The previous two articles describes the analysis of customer loyalty based on click-stream data for all sites and used in e-commerce website user analysis , and intuitive radar chart can show the user in various indicators of performance, we can The area surrounded by the radar map graphics to simply evaluate the value of a user. We may have this question, when each index is weight (such as e-commerce sites may pay more attention to the number of users to complete orders or volume of transactions in the period of time), in this case how to evaluate integrated value of the user? If you read my previous article - the Analytic Hierarchy Process (AHP) , I believe you have found a satisfactory answer, the AHP can solve this problem. The following were introduced using the analytic hierarchy process customer loyalty assessment and e-commerce website users value score:
Evaluate customer loyalty
Customer loyalty by four indicators - user access frequency, recent access time and mean residence time and average visit the page number of decisions, the four indicators there is no obvious category, and can be considered independent of each indicator at the same level , we only need to build a two-layer model:
We need to calculate the bottom of the weight of each of the indicators of customer loyalty, and only need to know the relative weight of the underlying indicators, this mainly through the expert group or survey research methods to obtain the data, it is assumed between any two of the four indicators the proportion of The following table (using 9 scale):
| User access frequency | Last access time | Mean residence time | Average browse page | |
| User access frequency | A | 7 | 3 | 3 |
| Last access time | 1/7 | A | 1/3 | 1/5 |
| Mean residence time | 1/3 | 3 | A | 1/3 |
| Average browse page | 1/3 | 5 | 3 | A |
- The data in the table is a reciprocal matrix, that is, if the proportion of users to access frequency "and" last access time "analysis that article 7 (1-9 scale, the definition of see level), then the corresponding line 2 of the form (3) the value of the value of the matrix diagonal symmetry of the form (line 3, 2) 1/7. Pairwise comparisons only need to complete the diagonal side of the data, the countdown on the other side to take the corresponding values.
Indicators between the pairwise comparison, the entire matrix of data inconsistencies, such as data on the table "user access frequency and the mean residence time", "average visit all three of the proportion of the number of pages. one should infer the proportion of the average residence time "and" Average Pageviews is 1, which is equally important, but below the proportion of the actual data is 1/3. So you can see the pairwise comparison process only care about the relationship between the two indicators, there is no derivation of the relationship, which would lead to the entire matrix inconsistency, the inconsistency coefficient matrix before computing the weight, only less than 0.1 The data matrix can be adopted.
Table pairwise comparison of the results by level analysis of law matrix operator calculation of each index rights weight, if you excel enough to be familiar with the words in excel which also can achieve here recommended using the AHP of analysis tools - the Expert the Choice in this tools inside the building similar to the above tree model, then input into the pairwise comparison results on the table of indicators, the software will automatically calculate the weight of each indicator relative to the target and the model CR (consistency probability). Import of the above model, pairwise comparisons of the indicators in the input form data, the software calculated the CR = 0.05 <0.1, so the data through the test, you can further access to the weight of each indicator, the following weighted-weighted formula:
Customer loyalty = user access frequency * 0.525 + recent access time * .056 + average residence time * .139 + average to browse page * 0.279
Based on these findings, we standardized each indicator score data weighted sum of user loyalty of the two users in a text, for example, using the 10 percentile scoring system, the calculation customer loyalty as follows:
| Score | User access frequency | Last access time | Mean residence time | Average browse page | Customer loyalty |
| User 1 | 1.0 | 5.0 | 3.0 | 3.8 | 2.28 |
| User 2 | 4.0 | 9.3 | 2.4 | 6.3 | 4.71 |
| User n | ...... | ...... | ...... | ...... | ...... |
Above the level of analysis, multiple evaluation score results weighted aggregated to a target indicators, a more direct evaluation of user, the above results can be calculated for each user loyalty rating, can sort, select the loyalty top expand directed marketing.
Evaluate user value of e-commerce website
Instance by the above analysis, the application of the analytic hierarchy process should have a certain familiarity, then the more complex multi-level analysis model, the analytic hierarchy process is how to achieve it? The following index system, for example in this article to a brief introduction to e-commerce website users . The first building evaluation index system model, the indicators obtained from the e-commerce transaction data is more abundant, further stratified by the index of indicators elaborated according to the paper, you can establish a three-layer structure model, as shown:
Based on this model, you need to use three times AHP to calculate:
- Loyalty and spending power of the weight of the user value;
- Recently purchased, purchase frequency and purchase product categories loyalty weight;
- The average each spending and the highest single amount of spending on the spending power of weight.
Indicators pairwise comparison weights were obtained at each step, the consistency of the probability of CR to calculate the matrix, calculate the weight coefficients of each indicator on the upper corresponding indicators, and ultimately we can get the following results:
User value = loyalty * 0.67 + consumption capacity * 0.33
Loyalty = recent purchase time * 0.12 + frequency of purchase * 0.64 + Buy product categories * 0.24
Spending power = average each spending * 0.67 + word * 0.33 of the highest spending
User value can derive a direct formula:
User value = (the most recent purchase time * 0.12 + frequency of purchase * 0.64 + purchase of product categories * 0.24) * 0.67 + (average consumption of each amount * 0.67 + words maximum spending * 0.33) * 0.33 →
User value = buying time * 0.08 + frequency of purchase * 0.43 + Buy product categories * 0.16 + * 0.22 + words spending * 0.11 average each spending
Example two users to calculate their overall value score, the following table:
| Score | Recently purchased a Time | Frequency of purchase | Purchase of product categories | Average each spending | Word the highest spending | Loyalty | Spending power | User value |
| User 1 | 2 | 3 | 3 | 8 | 9 | 2.88 | 8.33 | 4.68 |
| User 2 | 7 | 7 | 8 | 6 | 5 | 7.24 | 5.67 | 6.72 |
| User n | ...... | ...... | ...... | ...... | ...... | ...... | ...... | ...... |
It can be seen from the table, not only to calculate the ultimate goal indicators (customer value) ratings in the analysis of results obtained using analytical hierarchy process, but also able to calculate the index of get the model in the middle layer (loyalty and spending power) ratings, so that we can not only direct comparison of the user's value score for an important user of the website, loyalty and spending power ratings for the user segments of a strong quantitative numerical reference, as shown below:
- 100 random samples of data scatter diagram data for reference only and does not represent any actual
The scatter plot of the above is a simple display of customer loyalty and spending power, gathered from the graph the midpoint of the intensity distribution (or points) can diagram simple divided into four, to meet the points of the sub-block within the the shortest distance (most dense) and block distribution of the longest (most discrete), and in fact can be seen as one of the most simple clustering can be seen from the distribution of e-commerce website user features:
- Distributed more loyalty and spending power ratings for the three near the region, but also the most common site customer base can be seen from the C region;
- Users in region B is the site of the most valuable customers (VIP), but the number is quite rare, may be less than 10%;
- A point of intensive interval in the A region (loyalty 1-2, the spending power of 8-9), that is the site of advanced consumer user base, not more than they consume, but spending is high, if your site is to provide the luxury goods wholesale, buy services, they may be is that customer base;
- Users of the region D spending power is not strong, but they are loyal Fans of your site, do not ignore these users, they tend to be favorable supporter of the site offline marketing and brand reputation spread.
Similar to the above analysis, we found that certain characteristics of e-commerce site to provide decision support, operational direction and marketing strategy for the site.
This article which focuses on the Applications of the Analytic Hierarchy Process an evaluation of the website user, in fact, the analytic hierarchy process applies not only to the evaluation of the site users, the same applies to the web site pages, products, sources, keywords and any other involved more than evaluation of the indicators can be layered, the key is how to establish a system of indicators evaluation system. If you have a better idea of the expansion, and welcome the exchange of comments with me.
»In this paper, the BY-NC-SA agreement, reproduced please specify source: The data analysis » "user value rating system
Related Articles:
The corporate marketing department is absolutely need it, do their own planning.
Thank you to share
Write well, and hope join the bihuman, blogger is waited on by the Internet industry
Cipian "user value analysis in conjunction with the" user life cycle management, I do not know how do you think?
The _AT_ zhilavie : User value can be used as a quantitative standard for assessing the value of the user life cycle.
Later want a not so understanding of the user life cycle, including a time factor, in which life cycle by determining the user to provide the outreach program, to meet or continue to maintain high-value
The the _AT_ zhilavie : user's life cycle theory which is very important part of customer relationship management through the analysis of user behavior to assess what stage the user may be in the life cycle, and expand the targeted personalized marketing.
But I am here mainly to quantitative calculation of the value created by the user for the site throughout the life cycle.
, The user value can understand the structure of the site's users, user value distribution to understand the site is currently operating conditions
Have only recently started to learn to do site inadvertently found through Google webdataanalysis.net, very like the style of your blog and read your book feeling very fruitful, your text is very practical. Is also very easy to understand. I calculate this article is about the fifth move of the younger brother to comprehend on your blog! Lucky :) redouble their efforts, Oh! Really looking forward to the flourishing of your blog.
_AT_ happy kids karaoke : I would like to thank you for your concern.
This did not understand. . Requires very specialized knowledge. For example, statistically. . .
_AT_ mengyi : in fact, will not be difficult, and may involve a number of statistical and analytical methods, but mainly based on the implementation and application point of view, in fact, the important thing is not the complexity of the method, but choose the appropriate method to achieve valuable analytical goals .
You write user loyalty evaluation sheet data did not understand this, and I counted not the same. Look forward to doubts.
_AT_ nancy : in fact, either loyalty or value of the score is a weighted sum of the process, should be noted that each impact indicators need to be transformed into the form of the component values, rather than its actual value, the other weights make sure I used here AHP, of course, can also be used in other ways.
Well, I think I ask if you are a SNS site you want to analyze its usage, can be used AHP to analyze? Assume that the product: the blog, pictures, meager, song list, the user to use these products in order to realize the value of the user, we use AHP to analyze? Or other methods, can give me some suggestions you?
_AT_ nancy : AHP only to determine the weight of one of the methods can use other methods, AHP is just one of a choice, this is not important, it is important to determine underlying impact indicators, such as you say your product has a lot of How to quantitatively measure the user's use of these products, but these indicators must be user-generated value influential, and thus is to determine the index weights.
I can not give a specific solution, just an analysis of the idea of the specific application is determined depending on the situation.
I see, thank you for your advice.
In fact, I feel that the set weight seems highly subjective, objectivity is relatively small. The composition of the products is also the site of site settings and weight settings in the formation of the product composition is still identified. As for the method used to set the weight of more scientific this statement seems not important. Do you think?
Is considering the establishment of the SEM index evaluation system, and very enlightening for me, thank you!
Kaka ~ ~ ~
Awoken him ~ ~ ~~
Before I'm sure the other weight questionnaire analytic hierarchy process to determine the weight did not think
Learned and reproduced ~ ~
Ask a question: Analytical Hierarchy Process is a very important one is the expert evaluation system, evaluation of many experts, how ~ ~~
Vincent : is like a general pairwise comparison results will be through the diverse collection, the consistency test each sample individual does not pass the consistency test (CR> 1.0) sample data first need to eliminate, and then average calculated for all samples taken on it.
I would like to ask the test matrix is inconsistent how to do ah?
Melisa : consistency test to pass data should be removed
Thank the bloggers to share, there is a question to ask, loyalty, there are types of purchases this property how will the response to customer loyalty?
The _AT_ wingers : here the choice of indicators is not a common set of standards, the site can be adjusted according to their own circumstances. Purchase types classified loyalty is to consider the types of goods purchased more than the purchase of goods is relatively simple and centralized user trust and dependence of the site will be stronger, so into them. Because some users may tend to the particular class of goods on this site to buy other goods to select channels in other websites or line.
The last access time is an indicator?
@ industry fire : "last access time" refers to a user recently visited the site of the point in time, the original is a point in time, last access time for quantification from the current interval, can take the number of days between. This article should be in the customer loyalty analysis described here without making repeat the instructions, may cause misunderstanding, it is recommended if you are interested can read the reference below.