Data operation: How does big data make users more willing to pay?

When a product reaches a certain stage of development, commercialization is an inevitable issue. It is a bit rogue to talk about feelings with investors without making money.

However, the topic of commercialization is too broad, so today we will first discuss how to use big data capabilities to encourage customer payment . The following are some thoughts on the project.

1. Introduction

1. Background: After acquiring customers, we encountered challenges in converting them into paying users. Here are some of the attempts and thoughts we made under these circumstances.

2. Applicable to: If you are also thinking about how to use some product strategies to convert users, you can discuss this topic together, which is more about e-commerce products;

3. Content framework: divided into three parts: preliminary exploration, problem exposure, and solution ideas.

2. Early Exploration

In order to convert users, we made some attempts with our algorithm colleagues in the early stage. The main steps were: find the target group - design activities - reach users to promote their conversion. The following restores part of the practical process.

1. Find target audience

Based on business experience, collect a list of highly correlated customers with the highest likelihood of conversion from the database and create a whitelist customer tag. For example: based on the basic social attributes of customers, such as age, gender, region, etc.; based on the transaction characteristics of customers, such as whether they hold a card, whether they have points, etc.

1. Design Activities

Select the hot items that these customers may like, design corresponding special events or find seasonal events, and package the content

1. Find customer touchpoints

Reach out to customers through channels such as app push, SMS, and WeChat suffixes

1. Analyze the data

Collect a list of reachable groups, track the changes in their transaction behaviors over the past three days, and analyze the conversion funnel

3. Exposure Problems

At first glance, the overall idea above seems to have no problem, but in practice, there are actually many problems, which are mainly reflected in the following three points:

1. The crowd is not accurate

① Fear of brain-based analysis: The single-dimensional crowd found through business experience cannot effectively judge the user's transaction intention.

② The population is divided in a scattered manner, with high overlap, and the action groups cannot be distinguished, which actually means that it is impossible to effectively manage and operate

1. Lack of a sustainable and effective conversion engine

The problem lies in two main aspects: unsustainability and difficulty in measuring effectiveness

(1) Sustainability

① Collecting data to find the target population is too slow: Finding the target population by guessing, every time you screen users, you need big data students to collect data once. On the one hand, the crowd list is one-time and has low reusability. On the other hand, the algorithm students’ research model has not been fully utilized.

② Long activity design cycle: Finding crowds, planning activities, designing pages, and launching activities. This cycle is too long, with low timeliness and difficult to meet

③ Manual outreach is time-consuming: One-time event outreach is very challenging for operators to conduct each event, and they simply cannot spare the time.

(2) Unable to measure effectiveness

① The activities of each contact are scattered, similar to guerrilla warfare, and it is impossible to continuously observe whether the strategy is effective.

② Lack of experimental thinking: Failure to form overall conversion monitoring indicators, and no control group was set aside for the reach activities.

1. Difficulty in data analysis

Each activity analysis requires data collection, and the timeliness of the analysis cannot be guaranteed; data at each level belongs to different systems, and the underlying data tables cannot be linked.

4. Solution

The series of attempts and problems exposed previously provide a good foundation for us to determine our next ideas. Let’s get back to the topic of this article. How do we use the power of big data to make it easier for users to become our paying customers?

The core idea here is: spend less time finding the most easily converted customers and get the most suitable offers to touch the customers.

What is an offer? Let me explain here that the scope of our definition includes: functional product advantages (that is, the virtual value of the product) and profit-sharing hot-selling activities (discounts, new customer gift packages, hot-selling products).

So what is the specific operational idea? Let’s see it broken down step by step.

1. Find your strengths

Many times, a product (platform)’s own advantages are packaged by itself, but is this really the case based on the feedback from the market/users? Why did new customers choose your platform?

To understand this problem, we did two things:

① Pull the order data of the first conversion of new users in the past year: Analyze the decision logic behind user conversion from the data. The result is quite obvious. X% of users are for A, X% of users are for B... I will not elaborate on the specific data.

② Survey users: What is their perception of the platform and their motivation for making their first purchase on the platform?

Based on massive historical data and user surveys, we can find out what aspects of our platform are attractive to users.

1. Population stratification

The user conversion project is a teamwork that includes: algorithms, business, operations, and products. As mentioned earlier, if users cannot be divided, the team will be chaotic and lack goals. This way we won’t be able to manage our users well. Here we adopt a four-quadrant stratification, user perception of the platform and user preference

High propensity: People who have a high propensity for platform products. This can be screened through the propensity model.

High cognition: cognition is divided according to the closeness of the connection between the user and the platform. For example, product purchase breakpoint customers are people who know more about the platform. For new users who have just registered, their awareness is very low and they need to receive cognitive education first.

Crowd segmentation is only the first step. The second step is to find out the user volume of each group. You can first divide it based on user cognition, for example: breakpoint customers, app active customers, app registered users, WeChat followers, etc., and then set conversion goals for each layer, and each action team will claim the task.

1. Algorithm model support

The business requirements for the algorithm are actually: set a business goal, for example: users who are most likely to buy Class A products (Note: Class A products here must be set in combination with the advantages of their own products), and then algorithm colleagues will model according to the business goal, including: problem modeling (indicator evaluation, sample selection, cross-validation), feature engineering, model selection, model fusion, and finally model verification.

Of course, the above is a complete modeling process. In practical applications, we always start with small steps. First, use a relatively simple IV value estimate to find out the strongly correlated customer dimensions, and give them to the business side for trial.

1. Automated operations based on deep-dive scenarios

This part, I think, is the most critical one. The process of user conversion must be: in certain triggering scenarios, through habitual behavioral routes, to obtain the desired content. Here it will be divided into three parts: high-frequency scenario mining, automated strategy support, and strategy execution.

(1) Find high-frequency scenarios based on population stratification

① Understand user behavior preferences. Online behavior: active users regularly browse a channel, a function, or search for certain content on the app; offline behavior: based on lbs, understand the cities where users appear

② Understand the transaction preferences of this group of users: After knowing their behavioral preferences, the next step is to understand the users’ transaction preferences, what they like, and what they like.

(2) Automated business strategy

Manually set push is time-consuming and labor-intensive and cannot support customer conversion needs at all. Therefore, you can set up some automation strategies based on the key scenarios and routes of new customer conversion, such as the following:

① Customer breakpoint automated recovery strategy: Customers who have browsed/commented/followed/unused coupons, etc., can be automatically collected and targeted for recall. For example, if a customer has browsed a certain product in the last three days, you can give the customer some benefit points and then reach out to them.

② Full-link ambush strategy for key scenarios: Ambush users with conversion content before, during, and after they complete a task. Of course, prioritize reaching high-propensity customer groups and avoid excessively disturbing them.

③ In addition, there are: customer offline setting T+x day contact strategy, customer silent setting profit-sharing strategy, etc.

(3) Strategy Execution

The comparison dimensions of the experiment setting can be: channel, customer group, and offer. One point to emphasize here is that it is necessary to clearly define the goals of the entire strategy and set aside a control group. Otherwise, subsequent data will be difficult to be convincing and it will be impossible to judge whether the strategy should continue to be solidified.

5. Data Analysis

In this part, the focus is to connect the data of each node and realize visual analysis and comparison based on each experiment. This depends on whether the platform's own data management is standardized. If data collection is not standardized, analysis will be disadvantageous in many cases. I won’t go into details here, best wishes~

Throughout the entire process, big data assisted the business in identifying its own advantages, building models to find high-propensity customer groups, and provided greater support in experimental data analysis, making business operations more refined. When it comes to user conversion, it is also more rigorous and methodical.

The above are just some staged thoughts, I hope they will be helpful to you~

Attached is the full text thinking map:

Source: Nobita carries his bag (dxbqxngo)