How to recall users? Reviewing the 5 major strategies will teach you how to return scientifically!

The biggest headache for many operations colleagues is recalling lost users. Marketing partners are busy building channels to attract new users, and they have worked hard to attract users at high customer acquisition costs. However, if there is any negligence, such as problems with user experience or revenue, users will no longer participate in the platform.

The operation department needs to find these lost users who have abandoned us and win them back under controllable costs. The marketing department is like the one shouting "come and play, uncle", while the operations department also has to tell the uncle to come and play often when he is free.

Everyone understands the principle of increasing revenue and reducing expenditure, but it is not easy to put it into practice. There are users above who demand high returns, and there are leaders and finance below who clamor for reducing the budget. How to spend money wisely is a science.

In fact, as I understand it, user growth is about doing two things: one is to achieve the same amount with less money, and the second is to bring more amount with the same money.

Today we will start with the strategy of recalling lost users. There are many names for recalling lost users, such as "resurrection", "awakening", "reflow", etc. In the following text, we will refer to recalling lost users as reflow.

This article reviews the return strategy and analyzes future prospects to share my experience in operating return users over the past eight months. It can be said that this is the user growth experience that I have continuously explored and summarized during this period.

First, I use the rhetorical question analysis method that I often use to analyze the iteration of the strategy through four steps. The specific analysis mind map is as follows:

Analysis Overview

1. How much budget does the company give for reflow?

2. What strategies did we use this money to implement?

3. How effective are these strategies?

4. How to improve the strategy later

1. Give each returning user a budget

We can calculate the budget for returning users using the following formula:

1. Return cost = user contribution cost - market cost - operating cost

2. User contribution cost = (investment period/12) life cycle cost rate

3. Life cycle cost rate: 4% in 1 year, 2.5% in 2 years, 1.15% in 3 years

Logical explanation: The total cost that a user may pay in his or her life cycle is determined by the user contribution value and the life cycle cost rate. The total cost minus the initial market investment cost and operating cost equals the return cost.

We combine formulas 1, 2, and 3, and take the life cycle cost rate as 2 years, so we have the following formula:

Upper limit of return cost = (investment period/12) life cycle cost rate - market cost - operating cost = 150 yuan

(PS: The life cycle cost rate is calculated based on 2.5% for 2 years)

That is to say, the company can allow spending (CPA) of 150 yuan on each lost user, but because I am a stingy person, under my control, the average CPA per person is only 85 yuan. That is to say, in the future, I can be more flexible and not be too stingy while ensuring the return effect.

2. What strategies did we use this money to implement?

Activity Gantt Chart

The Gantt chart above lists the investment-to-production ratio, utilization rate, upper limit and average investment per user (ARPU) of each activity. After continuous exploration and experimentation, we finally use the C+D strategy. Why do we choose this strategy? Please see the figure below:

Rendering

The horizontal axis is the utilization rate (quantity), the vertical axis is the ARPU value (quality), and the diameter of the circle represents the production ratio

1. Strategy A has a low investment-output ratio, while the user ARPU (average purchase value) reaches 15,000 yuan. It is not difficult to see that the upper limit of A cannot meet user demand. However, the upper limit of 10,000 yuan will also cause many users to only invest 10,000 yuan (and leave after using up the red envelopes). This is actually the case, so it reflects the irrationality of A. What we need to do is to increase the upper limit, reduce the fee ratio, and increase the user's ARPU value.

2.B is the most commonly used strategy at present. Based on A, it increases the upper limit, reduces costs, and adds rules for the use of short-term products. What follows is a low proportion of long-term products, which means that users purchase more short-term products after returning. This is something we don’t want to see because relevant research shows that long-term users are of higher quality than short-term users.

3. C restricts short-term products based on B and can only purchase long-term products. Although the utilization rate is not as high as B, it guarantees the ARPU value, the proportion of core products (long-term product purchase rate), and reduces the cost ratio. It is a relatively balanced strategy.

4.D is a red envelope set for high-quality users. It has the lowest usage rate, but the quality of users who use it is very high.

5. C+D is the strategy we have chosen so far. It is very balanced in terms of the upper limit of the amount, the core rate, and the cost, taking into account both low-quality users and high-quality users.

The above are the strategies we have implemented and the evolution history of the strategies. We have analyzed the overall activities above. Next, we will go into the user level, from macro to micro, and deeply analyze the effectiveness of the activity strategy.

3. How effective is the activity?

Using the RFM model and whether the user is core (whether they have purchased products for more than 12 months before returning) for analysis, I named this model the C-RFM model.

C-RFM model analysis method

1. Core users (core)

Let’s first look at the comparison between core and non-core users:

Comparison chart of core and non-core users

(1) The ARPU value of returning core users is 1 times that of non-core users, and the reinvestment rate is slightly higher than that of non-core users;

(2) However, the CPA of core users is 30% higher than that of non-core users.

Summary: Whether the user is core before purchase is the main criterion for judging the quality of users after return.

2. Recency

Loss time distribution chart

(1) The shorter the reflow time, the easier it is to recall (derived from internal reflow analysis);

(2) CPA will be higher in the churn period from April to November, and lower in January to March and December and above;

(3) The cost-to-performance ratio shows a decreasing trend as the loss time increases;

(4) The capital reinvestment rate (subsequent investment/returned amount) will show an increasing trend after the outflow exceeds 7 months.

Summary: June and December are the return flow cycles. The money of lost users is nothing more than capital shortage or capital inflow into other platforms. When the user's funds are no longer scarce or the product funds are redeemed, he is actually in the decision-making period for purchasing the next product. Therefore, June+ and December+ are the return flow cycles. The cost can be increased or the cost can be tilted on this type of users to prevent funds from flowing into other platforms again. At the same time, the quality of this type of users after returning is relatively high, which can bring higher profits to the platform.

3. Frequency of investment before repatriation

Frequency distribution diagram

(1) The correlation between investment frequency before repatriation and quality after repatriation is very low;

(2) The frequency before reflow is proportional to the CPA at the time of reflow. The more times a user purchases before reflow, the higher his or her reflow CPA.

Summary: As a variable of user return quality, the importance of pre-churn investment frequency is very low and there is no obvious regularity.

4. Pre-reflow peak (Monetary)

(1) The peak value before repatriation is positively correlated with the red envelope ARPU value. Users with higher peak values ​​will have more repatriation investment.

(2) The reinvestment rate of funds is relatively high among users in the peak range of 40,000 to 100,000 yuan, proving that there is room for development among users in this range.

Summary: Historical peaks are strongly correlated with user quality and can be used to determine their quality after return.

The real value of analysis is to understand the past and look forward to the future. By analyzing the C-RFM dimensions above, we can find the patterns of users and understand them more deeply, which will greatly promote the subsequent strategy iterations.

IV. How to improve in the future

(1) AB test of SMS contact copy:

Through the AB test of the reach SMS, we can timely adjust the optimal SMS, notify users, and increase the return rate. Sometimes it is not that the user does not want to come back, but that the user did not notice or ignored the red envelope sent to him. By the time he finds out, it is too late and the red envelope has expired, missing the golden return time. Therefore, how to improve the presence of the reach SMS is our future topic.

(2) Change the existing red envelope strategy:

Strategy one is to increase the face value on the surface but actually reduce the fee ratio. For example, a red envelope with a 100 yuan discount for orders over 20,000 yuan and a 200 yuan discount for orders over 50,000 yuan may seem more tempting at 200 yuan, but in fact the fee ratio of the latter is 0.1% lower than the former, which is the reduction shown by the orange dotted line in Strategy 1 in the figure below. The fee ratio is reduced while ensuring the utilization rate and ARPU value (in fact, I am already stingy enough now, I didn’t expect that I can be even stingier, haha); Strategy two is to increase the fee ratio in a real way, spend more money, and improve both quality and quantity. As long as it is within the budget, you can try it in the future.

(3) C-RFM, a precision operation model

We have just analyzed user portraits from the perspective of the C-RFM model. Next, we can establish a return value model to carry out targeted marketing for different users. The score weight can be adjusted according to the actual situation. For example, at present, the importance of variables is:

Core or not > Historical peak > Loss duration > Investment frequency before return

Value labels are added to lost users based on their scores, as shown in the figure below, and users are ultimately classified into different levels for precision marketing.

Precise C-RFM model

(4) Precision Operation Model Logistic

There are many models that can be used to build models using existing data, such as decision trees. Here we recommend using the Logistic model to predict the user's return probability and perform precision marketing based on different probabilities.

Let's review today's entire reflux analysis process:

(1) How much budget does the company allocate for user re-engagement?

(2) What strategies did we use this money to implement?

(3) How effective are these strategies?

(4) How to improve the strategy later

The entire analysis process has been completed. This is my 8-month return strategy summary. During these 8 months, I have been constantly iterating the strategy, constantly trial and error, and seeking the most ideal strategy. However, no matter how busy I am, I must stop to analyze the past and look forward to the future. We must clearly know the reasons for good and bad strategies, how good ones can be better, how bad ones can be avoided, and actively embrace changes and not be eliminated by the times.

Postscript: I hope this article can help the majority of Internet finance operators and enable the majority of Internet finance users to understand how the platform operates. At the same time, I welcome Internet finance colleagues and enthusiasts to communicate and learn together and put forward your valuable suggestions.

Source: Jiang Di