User Recall and Activation: 7 Case Studies on User Loss and Activation

Activation and recall are also top priorities in operations , because the cost of activating old users is roughly 5:1 (or even more) than the cost of acquiring new users. In view of the customer acquisition cost, it is more hoped to recall valuable users and upgrade ordinary users to active users.

1. User recall

It is necessary to classify lost users, establish a user churn model, and then establish a churn warning mechanism to predict "when and with what characteristics" users may churn, and then adopt corresponding strategies to retain them.

1. What kind of users are suitable for recall?

Just like new user additions, user churn may occur at any time. When "new user additions > user churn", the user base will increase, otherwise it will decrease. Therefore, while we are attracting new users and increasing retention , we must also pay more attention to recalling lost users. When recalling users, try to retain valuable users as much as possible. After users come into contact with the product for the first time, we have already screened them. Users who "do not meet the needs and are not interested in the product" do not need to be reached again; more emphasis is placed on users who leave due to "poor product functions, competitive product attacks, etc." These users have already contributed value to the product and are the target users of the product. They are lost only because our products are not good enough or because of the competitive product advantage effect.

(1) When registering and logging in, in the early stages of a product, setting the login threshold too high is not conducive to product growth. In today's prosperous mobile Internet era, accounts are gradually developing towards a shared model. Weibo accounts can be used to log in to various apps such as NetEase Cloud Music and Jianshu .

For users, this login method is low-cost and reduces their defensiveness, so they don't have to worry about the security of their personal information. For ABCD's products, letting users get in touch with the products first and then encouraging them to register platform accounts is a shortcut to save the country by taking a detour. At the same time, based on the user's Weibo ID, some user characteristics can be captured in the early stages with some effort. For Weibo, on the one hand, it reflects Weibo's popularity and user trust, and on the other hand, it can help Weibo better understand what types of apps users are using. The frequency can be captured, but usage behavior data can only be owned by the platform.

After we designed the visitor mode and third-party login on the registration and login page, users still left directly. This kind of lost group of people are not considered as product users, and we do not have their information. We cannot reach them again through recall. They may download because of the benefits, or they may give up after seeing the UI style of the product after downloading. Whatever the reason, we can only reach this type of users again when we attract new users.

Case: The comprehensiveness of Tiger Brokers ’ login method

Tiger Brokers is a product for online trading of U.S. and Hong Kong stocks. Regardless of whether we play stocks or not, we should intuitively know the security of opening an account and the strictness of the process. Among them, the security of the account is particularly important. If the account is stolen or the transaction goes wrong, there will be a loss of real money and the leakage of important identity and financial information. But at the same time, registering on such a platform will also cause users to have security concerns. Many users just use it to check stock trends and look at financial reports, and have no intention of opening an account (Tiger's interest and transaction fees are relatively high). For example, I just use it to check stock trends, and there is no need to register. Therefore, registration and login of Tiger Brokers does not affect the use of products (except the functions of opening an account and trading). You don’t even need to log in from a third party, you can just go into “visitor mode”.

Maimai also authorized a Weibo account in its early days. Later, it grew stronger and posed a certain threat to Weibo's social status. Weibo banned the authorization. Maimai also established its own account system based on its own accumulation. Now it no longer relies on third-party accounts. This is also a successful case. It is conceivable that we can make some changes regarding registration and login.

(2) New user guidance and preference prediction are both ways for the product to introduce itself to the user. This type of introduction is usually sufficient for the first use. If you are required to see your self-introduction every time you enter, this is a design error of the "guidance function". In this process, there will be a lot of user loss; if the user does not like the interest tags you give him, just like today's headlines , the user wants in-depth reports, but all you provide is one-sided and eye-catching content, and all you get is a complaint. If he has not registered a Toutiao account , I believe that such a serious mismatched user will not use Toutiao.

For this type of lost users, although they may have registered or logged in with third-party information, they did not use the product afterwards. We can investigate the reasons through some phone inquiries or private messages to the account. If the product content itself is inappropriate, there is no need to recall the user. If the recommendation page or guide page is not well done, we can improve it and invite the user again.

Case: Diversity of Baidu column classification

Baidu APP divides columns into Q&A, forums , pictures, videos, technology, society and other columns according to "Baidu product series", "content medium attributes" and "content fields". It is a bit messy, but you can always find a column that suits you. Baidu's user base is too large, and it is effective in terms of applicability to a large number of users.

(3) For users who churn later, regardless of the reason, they have all used the product for a period of time and are most likely the target users of our product. The reason for their churn is not demand asymmetry, but may be due to problems with competitors or the iteration of their own product features. At this time, we need to analyze the lost users, find out the reasons, and then take appropriate measures to correct them, reduce the loss, and then find ways to bring users back.

How to conduct lost user analysis? Whether this step is done well actually has a lot to do with whether you have tried to understand the user when he first came. The official account can be labeled, the APP can also be labeled, and the intersection and union of features can be done. We have done enough analysis of the users at the beginning. When the users change, we can know which type of users have been lost and what the proportion is through the changes in the data of the feature users.

What else can we do if we don’t do user characteristics research? My suggestion is to retrieve the data on lost users and conduct a sample survey. These users have used the product for a period of time, or at least have logged in with a third-party ID. Whether it is a telephone interview or a private chat on a third-party platform, you should find some of them, understand the reasons why they left, and show your sincerity and the benefits they can get .

As we gradually understand the users who churn, we accumulate some of their behavioral data. By analyzing the behavioral data, we can find out under what circumstances users will churn. Whether it’s “constantly clicking ‘not interested’ in information apps”, “not getting a response to complaints in tool products”, “not placing an order in e-commerce apps for half a year”, or “not interacting with friends in social software for a month”, as long as these reasons exist, we can get the answers from behavioral data. Therefore, we use these indicators as indicators of the churned user model. Their weights should be adjusted according to the proportion of churned users. The model should be iterated according to the continuous changes in behavioral data, so as to predict as much as possible that a certain type of user is about to churn, and we have to take corresponding measures to retain them. Based on the retention effect, we can add the retention measures and retention effects of relevant lost users into the lost user model to form an automated early warning of "prediction-retention measures" as much as possible.

Case: Public account user unfollows

First of all, by comparing "real-time data" with "previously added data", we can find that the number of new users lost from the 12th to the 23rd was 26 as of the 24th, and the total number of users lost from the 12th to the 23rd was 32, which means that the loss of new users accounted for 81.25%. The main problem is that there is a problem in gaining the trust of new users. Secondly, the users who have recently unfollowed me are mainly those who followed me a week ago. My fastest update frequency is "one article per week". It is very likely that some users unfollowed me because they felt that I might not be updated anymore. This reason cannot be avoided for the time being. If the effect is good in the future, I will strive to integrate high-quality articles, etc., but I have no plans to do so for the time being.

Furthermore, I guessed that the main reason was that "there was a conference recently and all the information sharing links were invalid." I went to the background messages to verify it. Because there were a lot of background messages, I still found users who unfollowed me because they felt cheated by not receiving timely information.

Finally, since I found the reason for user loss, I must find a way to reduce this experience of being "cheated", so I added this sentence in the message reply: "If the shared link is invalid, I'm sorry. I'm holding a conference recently and the shared link is always cancelled. I will update it from time to time. Because the frequency is too high, I can't update it in time. Please forgive me. It's best to get it after next Wednesday."

Of course, I did not deceive the users of the official account, so I will update the links from time to time, and directly reply with the links that can be jumped in the background messages, trying my best to restore their impression of the official account. I would also like to express my gratitude for everyone's understanding here.

2. How to recall users?

Recalled users should be treated like new users, just like getting to know a friend again. Without further ado, let’s first look at how to conduct a recall. The recall methods I have encountered so far are still these three: "email-EDM marketing ", "SMS marketing", and "APP push ". They each have their own advantages and disadvantages. Generally, the three methods are used in combination, with different strategy combinations adopted for different users. No matter what method is used, we still have to focus on data in the end, that is, the reach rate of the content, the opening rate, the conversion rate of the link jumping to the APP download page (web page), etc., and use the data to measure the effectiveness of the content copy and the channel .

(1) SMS

We receive different text messages every day, which attract you to download APP in various unlimited ways. Generally speaking, there are several basic requirements for text message recall: "platform", "user interests", "download link" and other three parts. Needless to say, users only stay on the text message for 1-3 seconds.

Examples:

Vipshop , discounts, no download link to APPstore (I am using iOS)

Dianping.com, 9 yuan limited time offer, download link → "Perfect"

JD.com , overly complex discount descriptions (you can simply say "DW watches up to 50% off, 100 off for orders over 799, and a limited-time 500 yuan coupon"), download links

(2) Email

Nowadays, users have a passing impression of email marketing . Even if clicks are generated, it is difficult to convert them into product downloads. However, this method can still be used as a supplement. Mainly event invitations, discount notifications, etc., the requirement is usually best to "complete the purpose introduction with a poster image". Users spend less time reading emails than text messages, often just a click for 1 second.

Examples:

The disadvantage of Fanruan's event invitation email is that the text is too complicated and people don't have the patience to read it all in various scenarios.

NetEase Finance 's email marketing, one picture tells me the benefits, and the picture links to the financial page → "Perfect"

(3) APP push

The users to whom you push have not uninstalled the product, so push is mainly to remind users not to forget "her own needs" rather than telling users what you need from her. The push frequency should be moderate so as not to make users feel disgusted, especially Android users. They have no way to block your push, nor can they intercept other pushes. Do not push excessively if they are already disturbed. iOS users themselves block you, so if you disturb them excessively, they will block you and you will never be able to push to them again. It must be moderate. Pushing 10 messages a day like Toutiao and UC Shock Department is not suitable for most products.

Examples:

The content pushed by Keep is "It's time for me to exercise", "My weekly exercise report", "What is the best way to exercise, how to make fitness meals", etc. It takes me as the center point, and I did not block it. It is a good push case.

(4) Product linkage

Many products cannot achieve linkage within their own product system, so just simply understand it.

Examples:

WeChat often pushes the "Weekly Battle Report" of Honor of Kings . Given the active volume of WeChat, the conversion rate should be pretty good. At least when I couldn't remember the game "Heroes of the King", it was the battle report that brought me back. (Yeah, let’s play together when we have time)

2. User activation

Although a large number of users have not uninstalled the product, they are usually just ordinary users. We need to find ways to upgrade them to active users as much as possible. This is called activation promotion.

(1) Three reasons why users abandon your product

"Forgot", "No interest", "Upgrading is too difficult", etc., usually these three reasons are the majority.

Take myself as an example. After I downloaded "3wschool", I forgot to open it and use it. It really never pushed any information to me. Maybe they think that programmers are a group of people who are self-conscious and will open the APP to check the method when they need it. (I am not a programmer, I just want to understand the relevant technology)

As I get older, I have lost interest in some of the games I played before and haven’t played them for a few years. I have lost interest in shopping apps since I found out that Vipshop’s products are of poor quality. There are too many pitfalls in the ranking of Nongyao and it is too difficult to upgrade, so I gave up playing it. The apps that I haven’t uninstalled but don’t open are all because I “forgot”, “lost interest”, or “it’s too difficult to upgrade”.

(2) How to classify user reading

As long as we are doing analysis, we should never leave the data. There are corresponding indicators for activity: "daily active users", "weekly active users", and "monthly active users". The measurement standards for each data indicator should be suitable for different products, and the word "active" should be broken down into behavioral details, which means "what kind of behavior does a user have for us to consider him active?" Information apps may have opening, liking, collecting, sharing, commenting, etc.; e-commerce apps may have opening, searching, clicking on detail pages, clicking on comment pages, placing orders, paying, receiving goods, etc.; tool apps may have opening, using a certain function, sharing to show off, etc.; social apps may have opening, searching, interacting, etc. Each product has its own similar and unique behavioral indicators.

Case: Behavior Analysis of Music APP Users

(I have never made a music app, so please let me know if there are any major differences)

In my understanding, a user of a music app will download and log in, then select their preferences in the app, and then listen to the songs. If they like them, they will download them. However, people's preferences for music change over time, so downloading and deleting are constantly interacting with each other. Once users can't find the music they like, they will lose interest in the product and uninstall it. Which specific behaviors can be considered as active users? I think users who interact between listening to songs and deleting them are active users. If users keep deleting and filtering their preferences, we should be alert to possible user loss.

(3) What are the ways to increase user activity?

Keep giving to those who want, and try to give to those who might like it.

Taking the above-mentioned music APP as an example, in addition to constantly providing users with the types of songs they want, we also try to push songs, MVs, etc. that users may like. The connection between these is the recommendation algorithm, and the core is actually data mining. We must believe in the correlation shown by the data while maintaining common sense thinking. I won’t talk about data mining here, as this is a deeper topic that we can discuss in depth later.

Establish an incentive system:

"Motivate" users to take an action, and this action can help us increase user activity and provide benefits to users. Understand what the user wants and give her the benefits she wants. The membership points system + points redemption mall is the commonly used method. In the game, completing tasks and getting rewards also follows a similar logic.

Examples:

The sign-in system of China Merchants Bank’s mobile life allows you to earn more points by signing in continuously. Points can be used for lucky draws and can be exchanged for gifts.

Alipay is similar to Palm Life, which also includes points + mall.

Game task incentives represented by Honor of Kings.

Whether it is signing in to get points, or completing tasks to get coins, etc., all of them are designed through their own account system to guide behavior + rewards. This is a long-term and relatively large project, which will not be expanded in this article.

Having written here, I have finished writing about user recall and activation. I still hope that you can interact with me through storytelling and comprehensive knowledge.