so it is recommended to check before the normalization of the indicators of the existence of outliers, if there is, The exception value can be converted or filtered , while the most recent visit interval is also applicable to the "day" as the unit, attention to the normalization of the time required special treatment, because the greater the number of intervals, the corresponding score should be smaller, different from the other 3 indicators, the other 3 indicators using the formula (x -min)/(Max-min), the number of days of the most recent visit is handled using (max-x)/(Max-min). We use nearly one months of user access to data, select 3 users to enumerate the user behavior data processing, see table 6-2.

Table 6-1 User Loyalty Index score

 

 

Frequency of Access

Last access Interval

Average length of stay

Average number of pages viewed

User 1

Data

3 plays

15 days

150 seconds

Page 3

Standardization

0.10

0.50

0.30

0.22

Score

1.0

5.0

3.0

2.2

User 2

Data

12 plays

2 days

120 seconds

Page 4

Standardization

0.55

0.93

0.24

0.33

Score

5.5

9.3

2.4

3.3

User 3

Data

1 plays

21 days

300 seconds

Page 6

Standardization

0.00

0.30

0.60

0.55

Score

0.0

3.0

6.0

5.5

Table 6-2, the user loyalty of the 4 analysis indicators after standardized processing unified in a very formal output, so you can directly distinguish each user's performance of each indicator. Based on the scoring of each indicator, users can be screened, for example, the marketing department focus on follow-up visits to the site users, you can choose to access frequency score more than 3 points of users, or focus on follow-up users to participate in a higher participation of users, can filter average stay time and average number of pages visited more than 3 points of users, This will help the marketing department quickly locate loyal users.

Here we use 4 user behavior indicators to evaluate the user's loyalty, this kind of multi-indicator based on multi-angle evaluation of the most common way to show is the radar chart , or spider, in computer games is more common, such as some football games using radar charts to show the players of various aspects of the ability Index, such as defense, offensive, technology, strength, spirit, etc., so here can also borrow a radar chart with 4 indicators to show the user's loyalty performance, 6-18 shows.

Figure 6-1 User Loyalty Radar chart

Figure 6-18 uses the scoring data of three users in table 6-2 to be drawn, can be very visually displayed user loyalty in the performance of the indicators, user 1 of the overall loyalty is low, user 2 in the frequency of access and access interval has a good performance, and user 3 access has a relatively high degree of participation. Using radar charts to analyze user loyalty has the following advantages:

u can display all the evaluation indicators in full;

U show the user's bias in each index score, in which aspect performance is better;

You can simply observe the user's overall loyalty, that is, the size of the graph around the area (assuming that the weight of the 4 indicators is equal, if there is significant difference in importance, it can not be measured by the area);

U can be used to compare the loyalty between users.

So, based on the radar chart to show the user's loyalty, the marketing department can directly see which users have good loyalty, which users are worthy of their focus on follow-up.

A comprehensive score based on user behavior

The user loyalty analysis described above uses the user's 4 behavioral indicators to evaluate, but we can only see the performance of the indicators, unable to assess the overall level of user loyalty, so need to all the relevant indicators to do a summary processing, to obtain a comprehensive score, like a football game player's comprehensive ability value (Overall). The above-mentioned loyalty indicators have been standardized to standardize the measurement interval, the simplest method is to take all relevant indicators to score the average value to calculate the loyalty score, such treatment will all indicators with the same importance of treatment, but in reality, the impact of different indicators on the comprehensive score is not the same, Some indicators are more critical, some are relatively minor, so the introduction of AHP method to set the weights of different indicators.

AHP (analytic Hierarchy process) is an American operational research company T. L. Professor Saaty in the early 1970s, AHP is a simple, flexible and practical multi-criteria decision-making method for quantitative analysis of qualitative problems. It is characterized by dividing the various factors in the complex problem into an orderly level of interconnectedness, making it organized, and quantitatively describing the importance of comparing each level element 22 to the subjective judgment of a certain objective reality. Then, a mathematical method is used to calculate weights that reflect the relative importance order of each level element, and the relative weights of all elements are calculated and sorted by the total ordering between all levels. The analytic hierarchy process is applicable to multi-objective decision-making, and it is used to evaluate the merits and demerits of each case in case of multiple impact indicators. Analytic hierarchy process can be used when a decision is influenced by multiple factors and there are hierarchical relationships between them, or there is a clear classification of categories, and the degree of influence of each index on the final evaluation cannot be quantified directly by sufficient data.

After understanding the AHP, we take the above loyalty score as an example, first introduce the application of AHP. First, based on the loyalty of the impact indicators to build a hierarchical model, where only two layers, the upper level is loyalty, the lower level is the impact of the 4 indicators of loyalty, 6-19 is shown.

Figure 6-2 Loyalty scoring hierarchy model

We need to calculate the impact weight of the underlying 4 indicators on loyalty, we need to build a contrast matrix, that is, the use of 9 scale on the need to empower the same layer of the impact of the 22 comparison, such as the model of the element I relative to the importance of the upper layer J, 1 means I and J are equally important, 3 is more important than J 5 means I is more important than J, 7 means I is much more important than J, 9 means I is more important than J, can be expressed with WI/WJ, 22 can be obtained following matrix after comparison:

The results of 22 comparisons can be obtained from the diagonal of the matrix, and the symmetric elements on both sides of the matrix are reciprocal, and the values of all the elements of the diagonal are 1, so we can get the whole matrix by the numerical value on the diagonal side. Since the value of the matrix is the result of 22 comparisons, it is possible that a element is more important than the B element, the B element is more important than the C element, but the C element is more important than the a element, that is, the inconsistency of the matrix, so first we need to verify the consistency of the contrast matrix. The consistency of matrices can be measured by the method of calculating the maximum eigenvalue of the matrix, the correlation index has the consistency index CI, the stochastic consistency index RI, the consistency ratio cr=ci/ri, generally when cr<0.1, we think that the consistency of the contrast matrix can be accepted. If the consistency of the matrix satisfies the requirement, the corresponding eigenvector can be further computed according to the maximum eigenvalue of the matrix, and the eigenvector is normalized (and the 1 of each component in the eigenvector) is transformed into a weight vector, which is the result of our request. Each component in the weight vector reflects the weight of each element's influence on its corresponding upper-layer features.

Because the analytic hierarchy process of AHP to design some advanced mathematics related knowledge, need to learn more about the statistics, operations research and decision-making books and materials can also be directly on the Internet to search the analytic software of AHP, Some tools support the results of the comparison of the input indicator 22 can be directly output consistency test results and the weight coefficients of the various levels of indicators.

The above-mentioned loyalty scoring system uses the AHP method to calculate the impact weight of the underlying 4 indicators on loyalty:

Loyalty Score = Access Frequency score x0.4 + recent access interval score x0.25 +

Average length of stay score x0.2 + average browse pages score x0.15

After calculating the weight of the affected indicator, the final loyalty score can be calculated by the weighted sum method, as shown in table 6-3.

Table 6-2 user Loyalty weighted score

Access Frequency Score

Recent Access Interval score

Average length of stay score

Average page views Score

Loyalty ratings

User 1

1

5

3

2.2

2.6

User 2

5.5

9.3

2.4

3.3

5.5

User 3

0

3

6

5.5

2.8

Table 6-3, the weighted way to calculate the user loyalty score, you can directly compare the loyalty score to evaluate which user's loyalty value is higher, which is lower, the marketing department colleagues have a more direct basis for the user's choice.

Above is only the user's loyalty to do the assessment, can not reflect the value created by the user, and the second demand of the marketing department is the evaluation of the comprehensive value of users, such as e-commerce site users may have a certain degree of loyalty, but if only to see not buy, still can not bring enough value for the site, Therefore, the need to further assess the value of the output of users, e-commerce sites in particular to pay attention to this point. In order to reflect the user's value output, we need to consider the indicators related to the purchase of consumer consumption indicators, here is a list of 5 indicators for reference:

    1. 1. Recent purchase interval : You can take the user's most recent purchase from the current number of days, reflecting whether the user continues to maintain the consumption on the site;
    2. 2. frequency of purchase : The number of times the user buys in a period of time, focusing on the user's consumption viscosity;
    3. 3. Purchase Product Type : The user buys the product type or the commodity large class in a period of time, reflects the user demand breadth, may analyze the user value output diversity and the expansion space;
    4. 4. Average per consumption : The user's total consumption in a period of time ÷ the number of consumption, that is, the guest unit price, reflecting the average consumption capacity of users;
    5. 5. One-time maximum consumption : Users in a period of time to purchase a single maximum payment amount, reflecting the user's ability to pay, but also reflect the user's trust in the site.

The above 5 indicators reflect the user's value output from different angles, and are quantifiable statistics obtained, there is also a time interval limit, need to pay attention to choose the appropriate time period length. In order to be able to measure value uniformly, it is also necessary to standardize the above 5 indicators, using a 10-point output for evaluation, or using radar charts, as shown in 6-20.

Figure 6-3 User Value Radar chart

Figure 6-20 uses a radar chart to show the data of 3 user indicators to reflect the user's value characteristics, according to the attributes of each indicator can be further divided into two pieces of user value, including the most recent purchase interval, purchase frequency and purchase of product categories to show the user's purchase loyalty , While the average amount of each consumption and a single maximum consumption to reflect the user's spending capacity , in Figure 6-20 box up the two block area, the upper part of the radar chart is used to express the user's purchase loyalty, the lower part of the performance of the user's consumption capacity, 3 users of the data for analysis, The overall value of user 3 is lower, the value of users 1 and 2 is higher, and the value of user 1 is concentrated in the higher consumption capacity, the value of user 2 is more reflected in the higher purchase loyalty.

Radar chart is a good display of user value in different indicators of the embodiment, combined with analytic hierarchy process, can be a comprehensive evaluation of user value, the basis of data from the above 5 indicators of the score results, the use of AHP can not only get the final user value score, At the same time can also get the above purchase loyalty and spending ability of these two aspects of the score.

Figure 6-4 User value scoring hierarchy model

Figure 6-21 is the use of AHP to build a user value scoring hierarchy model, the bottom is 5 basic indicators, the middle tier is the user value of two aspects, respectively, corresponding to their respective indicators, the top is the user's comprehensive value. Here you need to use 3 AHP to calculate:

    1. Purchase loyalty and consumption capacity of the impact on user value weight;
    2. The weight of the recent purchase interval, the frequency of purchase and the impact of the purchased product category on purchase loyalty;
    3. The weight of the impact on the consumption capacity of the average per-consumption and single-time maximum consumption.

After 3 times 22 comparison calculation can be obtained on the graph of each layer of indicators on the impact of the last weight, just as the value of the connection line labeled, the result of conversion to a formula is as follows:

User value = Purchase Loyalty x0.67 + spending power x0.33

Loyalty = Last purchase time x0.12 + Purchase frequency x0.64 + purchase product Type x0.24

Consumption capacity = average per consumption x0.67 + words maximum credit x0.33

After deduction, we can calculate the user's comprehensive value directly with the score of the bottom 5 indexes:

user Comprehensive value score = (recent purchase interval score x0.12+ Purchase frequency score x0.64+ Purchase product category score x0.24) x0.67+ (average per credit score x0.67+ single maximum credit score x0.33) x0.33

user Comprehensive value score = Recent purchase interval score x0.08+ Purchase frequency score x0.43+ Purchase product category score x0.16+ average per credit score x0.22+ single maximum credit score x0.11

With the above formula, all levels of the score in Figure 6-21 can be calculated, we calculate their overall score according to the data of 3 users of Redatu example, see table 6-4.

Table 6-3 user Value weighted score

Recent Purchase interval score

Purchase Frequency Score

Average per-credit score

One-time maximum credit rating

Buy Product Category Score

Purchase Loyalty Ratings

Consumption Competency Score

Comprehensive Value Score

User 1

2

3

8

9

3

2.88

8.33

4.6785

User 2

7

7

6

5

8

7.24

5.67

6.7219

User 3

5

1

3

2

1

1.48

2.67

1.8727

The table not only calculates the comprehensive value score, but also obtains the two middle-tier points of loyalty and consumption ability, so that we can not only get the important users of the website by directly comparing the user's comprehensive value score , At the same time, the scoring of loyalty and consumption ability provides a powerful quantitative reference for the segmentation of users, which is shown in 6-22.

Figure 6-5 Subdivision of user value evaluation

The figure shows the value score data of 100 users, divided into 4 pieces according to the score of purchase loyalty and spending ability, from which we can see the distribution of user characteristics of e-commerce website:

    • From the C region, we can see that the user is more distributed in the area of loyalty and spending ability scoring 3, and also the most common customer base of the website;
    • The user of Area B is the most valuable customer (VIP) of the website, but the number is quite rare, possibly less than 10%;
    • In the a area there is a point dense interval (loyalty, 8~9), can be considered as the site of the senior consumer groups, they do not consume much, but the consumption is very high, if your site provides expensive consumer goods, bulk purchase services, then they may be the customer base;
    • Although the users in the D region are not strong consumers, they are loyal fans of the website, do not neglect these users, they are often the site offline marketing and brand word-of-mouth dissemination of favorable advocates.

Through the analysis process like above, we can discover some characteristics of the users of e-commerce website, and provide some decision support for the operation direction and marketing strategy of the website. If you want to develop a marketing strategy for the user, what kind of first class are you going to start with in Category A, B, C, D, 4 groups of users?

This article extracts from the "site analysis of the actual combat-how to data-driven decision-making, enhance the value of the site (full colour)"

User value Analysis

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.