This logical derivation is relatively easy, but it is still very challenging in practice. The main difficulty is for specific products, what are the key behaviors that form expectations A and B, and how to create them without errors?
Take our product as an example, we are a content consumption product. The formation of Expectation A is difficult for users to understand through traditional brand promotion or preaching; we usually select representative content from the product and use it as an advertising material after certain processing.
For the formation of expectation B, content products must be the first few contents browsed by users are crucial to the formation of expectations, but it is more difficult to define what the expectation represented by each content is; we tried to associate country email list and define several times , and finally found a good pattern in the data.
In order to ensure the consistency of expectations, we assume that the expectations A formed by a material content type are the same from the users of the creative channel, and then count the content browsing of users when they use the product for the first time, which is the formation of expectations B; users The proportion in the browsing history that is consistent with the content form of the material, which can also be called the degree of coincidence, is approximately the expected degree of coincidence.
In order to avoid the influence of the length of user browsing records on the retention rate, we selected users in the same range of content browsing distribution for comparison (the data has been shifted and desensitized, the specific values have no reference meaning, only indicate the trend), and finally the average obtained The relationship is shown in the figure below.
How to use "expectation management" to improve "user retention" and "life retention"?
From the figure, I can see that when the content type that new users see after opening for the first time has a higher degree of overlap with the creative type they saw at the beginning, the retention will change significantly, and the degree of overlap will increase from 10% to 50% In the process of user retention, the user retention has improved significantly, and then the degree of overlap is increased, and there is no further improvement trend in retention.
This is basically in line with our previous expected management view. For example, if "Pharaoh" downloads our product after seeing the content of a young lady dancing, his expectation may be "there are a lot of beautiful young ladies here, let me take a look".
If you see very little "Miss" after the download is opened, it is basically inconsistent with expectations. "Pharaoh" will basically choose to swear and then uninstall it; but if the degree of overlap is high, you can see more "Gao Yan" "Miss Value", which is more in line with the user's expectations, or even exceeded, and the retention will also increase.
03 "Expectation management" to cure the poor and retain
According to the analysis of the aforementioned reasons for retention, we can conclude that the principle is to give "Pharaoh" a correct and reasonable expectation and truly realize this expectation, and don't fool him.
Therefore, the problems of "expectation management" can be analyzed from the following two aspects:
1) Create a correct and reasonable expectation A, and try to avoid the situation of "I have taken off my pants, can you show me this?"
And found that many materials with lower unit prices have a characteristic, that is, the display of particularly exquisite video content, many of which are still life and scenery; and accompanied by beautiful music, this material is extremely It is easy for users to create the illusion that this is a material download library, with a lot of ambiguity.
2) If the advertising material creates a more reasonable expectation A, the actual success of the product must be acceptable, and there should be no problem of "Li Gui for Li Kui".
For some new broken circle materials, we looked at the content browsing situation after users entered for the first time, and the proportion of overlapping materials with broken circle materials is very low, especially in the consumption of the first three content; I hoped to see "Miss and Sister" sing, but I saw a group of "Grandpas and Aunts" doing square dances.