User operation practice: How to build a user recall system in 3 steps?

A product is like a traffic pool, with fresh blood continuously delivered through various channels every day, but every addition is accompanied by loss. Therefore, how to increase revenue and reduce expenditure has become a hard skill subject for operations students. The author will introduce to you: How to establish a lost customer recovery system.

1. User Return Strategy

The corresponding strategies and systems for user return are a big topic. From an operational perspective, we can divide it into two main directions:

1. Make sure the product is active and stop the flow

No matter how big or small your product traffic pool is currently, if you fail to do a good job of the interconnected product operations in the early stages, the cost of lost customer recalls borne by the product will become higher and higher in the later stages.

A thousand text messages don’t cost much, but the cost of 30,000, 50,000... text messages per month is an explicit operating expense.

1. Build a lost recall system

Products are always iterating, and the user life cycle will always get longer and longer. By matching the constantly iterating products with the various stages of the user life cycle, the churn rate will be reduced and remain stable.

Secondly, in the early stage, quickly conduct trial and error with minimal cost, and formulate or explore effective churn recall strategies; in the later stage, operational recall costs and team pressure will gradually be minimized and remain stable.

One more thing: the highest level of operation is not to repeat the activities, retention, recall, etc., but to build a stable and reusable operation model for each section; then the team's value overflows and try more meaningful new things, right?

Here I will mainly talk to you about the latter, how to build a different user churn recall system?

Before we begin, let’s first understand the characteristics of common recall contact channels:

2. Choice of user reach channels

Common recall channels include push, SMS, and email. Each has its own advantages and disadvantages and needs to be selected based on your own product portfolio.

(1) SMS

The main advantage of SMS recall is its high reach rate, which is undoubtedly the best choice for recalling users who have uninstalled the APP. However, since SMS recall has sending costs, when conducting SMS recall, try to screen out people who are easier to recall, increase the recall rate, and reduce ineffective investment;

Otherwise, if SMS recalls are carried out in large quantities, a small number of non-core users will feel that their privacy has been leaked, and frequent contact may cause some negative emotions and influences (for example: complaining in various channels such as post bars or forums, and this hidden cost must also be considered).

(2) Email

The main advantages of email are free cost and large volume. Email recall is a common method in foreign countries, but since domestic users have a low usage rate of mailboxes, and ordinary users have an even lower usage rate, I think SAAS/education products can try this method.

(3) APP push

APP push is the most common and widely used push method in daily life. Its main advantages are high opening rate and no cost.

However, due to its obvious advantages, this approach is often abused by certain products. If the user is classified as a lost user but has not uninstalled the APP yet, a message will be pushed to the user for silent recall.

It is worth noting that inappropriate push times and high-frequency pushes will disgust users, and in serious cases, they will directly uninstall the APP.

Now that we know the three common user reach channels, how do we recall lost users?

The first prerequisite is: you can now clearly define it!

3. Sorting out user classification and churn definition

Our product user classification: Basic version

(1) User classification: Each product is divided into multi-level and multi-dimensional user roles according to the size of the user base, which will not be discussed in detail here.

Take our products as an example: the basic user categories can be broken down into new users (within the registration trial period) and old users (who have activated paid membership);

(2) Churn definition: Based on the trigger definitions such as the core functions of the product and indicator behaviors, determine whether the user is currently in a passive, silent, or churn state. Then, guidance, reflux and other actions are carried out according to the time band. For general industries, the triggering dimensions are at frequencies of 3 days, 5 days, 7 days, 14 days, 30 days, etc.

Now that you have stratified your product’s users and defined churn, the next step is to think about: what can you do to get this user to come back?

Don’t be impatient, let’s take a look first: what has he done on the product before? Now think about why he left?

4. Combine user portraits and develop behavior labeling

In simple terms, what is here is to establish refined behavioral labels for user portraits, which can be divided into four dimensions:

1. User attributes: refers to the basic objective attributes of the user. For example: gender, age, region, occupation, etc. This dimension tells us who he is and helps us formulate the push copy in the next step: whether he is a boy or a girl; 20 years old or 30 years old; an office worker or a student, your copy cannot be the same.

2. User status: refers to the user’s current status on the product. For example: whether it is the free users who have been lost, the paid users who have been lost due to expiration, or the active users with silent warnings as shown in the above figure, this dimension tells us which users triggered which warnings.

3. User level: refers to the user’s active status within the product. Users can be divided into high/medium/low users according to their activity level, and can be further divided into high-paying/medium-paying/low-paying user groups or annual/quarterly/monthly card users according to their payment amount. This dimension tells us his value on the platform.

4. User behavior: refers to the important behaviors that users have performed on the product. For example: preference for multiple participation in activities (love to communicate and share, sensitive to titles such as benefits, free, limited, etc.), price-sensitive preference for members (multiple renewal purchases under discount promotions, which is price-driven), fixed active cycle (users have fixed product usage habits, for example: online time on weekends will be higher than weekdays or only active at night), etc. This dimension tells us the important behaviors of users on the product.

Okay, now you have sorted out what lost users are and their last behavioral status. Now let’s start building an automated recall system!

5. User Recall System Automation

(This user pool is classified as original by the author. The mind map is prohibited from being used for commercial purposes. Please do not reprint it.)

1. User pool classification

(1) User pool: As the name suggests, the user pool is the user traffic aggregation pool of the entire product system.

(2) Warning pool: When a user receives a warning, the system will automatically filter and classify them into the warning pool, and automatically reach out to them based on the rules (push copy and frequency) set in advance.

(3) Maintenance pool: After lost users receive the SMS recall copy, they will return within 3 days.

At this time, the system should not automatically classify the user into the normal user pool. I will single it out and put it into the maintenance pool for 7 days to monitor whether there will be a second loss. Is the behavior normal? Indicators such as whether it is more active than before.

If the user's return performance is stable, he will return to the user pool; if he loses again, he will enter the loss pool.

(4) Churn pool: This pool is mainly used to store users who have churned twice or more, users who have not returned after being reached for more than 30 days, and users with zero value. These users have no obvious value to the product, so there is no need to waste SMS costs. Instead, good operational resource planning should be done.

1. Why do we need to segment the user pool?

(1) In the traditional recall model, users are defined from new-active-churn-recall. But think about it carefully: which one is more likely to succeed: waiting until the user leaves before recalling them or trying to retain them when they are about to leave?

Waiting until customers leave before taking the next step is a crude operational mindset.

The correct operational idea is - I want them all!

This brings us back to the open source flow control module mentioned at the beginning. Through data, we can monitor user activity. Let me give you two simple examples:

1. User A has been active (triggering key behaviors) 5 days a week for four consecutive weeks, 3 days last week and 1 day this week. This user has the possibility of being silent or churned, and can be triggered based on the overall user life cycle of the product.
2. User B's membership is about to expire in three days. He has logged in for two consecutive days but has not renewed his membership. This can be triggered to determine whether there is potential loss of membership.

The dual-engine mode of stopping the flow and recalling will greatly increase the recall rate of traditional thinking. This section will be left aside and written in another article.

(2) When the warning user enters the warning pool, the backend automatically groups the user according to their current status and behavior tags. This is similar to the PMF model, but with more refined dimensions:

1. The behaviors and labels of free users are relatively simple and can be split in multiple dimensions.

2. Paying users can refer to the PMF model for multi-dimensional matching, depending on the scale of product users, generally 8 or 16 are sufficient.

3. After grouping, clean up users with low-value tags and dimensions, such as those with low activity in the past 30 days, those who have not paid, those who have not participated in activities, and those who have not touched

The user will not be able to bring value to the product after being recalled, so it is not recommended to waste budget.

And so on, trigger the warning behavior - put it into the warning pool - classify - SMS self-trigger. In this way, there is no need to operate the SMS list - edit SMS text - submit triggers every day, week, and month, which greatly reduces labor costs and time.

(3) After a user is called back via a text message link or text message, he or she should not be able to return to the user pool normally. Why?

• For example: After user A is recalled, the system will automatically mark the user as a normal user. If the user triggers silent or churn behavior again within seven days, a wave of SMS triggers will be automatically sent to him, which is not suitable for a refined operation model.
• Lost and recalled users are generally unstable. In the traditional recall model, if the recalled user triggers the SMS and returns within 1-3 days, the recall is considered successful. But will this type of users churn a second time within 3-7 days? Do you want to follow up and observe? Is this considered an effective recall? Operations are needed to answer.

If you agree with my operational thinking above, then the maintenance pool has the following two uses:

1. The recalled users are placed uniformly and their behavior is observed for seven days after the recall to determine whether the processes such as the text messages, activities, and product guidance are optimal.
2. By placing recalled users in a unified manner, it is possible to distinguish: which users are retained within seven days and which are lost for the second time; which behaviors are triggered within seven days, leading to the different results, and whether it is possible to guide users to have an "aha moment".

(4) If users in the maintenance pool churn again within 3-7 days after returning, this means that such users have no obvious demand for the product, and vice versa, and should be permanently thrown into the churn pool; if they are classified as normal users, frequent text messages will cause harassment, which is detrimental to the product brand.

6. Design targeted copy

After cleaning and grouping the users in the early warning pool, we now need to have different push copywriting or activities to stimulate recall. I will take my own products as an example and briefly explain them as follows:

1. Title content

You know the upper limit of the number of words in a text message, but I know the maximum number of words that will be displayed when a text message is pushed and the message bar is not opened on common phone models. You have definitely not counted this (in fact, refined operations are not about doing big things quickly and decisively; this is what refinement means).

Therefore, in addition to the first paragraph of the text message must contain the product name, the second paragraph must learn to make good use of eye-catching words such as "benefits", "limited edition", "free", "promotion", "activity", etc. to increase the click-through rate of users such as activities or membership expiration.

1. Contact Content

In addition to using more commonly used eye-catching words for activity tags, for users with high activity and whose membership has not expired, it is more effective to connect the content with user needs.

To put it simply, this type of user usually enters a short-term dormant state due to temporary problems with product experience and functionality. You can try words related to product attributes such as "latest", "update", "iteration", and "fix".

1. Confessional Content

When the user remains indifferent after being recalled through the appeal content and has been lost for more than 30 days, the user can be automatically placed in the loss pool to avoid SMS harassment and cost waste, because the user has made his attitude clear.

The last text message can be used to play the emotional card. Since you can’t become friends, you should leave a good impression of your product: warm and humane, rather than the kind of product that keeps pushing text messages until the end.

For example: Dear sir, you have been away for a long time. The butler is very grateful for your use and support. I can’t help but quietly renew your 7-day membership and look forward to your return at any time! Be happy~

To achieve more content reach, you can do more AB testing on copywriting and landing page activities based on the product tone.

Finally, regarding the control of the time nodes for SMS push, it is traditionally between 19:00 and 21:00. You can also conduct customized tests based on the peak user activity periods of your own products. Same as above, modify the copy more often, do more tests, and speak according to the data until you get the best result.

Readers of my articles know that each of my articles is based on my own practical experience in product operation, and sharing them is also a review and summary of my own work. Therefore, telling the truth with data is the tone and commitment of my articles; I refuse to use theoretical insights throughout the article and lead readers to imagine. It’s better to attach a picture of yourself:

Attached is the return effect of the recall system for some lost products, only the SMS recall end.

Related reading:

1. User operation: How to build a system to recall lost users?

2.5 steps to build a complete user recall system!

3. How to build a user recall system from 0 to 1?

4. User operation examples to create a lost user recall system

5. How to recall lost users at low cost?

Author: Mao Li

Source: Operational Growth