Do users uninstall the APP as soon as it is pushed? Pay attention to these 3 key points when doing push

As we all know, the cost of acquiring users is getting higher and higher. With the rise of mini programs , users' willingness and cost to download an app, register, and fill in information are also increasing exponentially.

In response to this trend, we should first cater to the needs of new users. For new users, we should simplify the product usage. For old users, we should try our best to prevent them from uninstalling or leaving us.

First, let’s talk about ways to reduce uninstalls using message push , a very common and important method in app operations .

First of all, we need to find out the dimensions that affect the quality of App push , what specific actions are included, and what are the key nodes for App uninstallation.

1. Three key points that affect the app uninstall rate after app push

(Right click and save as to view the high-resolution image)

2. Determine push scenarios based on push copywriting , sending technology and frequency, and user attribute grouping

1. Push copywriting

Classify push copywriting types, determine company attributes, and then choose the copywriting style and wording. Push types are mainly divided into IM type and sub-IM type. Here we mainly talk about non-IM categories.

(1) Activity notification

The purpose is to tell users that an activity or something important is about to begin and hope that users will participate. To increase user engagement, the contradiction lies in the degree of participation intention between operations and users.

The operation hopes that users will participate as much as possible, and will be happy only if the content is pushed 8 times a day. The user's perception, apart from the interest in the activity itself, is the guidance of the copywriting. Therefore, the copy should attract the user's attention in the shortest possible time and stimulate the user

Click. This type of push is most likely to cause user disgust.

It usually starts with a number + a recognizable interjection such as "Oh my God, I can't believe it".

(2) Information consulting

Mainly covers knowledge + news apps. It usually focuses on hot current affairs news and personalized subscriptions.

Therefore, this type of push mainly describes an event in concise and clear language.

(3) Product promotion

Its main purpose is to encourage users to make purchases, and it is mostly used in e-commerce and financial products. The purpose is to create a sense of tension for users and guide them to place orders from an emotional perspective.

Usually it is mainly based on tense words such as product brand + about to be sold out / descriptive words such as appearance and practicality.

(4) System prompts

The target users are mainly the owners of the devices, and it displays the upcoming behaviors of the mobile phone system or the behaviors that the owners have done based on the app.

This type of information mainly describes facts, and accuracy of information is the first requirement.

2. How to arouse user interest through push notifications

Push notifications can be roughly divided into the four categories mentioned above, but based on the four categories, the expressions vary greatly. Therefore, when determining the copy, we need to think from the user's perspective and think about what kind of content he wants.

Basically use the following 5 points of thinking:

(1) Word limit for push copy

(2) Use of punctuation marks, such as “～、》？……”, etc.

(3) Use of emoji

(4) Use of numbers

(5) Use of emotional content in text

To attract users’ attention, you can refer to the AI ​​DA model:

Detailed explanation

(Right click and save as to view the high-resolution image)

3. Push method and frequency selection

A successful push message meets three dimensions: the right time + the appropriate scenario + text that meets the needs. So how can we find a way to satisfy these three points? There are four ways to try:

(1) Multi-copy test push

This method is to click the latter formula first, test it on a small scale, and then push it all. First, the operations staff is required to write multiple copies based on the needs and current product user groups, and then randomly select a number of users from the user pool and distribute them equally.

This method can effectively and quickly find the tonal copy that users like. The choice is given to the user, not decided by the operator or the boss. Avoid generalizing and being self-satisfied blindly.

Example: Push user number 5 million

(2) User personalized push

This method is mostly used in platform/information integration apps. There are many types of users for this type of product, and their needs are inconsistent. E-commerce platforms include Xianyu, Taobao , and JD.com . Information integration categories are similar to Zhihu, Jike, etc. Educational categories include English Liulishuo, cctalk, etc.

Pushing corresponding copy based on the user's personalized needs can not only improve the conversion rate but also satisfy the user's wishes. Improve user satisfaction and loyalty to the product.

If you want to meet users’ personalized needs, you need to group them and conduct detailed user behavior analysis as the basis for personalized push.

Personalized push must be based on the understanding of the user. Online, it can be based on the app usage path, information browsing, etc., and offline, it can be based on the user's life radius to gain knowledge of the user.

(3) Semi-automatic push

Semi-automatic push has a wide range of applications, including in-station messages, emails, text messages, etc. When more and more information needs to be pushed, it is necessary to add push rule settings to save the manpower costs of operation members while ensuring the timeliness of information transmission.

It mainly focuses on time nodes + specific user groups + content determined in advance.

(4) Push filtering and deduplication

Now that we have explained how to push notifications to users, if a non-IM user receives n push notifications every day, he will go crazy even if his needs are met. Therefore, based on the push, we need to limit the frequency, basically 1 post per day, no more than 2 posts.

After deduplication, push classification issues need to be considered. Push priorities are determined to minimize harassment to users.

System-level push is the highest level, mainly focusing on major events. Attention-level push mainly pushes information that users themselves are interested in, such as news types and products available for purchase. The lowest level is the operational push, which is mainly based on parallel activities. It is not advisable to push too much to cause user interference.

4. How to choose the push time

The same copy will have different effects when pushed to users at different times. At this time, we need to break down the copywriting into which type of users it is for. At this time, we should consider four main points:

Among them, the most important is the work and rest time. We have two methods to test it.

(1) Browsing status by calendar day/week

As the name suggests, for a certain page, we need to eliminate many interference factors, such as organizing activities, sending push notifications, etc. Allow browsing behavior to be natural. At the same time, through the data selling points of R&D colleagues, the normal 24-hour PV/UV data of one or more pages is detected.

The purpose of detecting multiple data is to ensure the accuracy of the conclusion. It is much more difficult for other action data to interfere with one page than to interfere with multiple pages.

Through a survey questionnaire, we can detect the user activity time period:

Small tips : By detecting the pages with the highest user views, we check PV/UV. Sometimes bugs will be detected. First of all, we all know that uv/pv=1:2-3. Suddenly one day uv/pv=1:4-6, then there may be a problem with this page. Either the user can't understand it, or the page can't be refreshed.

(2) Message test

Even after data testing, we still have to try it through actual actions. Although data will not lie, it will hide some information. For example, high user browsing behavior does not mean high paid conversion , so it still needs to go through real testing. What the data allows us to do is to test within an effective time period, rather than blindly choosing time.