How to do a good job of user recall? Share 4 tips!

Last year, I was responsible for user recall. User recall actually depends largely on message push , that is, pushing the right content to the right person through the right channel at the right time. The right content and the right channel are quite simple things. The difficulty lies in the right time and the right person (for example, the number of App activations and big V users who have dropped for two consecutive weeks)

I have recently systematically optimized message push, so I decided to talk about message push.

For message push optimization, we first have a clear goal, which is to optimize the final effect, but the final effect = push success rate, reach rate, click rate, and user-content matching rate.

Therefore, if you want to optimize the final push effect, you have to optimize each of the above steps, optimize each element on the right side of the formula, increase the push population, improve the push success rate, improve the reach rate, improve the click rate, and improve the matching degree between users and content. Isn’t the goal much clearer after breaking it down in this way?

Next, let's talk about the operations required for message push in chronological order to see which steps can improve the above numbers. They can be divided into the following in chronological order:

——Before push: message creation, selection of push audience, selection of push channels

——Pushing: message push

——After push: message received, message clicked, data feedback (optimization can be performed based on data feedback)

The article actually has two lines. One is to split the message push into reach rate, click rate, conversion rate, etc. according to the formula. The other is to split the message push into each step according to the time dimension. Then, you can see at each step how to optimize the various data of the first line.

Before push

1. Message production

Here are a few points that need attention. How about the cover image of the push? Is the quality of the pushed title attractive? What is the quality of the pushed content? Are the title and content hot topics? Are the title and content relevant to the present situation? Are the title and content displayed taking into account the size of the phone? Is the landing page defined?

2. Select the audience to whom you want to send messages

It is actually quite difficult to select the corresponding group of people in this step. It is necessary to create various labels on the push system, and each time, the push users are decided by selecting the intersection or union of these labels.

You can also create a user set to avoid having to reselect the intersection or union of multiple tags every time. For example, select users with a level greater than 3 ∩ Male ∩ iOS system ∩ who have paid in the last three days. If you often send messages to such users, you can create a user set for them and select that user set each time you push messages.

Let's talk about how these tags should be defined.

The most talked about thing on the Internet is of course to divide by demographic perspective, such as city, age, gender, etc., but in the actual process, we will find that if we only divide it by demographic perspective, it will be particularly unreasonable, and we need to add many business-related categories , such as rewards in live broadcasts, rewards in dynamics, and rewards in private chats. These screening conditions can be divided into:

——Version condition: You can filter a specific version

——System conditions: Can filter Android, iOS

——User conditions: some demographic dimensions, such as age, gender, city, etc.

——Account conditions: total account, or account balance of a certain business, such as screening whether there is money in the wallet, whether there are gifts in the backpack, etc.

——Payment behavior: whether there is payment, how much is paid, and through which channel

——Active behavior: activating the App, posting updates, etc.

——Content consumption: enter the live broadcast room, interact dynamically, etc.

The specific details will be determined by your own product. The filtering conditions are generally composed of three parts: major filtering conditions, sub-filtering conditions, and time. For example, payment behavior can be filtered out or not, and the payment amount range and payment amount channel (such as live broadcast room, dynamics, private chat) can be filtered in the sub-conditions. In the time dimension, you can filter 1 day, 2 days, 3 days... and enter the number of days yourself.

Whether the granularity of the screening conditions should be so fine depends on the product stage and the specific product. There is no universal formula.

In addition to the above filter tag push, it is also necessary to support input ID for push

3. Select the push channel

The selection of push channels is quite simple. Generally, there are only a few channels: SMS, internal mail, email, push

4. Select push time

For larger companies, sending push messages may require several levels of approval, rather than being pushed out immediately after creation.

There is also a time limit for users. You need to limit which time periods cannot be pushed and which time periods can be pushed. This needs to be determined based on the specific usage scenarios of your product, but it is generally 7-11 o'clock (imagine receiving a push notification at 2 o'clock in the morning and waking you up, how would you react?)

For the sending time limit, you need to consider which messages should be counted in the limited time and which messages should not be counted. For example, advertising and marketing messages cannot be pushed to users between 23:00 and 7:00, but verification codes are not limited. You can't prevent users from receiving verification codes after 23:00.

Pushing

1. Message push

Let's first talk about how the message push process reaches the user's phone. There are some minor differences between Android and iOS:

iOS message push process:

——Manual push operation: initiate a push request (own server, or third party) → APNS (Apple server) → iPhone (terminal) → pop-up new message (terminal) → open App (mobile application) → enter landing page (mobile application) → start the process within the activity (mobile application)

——Trigger push: event trigger (own server, or third party) → APNS (Apple server) → iPhone (terminal) → new message pops up (terminal) → open App (mobile application) → enter landing page (mobile application) → start the process within the activity (mobile application)

Android message push process:

——Manual push operation: initiate a push request (own server, or third party) → Android (terminal) → pop up a new message (terminal) → open the App (mobile application) → enter the landing page (mobile application) → start the process within the activity (mobile application)

——Trigger push: event trigger (own server, or third party) → Android (terminal) → pop-up new message (terminal) → open App (mobile application) → enter landing page (mobile application) → start the process within the activity (mobile application)

It’s a little better for iOS, as they are all Apple products. At most, the user may have disabled the notification permission, which results in the failure to receive the notification.

Android has a lot of problems. You can't receive it when the process is ended, and you can't receive it when the notification permission is turned off. If you integrate the manufacturer's push (the advantage is that the end of the process can also be pushed to the user's phone), the disadvantage is that if you only connect to one manufacturer, other Android phone brands may not receive it. For example, if you connect to OPPO or Huawei phones, they may not receive it.

Here we only consider allowing Android users to receive messages when the process ends. There are many other issues to consider in the message push process. I will list them down for you to consider.

For example: Which third-party company to connect to? Are there any conflicts between the company that receives push and the company that receives im (the Jiguang and Yunxin that our company accesses are in conflict and cannot be accessed at the same time)? How much concurrency is supported? Will there be a delay of several hours before users receive the message? What are the policies of various manufacturers?

After push

1. Message received

There are mainly a few questions. Let’s list them first and then talk about them in detail.

Frequency Limit

——Wake up each other

——The impact of push

——Guide to open notification permissions in the App

1) Limit frequency

To give an extreme example, what would you do if Taobao pushed 500 messages to you every day? Maybe they just uninstalled the app directly, so the frequency of message push to users needs to be considered.

For example, a maximum of two messages can be pushed to users every day. If more than two messages are pushed, the automatic push will fail.

However, you need to consider which messages should be counted. For example, advertising and marketing should be counted, but verification codes should not. It may be normal for users to receive 5 verification codes per day.

One point to note is that some students may think that other apps push more than ten messages a day, but I only push one or two. Doesn’t this take up less user time?

Message push cannot be thought of in terms of the tragedy of the commons (you can search for tragedy of the commons on Baidu). When a user uses an app, he can choose to uninstall it or turn off notification permissions, unlike in the tragedy of the commons where there is no choice.

2) Apps wake each other up

Assume that both Jike App and Taqu App are connected to the same third-party push manufacturer. If Jike App does not terminate the process and Taqu App terminates the process, when a message is pushed to Taqu App, Jike App will wake up Taqu App so that Taqu App can receive the message.

However, Google has restricted this wake-up mechanism in Android N, so this method can be discontinued. Now it is more reliable to connect to mobile phone manufacturers and push through manufacturers.

3) The impact of push notifications has both advantages and disadvantages

Benefits: information notification and reminder, promotion of activity, enhancement of stickiness, awakening of silent users, improvement of retention, and increase of functional module usage rate

Disadvantages: harassing users, increasing uninstall rate, overdrawing trust, "the boy who cried wolf", too much worthless content, causing users to be disgusted or even numb

4) Instruct users to enable notification permissions within the app

This is a problem that exists in both iOS and Android. If the user manually turns off notification permissions, push notifications cannot be delivered to the user anyway, so when the user uses the App, you need to try your best to guide the user to enable permissions.

If the user has turned off notifications, you need to determine at some point whether the notification permission has been turned off. If it has been turned off, you need to guide the user to turn it on.

2. Message click

Points to consider when clicking on a message: What is the quality of the content? Is the title appealing? How well do users match content? Is it hot content? Does the notification contain personalized content such as user nickname? Is it supported to jump to the corresponding page? Is the jump smooth? What is the quality of the landing page?

This is actually the result. The quality of the result here will depend on the steps before sending, whether the right information is sent to the right user when screening users, the quality of the message when it is produced, etc.

3. Data Feedback

After sending, you need to count the data of each step, compare it with the industry, and compare it with your own data on a month-on-month and synchronous basis. The data that need to be counted include the number of people who should be sent, the number of people who were actually sent, the number of users who successfully sent, the number of users who successfully reached, the number of users who clicked, and the number of conversions at each step after jumping to the landing page. In addition to these data, you also need to pay attention to the user's uninstall rate (the number of users who uninstalled the App within 1 hour after receiving the push / the number of users who received the push).

4. Push records

When designing a message push system, one part is message push, and the other part is message record after push. This is the prototype of drawing. It is very simple. I just want to mention a few points that need attention.

——Push progress, such as creation success (supports cancellation of push), pushing, push success, deleted, etc.

——Data after a single message is pushed, such as click-through rate, conversion rate, number of push users, etc.

other

How to monitor uninstall rate (I don’t know)

——How to determine the priority of messages? Since the frequency of message push is limited every day, messages with high priority will definitely be pushed first.

——How to make each user receive the maximum number of messages he can bear (for example, some users can accept a maximum of 2 messages per day, and they will uninstall the app after more than 2 messages, so just push 2 messages to them every day; some users can accept a maximum of 8 messages per day, and they will uninstall the app after more than 8 messages, so push 8 messages to them every day)

——How to do A/B testing, for example, take out 10% of users, push the 5 prepared copy to 2% of users respectively (10%/5=2%), and see which copy has a higher click rate in a short period of time, and then push this copy to the remaining 90% of users

——Distinguish between manual push and triggered push. Most of the above are manual push. There is also a category called triggered push, which pushes a message to the user when various conditions (an event) are met, such as today is your birthday, you haven't logged into the App for 3 days, recharge is successful, coupons expire, etc.

——If the mobile terminal and PC terminal are online at the same time, what is the push logic?

Author: Zeng Jun

Source: Zeng Jun