Many people are using recall campaigns to retain users, but few are able to do so efficiently. The author emphasizes that efficient recall requires not only good preparation work in the early stages, but also data follow-up on the recall results in the later stages to provide optimization suggestions for the next event. Let us explain how we do it.

In the second half of the mobile Internet, the demographic dividend gradually disappears and the cost of acquiring customers becomes increasingly high. Reducing costs through refined operations, cultivating existing users, and recovering lost users is a direction worthy of in-depth research and practical application.

After all, “recalling lost old users” is cheaper than “paying to promote new users”.

The recall of old users can be treated as a separate project, and the best operation plan can be achieved through continuous trial and error optimization.

The first step is user stratification, and the segmentation must be based on user usage scenarios; the next step is refined operations for segmented users, including recalling some of the lost users.

Only by running through the entire process can the entire set of work be of practical significance.

For example: important valuable customers and potential customers need to be targeted with corresponding activity coupons; new customers need to be guided to develop user habits; important retained customers and lost customers need to be recalled, etc.

Today’s article mainly focuses on a previous recall activity case, talks about the recall method, and does a review by the way.

1. RFM Model

In October, in order to prepare for a platform event in early November, the department decided to carry out a large-scale recall campaign.

Platform activities and recall activities are actually a pair of good brothers. The more users participating in the activity, the better. Recalling users just increases the activity. Recall activities require a good reason, and platform activities just serve as a follow-up to the recall.

The first is user segmentation, and we use the RFM model to do this (due to business reasons, we do not need to segment too precisely, so the RFM model is sufficient).

The RFM model is one of many customer relationship management (CRM) analysis methods that can easily, quickly and effectively quantify user value and profitability.

The RFM model has three elements: Recency (the time since the last transaction), Frequency (transaction frequency), and Monetary (transaction amount).

R value (time since the last transaction) is the time since the user's last transaction. The shorter the interval, the larger the value. This type of customer is also the group most likely to respond to the campaign. F value (transaction frequency) is the number of times a user purchases within a limited period of time. The customers who buy most frequently are also the customers with the highest satisfaction and loyalty. M value (transaction amount) refers to the user's transaction amount, which can be divided into cumulative transaction amount and average transaction amount. Different data sources are used for modeling and analysis according to different purposes.

According to the indicators of the three values, the following 8 categories can be obtained:

"↑" indicates that the value is greater than the mean, and "↓" indicates that the value is less than the mean.

Calculate the average of R value, F value, and M value respectively, and then fill in the corresponding customer type based on the user's data.

2. Secondary Segmentation of Users

Before user segmentation, user "data cleaning" must be performed first.

Among the lost users, there are a lot of "zombie users". The characteristics of this type of users are: no matter how big the coupons are given through what channels, they will not respond at all, and they are users that can never be recalled.

Therefore, in order to reduce recall costs and improve recall efficiency, user data must be cleaned first.

Due to the limited length of this article, I will not go into detail on how to clean the data.

The main thing is to understand the company's business, specify screening indicators based on the company's business situation, and conduct data screening and cleaning.

For example: For this activity, we took the lead in excluding user data whose online time was usually low. After getting the cleaned user data, I felt refreshed, as if I saw the high recall rate waving at us.

User segmentation/user portrait is always based on a full understanding of the business. If you don’t understand business operations, you will never be able to correctly segment your users.

Based on a full understanding of the business, we segmented the recalled users according to their online time and recharge amount (the company's business is live streaming).

There will be no recall for users of categories C, E, and F; different copywriting will be used for recall of users of categories A, B, and D.

The more detailed the user classification is, the more targeted it is and the higher the recall rate is. Of course, the workload will be greater.

3. Recall copy

When planning the recall copy, it is necessary to conduct an in-depth analysis of the characteristics of lost users based on the characteristics of the product itself.

Deeper thinking from the user's perspective can arouse the user's resonance, curiosity, etc.; being able to directly inform the user of the interest relationship, are all good directions.

Generally speaking, product recall copywriting can be written from four aspects:

Give discounts and recommend more convenient services. For example: recharge discounts; the live broadcast platform has developed new functions; microphone connection function or anchor circle, etc.; free discounts and recommendations for more attractive content. For example: recharge discounts; a well-known game anchor has recently joined the platform; or a number of beautiful mixed-race anchors with high looks have been added; guiding users to understand product upgrades. For example: platform upgrade; added outdoor live broadcast section; show live broadcast section, etc.; guiding users to understand the product value. This article does not apply to this product.

Based on the analysis of the product business direction and lost users, we came up with three marketing copy and conducted a small-scale AB test.

After a small-scale test, Plan B was determined.

AB testing is a widely used working method, which will not be elaborated in detail here. It is very important to speak with data rather than making decisions based on your own guesswork in operations.

4. Recall Tool Selection

1. Email EDM

Advantages: low cost and large sending volume.

Disadvantages: Low open rate.

Except for specific products, email EDM is rarely used nowadays. The main reasons are low open rate and recall rate.

1. Short message

Advantages: high arrival rate and large sending volume.

Disadvantages: high cost.

SMS is currently the most mainstream recall method because it can reach users directly. The recall rate is closely related to the copywriting and product type.

Although the cost is relatively high, the status of SMS will remain unshakable until new channels emerge.

3. PUSH

Advantages: low cost, large sending volume, and high open rate.

Disadvantages: low reach rate, users can set rejection, which may easily lead to APP uninstallation.

Client PUSH is the method with the lowest cost and highest open rate. If the user has not uninstalled the app and allows push notifications, the app can reach the user directly. However, if the user has uninstalled the app or does not allow push notifications, the app will be ineffective.

The pros and cons of PUSH are very obvious. If the PUSH is too frequent or the PUSH content fails to attract users, it will have a counterproductive effect and even cause uninstallation.

Therefore, most companies mainly use SMS and PUSH as a supplement.

V. Analysis of recall effect

Regarding the recall data of 150,000 lost users, it is actually not rigorous to put SMS and PUSH together here, but because there is no way to verify through which channel the recalled users were obtained, this is a small mistake in this activity.

The correct approach should be:

Push SMS and PUSH separately, and then analyze the recalled data to get the recall efficiency of SMS and PUSH.

All activity data must be able to provide optimization suggestions for the next activity.

The activity has not ended yet at this point. Because no conclusions should be drawn before completing the value tracking of converted users.

Although the recall activity has a relatively low cost, it is still a subsidy of real money.

If there is no subsequent conversion and the ROI is not high enough, then the activity is a failure. No matter how high the conversion rate is, this campaign is a failure.

(Because it involves company confidentiality, all data are modified proportionally and do not represent the actual situation.)

Further tracking of the recall form resulted in the following picture:

The different consumption amounts and conversion rates of the three types of users, ABC, also conform to the initial user segmentation; as for the tracking time, it is determined according to the frequency of product use and user cycle, and each product is different.

At this point, we come back to this sentence - operations must have an in-depth understanding of the business.

6. Cost data

The cost of this activity is mainly in the following two aspects:

The cost of SMS is about 0.03 yuan per message; the subsidy cost is that all recharges require subsidies (ranging from 30%-50%).

Based on the recharge amount (GMV) tracked for 30 days, minus the activity cost, we can get the profit of this activity.

This can be further broken down: ROI is calculated for different users A, B, and C separately, and in future similar activities, individual recall or overall recall can be chosen.

Even if the recall copy is very attractive, users may not come back for various reasons. There is a rule in advertising called the "eight exposure strategy", which means that a brand can gradually enter the hearts of users only after multiple exposures, and multiple displays can increase brand awareness.

Therefore, for users who do not return after one recall, you can continue to recall them in the next campaign, reach out to them more times, and strive to influence them; you can build a product recall system, optimize based on data, and continuously carry out the work of user recall.

Source: User Operation Play (happy_yiyi517)