Retention analysis: systematic thinking sorting

2022-06-10 0 By

Editor’s guide: User retention is one of the most critical indicators of a product, and retention analysis is a very frequent business scenario in the product reversion.So how do you find and analyze retention issues more effectively?This article has systematically organized the idea of doing retention analysis, let’s take a look!Users must have don’t have to go into the value of the retained, even if the product has a high user growth, but if the user does not have to stay, finished products, as defined by the core behavior, is null and void growth, only the user to stay for the product continues to generate revenue, like community, games and other products, also need to keep enough users to maintain ecological products,Therefore, user retention requires frequent replay analysis.So how do you do retention analysis more efficiently and sequentially?This paper sorted out the system ideas as follows: Step 1: Observe retention data. Observe the retention data to see the user’s short-term, medium-term and long-term retention rates, which correspond to the three stages of the retention curve user life cycle (shock period, selection period and plateau period).Retention curves Retention rates at different stages of a user’s life cycle reflect the state of the product and user at different stages.Three periods respectively need to see how long retention rates, the use of this product and natural cycle, namely the user use the product of the frequency, such as Meituan take-away, zhihu, trill short-term retention rates and other daily use products mainly focus on the next day, like a cat’s eye film, barley, and 12306 weekly or every few weeks, the weeks left.There are three methods to determine the natural life cycle of a product: 1. According to the business experience mentioned in the above example, this method is relatively simple and can be judged without historical data.To analyze the daily retention rate, take new users within a period of time and observe their daily retention rate within 30 days. The first retention peak after the first day of retention is the product life cycle, which can be verified by the periodicality of subsequent retention peaks.Figure Daily retention rate of new users on a given day (virtual data) note:One day new users of retained = N, users of the day in the N number of active users/the day the new users, for example new subscribers to January 1, 100, the number of users active users to 60 in the third day, the third day of active retained rate = 60/100 = 60% pay attention if the use of the product cycle for a day, have no obvious peak,The daily retention rate is declining and then flattening out.The reason to observe the retention rate within 30 days is that products with a usage cycle of more than 30 days generally do not need to focus on user value through retention, but mainly focus on user experience, and drive users to become loyal users through membership mechanism and brand recognition, such as shell housing, goods and so on.3. Churn regression rate curve Churn regression rate curve can help us reasonably define a user, how long time span does not return product belongs to churn, and the product service cycle must be within this time span, define the upper limit of the product service cycle.Attrition regression mainly refers to attrition users who log in to the product again. Attrition users are those who have not logged in to the product in a period of time. This period is called attrition period.Damn homing rate = return users/subscribers * 100% through calculating the loss of different erosion period homing rate, we can draw a loss homing rate curve, with erosion stage, the greater the user homing rate is lower, after the erosion period beyond a certain point, the user homing rate will drop sharply reduce, leveling off, this point is the inflection point curve,The inflection point in the following figure is the 10th day, which means that when a user does not log in the product for 10 consecutive days, it can be judged as a real loss, and the service cycle of the product is no more than 10 days.Figure loss regression rate Curve Step 2: Judging retention Problems The key to judging whether there are retention problems lies in comparative analysis and comparison with different retention standards. If the retention standards are lower than relevant standards, there may be problems and further analysis is needed.The comparison criteria can be divided into the following three categories: 1. Comparing the time standard with its own historical performance, it should be noted that lower than the historical performance is not necessarily a problem, and it should be considered whether the product retention has periodic characteristics. For example, the retention of new users of game products on weekends is usually lower than that of new users on weekdays because the users are more extensive.2. Planning Standard Product planning retention is usually developed in conjunction with the business goals of the product.3. Specific standards are usually based on business experience, or compared with industry/competitive data.In addition, it should be noted that even if the overall retention reaches the standard, it does not mean there is no problem. We can split retention data of users with different channels/values for further investigation.Step 3: Retention problem analysis Retention problems mean user loss, so it is necessary to further analyze who lost, when, where and how to lose, in order to answer these questions, it is necessary to do the loss of user segmentation and loss point positioning.1. Segmentation of lost users to dissect the specific loss of users resulting in low retention, there are three ideas for user segmentation.(1) User sources are segmented according to user sources, including channel types and mobile phone systems. For example, a large loss of users in a certain channel may be caused by problems in the advertising strategy of the channel, which leads to the introduction of new users that are not the target users of the product.(2) Subdivision by user value There are two common methods to divide by user value, pyramid model and RFM model.Pyramid model: In order of importance, celebrity users > Professional users > Contributing users > Active users > regular users.RFM model: evaluate and rank user value by considering the three dimensions of Recency, Frequency and Monetary.RFM is relatively complex and can also be simplified for analysis according to business needs. For example, game products can be directly divided into several levels according to users’ recharge amount to distinguish between large recharge users and ordinary recharge users.User attributes include four categories: demographic characteristics, such as gender, age, occupation, education, etc.Social relations: marriage, children, old people, etc.Behavior characteristics: basic behavior (registration time, daily usage duration…), business behavior (bought special offers…).Business related: such as fitness product users’ fat and thin height, body fat percentage, daily average 8000 steps, etc.2. Attrition Point Locating Attrition point locating refers to identifying when, where, and how users lose users, and using this information to find the cause of user loss.(1) When to lose A user When to lose a user, there are three possible scenarios:In the use of product loss over a given period of time, as shown in March 21-27, during this time the new user in the seventh day retained a strikingly low compared with the expected level, can be observed at the same time 3 ~ 7 retained attenuation significantly faster than expected trends (ratio is lower, the faster attenuation), while 15, 30, is also lower than expected, but the attenuation trend with expectations,Therefore, we can basically judge that the problem of loss occurs around the seventh day after users use the product.Heat map figure retained rate (virtual) second: at a particular point in time the loss of the diagram below, on March 21-27, this paragraph of time the new user in the March 27-28 specific retention rates than this two days daily level obviously on the low side, can be further explored, this two days without exception, such as the APP is unable to login, improper operations affect the user experience, and other issues.Heat map figure retained rate (virtual data) and the third: one day/a certain period of time of the new user constantly low retained the diagram below, on March 25, the new user overall retained significantly lower than other new users of time left, this kind of situation usually import problem for new users, such as advertising, can be combined with loss of source of user segment were analyzed, and the further orientation.Heat map figure retained rate (virtual) (2) and where the loss which function module mainly refers to the user in the product, or use the product which the process of phase loss, usually see users to stay before the loss of function modules or processes, if discover the most damn users before they stay in one place, then you can focus on trying to identify the corresponding function or process.For example, if a game product finds that a high percentage of attrition players stay in a mission copy before attrition, that mission copy may be the cause of player attrition.(3) How to lose users You can usually analyze the operation logs of lost users before they are lost to check whether there are exceptions.In the third step, the clues of the reasons for user loss can be obtained by subdividing the lost users and locating the lost points. In the fourth step, the reasons for user loss can be further explored according to the obtained clues.There are two ways to explore the causes of loss: 1. Research with conditions and resources can give priority to research, including: conventional research inquiries: such as unloading research pages, questionnaires.Deep cause research: such as telephone, interview, community, forum.2. Hypothesis testing research is generally cost manpower, financial resources and time, in the majority of business analysis scenario, every time to do research, so in the case of limited “hypothesis testing” can use the second method, based on the third step for clues, according to the experience of the business do reason assumption, and then use the data validation.Examples such as the game players, lost and found before the players have a high proportion in the loss of loss of stay on a copy of a task, and these players are all 5 players, depending on the copy of business experience can assume the task of these players too difficult cause loss, specific copy can pass level 5 players in the task of passing data verification,If the pass rate is significantly low, the hypothesis can be tested.Step 5: Improve retention Once you understand the reasons for churn, how do you go about improving retention?It’s a case-by-case scenario, but the general principle, as described in The article “The Core Secret to Building a $1 Billion Product: A Hierarchical Model of User Engagement,” is to constantly increase the “benefits of continued use” and “losses from leaving the product.”(1) Demand satisfaction on the one hand, it seeks to match users’ needs more accurately. For example, for short video products, if the recommended video does not meet the preferences of some users and their needs are not met, this part of users are easy to lose;On the other hand, strive to meet the long-term needs of users, such as beauty camera products, constantly optimize products to launch new gameplay.(2) experience in there are other products can meet the demand of the same user, using the experience better, users won’t turn to choose other products, such as a taxi is also a software to meet the demand of daily take a taxi, service is a convergence, but the service experience can have differences, drops because the driver resources more, order response speed faster,Therefore, COMPARED with other taxi-hailing apps, I prefer to use Didi, which is also the barrier advantage for Didi to occupy the first place in the market for a long time.2. Improve user input of “loss of leaving the product” : Guide users to complete more key behaviors in the product. The essence of the idea is to let users invest more, including time, energy and emotion, that is, to improve the sunk cost of users as much as possible.When people decide whether to do something, they not only look at whether it is good for them, but also whether they have invested in it in the past. Therefore, the higher the sunk cost is, the more reluctant users are to abandon the product.For example, Evernote guides users to create a certain number of notes. The more notes users create, the higher the loss of users to switch to other note-taking products.In the end, the above sharing is based on work experience, spare time learning and personal thinking summary and comb, or there is not a thought, please feel free to comment.Part of the content reference books: This article was originally published by @Einstein fan Sister.The picture is from Unsplash, based on CC0 protocol.