Comparative analysis based on user segmentation

From the user level of the website, we subdivide the user into various types according to the user's behavior characteristic, because the user behavior is different, the behavior statistic index is different, the angle of analysis is different, so if want to subdivide the user, can realize a variety of different classification according to various rules from many angles, Have seen some data analysis reports do a variety of user segmentation, the analysis of various user behaviors, combined with various other dimensions, seems to be absolutely abundant, but it is difficult to understand what the results of these analyses are intended to illustrate, perhaps as a consulting report that reflects current overall trends and user characteristics that are indeed appropriate, But if you really want the results of data analysis to lead us to do something, or to do before the user segmentation to determine the purpose of analysis, clear business level requirements.

Since it is necessary to do a comparative analysis based on user segmentation, it is natural to clarify the differences between the behavioral characteristics of certain user groups and those of other users. This is mainly from the direction of the adjustment of content level, by comparing the different user segments of the content requirements of the differences, optimize the content of the operation, the quality of content or the content of the user's preferences recommended to the appropriate users.

Since it is based on user segmentation, first clear user segmentation rules, here are examples of 3 types of segmentation: the loss of users and retained users, new users and old users, a single purchase and two users, based on these 3 types of segmentation, each category of users to buy goods for comparative analysis, to identify which products more in line with the user's expectations.

Of course, to differentiate between lost and retained users, first of all must have a clear definition of user churn, the definition of lost users can refer to the blog before the article-the site's active users and lost users. With the definition we can do statistics and segmentation, or E-commerce site, for example, the content of the electrical business site is the merchandise, we calculate the purchase of these goods based on each product after the purchase caused by the loss of the user ratio, as follows:

The indicator definition here should be more specific, the proportion of the loss of each product should be the number of users lost after the purchase of the product in all the users of the purchase of the account, but only know that each product's loss of user ratio can not evaluate whether the product to the user to promote the retention, or to a certain extent, caused the loss of users, Only by comparing with the overall level can we draw the corresponding conclusion. So it's important to explain how the numerical value is calculated the percentages here are not the result of a direct subtraction, but the amplitude of a difference, where the overall user churn rate is assumed to be 56%, and for a commodity as an example, the result is: (58.13%–56%)/56 % = 3.8%, the same calculation method can also be used to compare the difference between the other goods and the overall range. Finally, the display, in Excel through the "conditional format" inside the data bar function can directly show the effect of the diagram, very convenient.

Obviously, the results of the above analysis are directly instructive to the operation adjustment, the aim is to promote user retention, so what we have to do is to save the goods for the user (F goods are significantly lower than the overall user turnover rate, indicating that f products more conducive to user retention) recommended to users, And the products that may lead to loss of the user (C goods) to optimize or the next frame.

Also, use the above method to differentiate the purchase bias of different user groups. The breakdown of new and old users is the most common user segmentation method, we can use a similar approach to look at new and old users of different preferences of goods:

What do you see in the picture above? The proportion of users who buy D products is significantly lower, perhaps new users do not like this product at all, and B and F goods clearly more in line with the new user's taste. If your site can be used to differentiate between new and old users of the targeted promotion, then the above analysis results will give you a lot of benefit.

Of course, the characteristics of this data presentation may have a certain relationship with the promotion channels of goods, for example, the above image of the D commodity more is the use of old users to compare the focus of the promotion channels (such as EDM), then the natural purchase of users in the proportion of old users will be high; or put some goods in the new user more concentrated landing page, the proportion of new users buying the product is also clearly high. Therefore, in doing such analysis, we need to pay attention to the differences in the promotion channels, specific problems specific analysis, can not generalize.

Using the same method can also lead to multiple purchases by the user. For E-commerce sites, the user's first shopping experience is very important, this will directly affect whether the user will be generated again or after a number of purchases, or whether the site can become loyal customers. If your site is focused on user relationship management, then you can try using the following analysis method:

It is to be noted that the base user base is set up for the first purchase of each product (not all), what we want to analyze is whether or not the user will initiate a second or even multiple purchase behavior when the product is first purchased, thus evaluating the impact of the product on the first purchase experience. As can be seen from the table above, B and f goods in the cause of two of the poor performance of the purchase, it is likely that the use of goods or quality issues affect the user's satisfaction, hindered the user to buy the footsteps. According to the analysis, we especially need to focus on those two times the purchase rate is much lower than the overall level of goods, but also need to be based on the characteristics of the product analysis, some goods are indeed easier to facilitate the purchase of two times, because there may be cross-selling and upward marketing.

In fact, I would like to split this article into a series of topics, because from the implementation level, each piece of user segmentation needs to be completed independently, and most of the data from the bottom of the calculation, if you from Google Analytics above from the search for similar data, The only thing that can be found is a new access scale, and the metrics that are subdivided into each page in the content module do not include the% new Visits (which is measured in traffic sources, geographical subdivisions), and of course you can customize the report to see the new percentage of each page in the site. The baseline of comparison is also the new access ratio of the site as a whole, and GA's presentation options directly provide the view "Comparison" with the overall comparison, and the following figure is my custom report:

The effect of GA above is similar to the result of customizing conditional formatting in Excel 2010 (2010 shows the red-green data bar with positive and negative values on the left and right sides of the axis, and 2007 does not seem to have implemented this feature), and this benchmarking comparison shows very intuitive use, In fact, the same can be used in other analyses. So what do you see from the comparison of the content of the new user in my blog? The trend in the first few articles is that the new user ratio of the conceptual methodology is higher than the average (of course, mainly by search engines), while the percentage of new users of opinion and analytical articles is below average (older users are more inclined to practice and apply), So if my blog can dynamically show different content to new and old users, then this analysis will be very valuable, maybe your site can try.

Finally, back to the beginning of the problem, it needs to be summed up is: The subdivision is used for comparison, the comparison is to reflect the differences in order to adjust the optimization, so the purpose of subdivision is ultimately to guide operational decision-making, this is the value of data analysis.