Compared with personalized push, the click rate of full push is lower, that is, under the limited number of push resources, oppo, vivo, and Xiaomi users receive more low-quality push content, so push open uv is reduced.
Apple and Huawei have no limit on the number of push items, and their push uvs are generally improved. However, compared with Apple, Huawei's push messages have no lock screen notification reminders, no ringing reminders by default, and there is a QPS limit on the push rate, which results in the growth rate of the number of push openings being less significant than that of Apple.
After reviewing this case, we decided to optimize the push strategy as follows:
For oppo, vivo, and Xiaomi users, try to arrange the personalized push strategy with high click-through rate to push in the morning to ensure the arrival rate of high-quality push content;
Replace the target distribution users of oppo, vivo, and Xiaomi manufacturers from "connected users in the past 90 days" with "connected users in the past 30 days" with higher activity.
After the new strategy was launched, the overall push open volume increased significantly.
The conversion link of push uv is "prepare to deliver-arrive-display-click", that is to say, even if push is pushed to the user, it may not reach the user's mobile phone; even if it does, it may not necessarily be perceived by the user.
From the perspective of funnel optimization, compared to optimizing the click conversion rate of the downstream "display-click" copy, if the stumbling block on the upstream "prepare-arrival-display" path can be eliminated from a technical or business perspective, the push The increase in uv has a more significant impact.
Usually, the reasons why push cannot reach the device are: the limit of the number of pushes in the Android manufacturer channel, and the permission to allow notifications is disabled by default after APP activation; Folding (Xiaomi), intelligent sorting of messages (vivo), no desktop red dot, background process is killed.
An example of how to optimize: For example, for reasons such as power saving and avoiding system freezes, the operating system of the Android manufacturer will automatically kill the background process of the app, so that the push cannot reach the user. The process is kept alive by means of association startup, self-startup, and mutual insurance alliance.
Be sure to carefully read the third-party push technical documents such as vendor channel documents and Getui Jiguang. Through the documents, we can understand the boundaries and limitations of major push services, and know how to solve optimization from a technical or business perspective.
At the same time, we can also learn about the unique functional advantages of various push services, which can help us find ways to improve push efficiency faster.
2. Rich push landing page materials
In addition to the regular news detail pages and thematic aggregation pages that can be used as push landing pages, topic homepages, ranking lists, channel pages, and recommendation result pages for search hot words can also be used as material pushes.
For example, 36Kr will push the list of daily popular content; Baidu App’s push titled “Beijing Congestion Situation” will be the search result of “Beijing Congestion Situation”; the Weibo push about celebrity gossip will not land on the dynamic details after opening. page, but the homepage-recommended channel page, and the first feed of this page is the dynamic content mentioned in push.
The clever landing page design can not only deliver richer content information to users, but also effectively improve the distribution of content services, which can be said to kill two birds with one stone.
3. Optimize user-content matching
Similar to the content recommendation system, the essence of personalized push is to recommend the most satisfactory content for users, so how to more accurately capture the user's point of interest, and optimize the "user-content" matching degree to motivate users to become active and become The important direction of push operation system optimization.
In order to clarify the matching process of "user-content" in push push, let's take a look at how the massive items in the content library are carefully country email list selected and finally reach the notification bar of the user's mobile phone?
Basically, users can be divided into three categories: interested in preference tags, no preference tags but static attribute tags, and no tags. These three types of users have different content recall methods:
1) Users of interest and preference tags
Usually they have expressed obvious demands for products and services on the site, so the system can recall content based on the matching relationship between content tags and user tags.
For example, users who frequently browse, search, like, and comment on the entity word "Wang Yibo" in Weibo will be recorded by Weibo as the preferred user of the entity word. When the latest hot information related to "Wang Yibo" is generated, it will give priority to This group of users is pushed.
Common recall methods include: interest-based classification, interest topic, interest entity word recall, etc. However, this recall method often depends on the granularity of the portrait data. The richer the label system, the more users’ preferences can be hit, and the higher the probability of users clicking on the push button.
2) Only static label users
If the user's historical behavior on the site is very sparse and has not formed any preference tags, content can be pushed based on the tags collected and generated by the App List, regional tags, social relationships, types of advertising materials, device models, and third-party user portraits.
For example, push beauty content to users with Xiaohongshu and Auntie App; push Beijing weather or hot news to Beijing users.