To build a data indicator system, it is most effective to master these three processes!

This article explains in detail how to build a data indicator system from the perspectives of business flow, management flow, and data flow, providing us with practical information for actual work. Next, let's learn how to use the data system to guide business implementation!

When it comes to data indicator systems, many people like to recite AARRR, RFM and the like. However, when it comes to work, they often find it difficult to meet business needs. For example, a few days ago, a classmate asked: How to build an indicator system for user churn?

The reason is: a company defined the user churn rate indicator as "no consumption for three consecutive months", but the business was confused when they saw this indicator:

• Now that we know that the user churn rate is 30%, what can we do?
• We know that we need to recall lost users, but is it cost-effective and worth doing?
• Why do we have to wait until users churn before we start working? Can’t we start working sooner?

Therefore, we want the data department to help us build a user churn indicator system to comprehensively reflect the problem and assist business decision-making. So what should we do? To build an indicator system that can be used by the business, we need to consider three processes: business flow, management flow, and data flow. Let's take a look at them one by one.

1. Sorting out business flows

Sorting out business flows means figuring out how many steps are needed to achieve business goals. Some business processes are very clear, such as the sales process, which is a large conversion funnel; for example, the customer service process, classifies and handles problems according to customer needs. Business processes are the basis of data indicators (as shown below):

In the churn analysis scenario in this article, the first challenge is that the business process is not clear . The business side itself does not know what to do, and certainly does not know which indicators to look at.

For user churn, common measures include:

1. Preventive measures: When users complain or return products, appease them promptly

2. Preventive measures: When users spend less and less every month, provide timely incentives

3. Pre-emptive prevention: When a user has not consumed for 1 month or 2 months (at this time, the user has not yet reached the churn standard), stimulate the user

4. Post-event remediation: Use promotional activities/new product launches/hot-selling products to recall and try to reactivate

You can make a list first, and then let the business choose: which direction do you want to start from. Attention! The business side is likely to say: "We don't understand the situation, so we can't choose for now." Then the next step is to first get the data corresponding to these situations, so that the business side can see the overall situation before taking action:

2. Sorting out management flows

Sorting out management flows means figuring out what the management wants the business to achieve. Note! Even if the business process is clear, the business goals may be diversified.

For example, in the sales process, there may be several assessment methods:

1. Only evaluate sales

2. Sales + Gross Profit

3. Sales + Gross Profit + Payment Collection

4. Sales + Specific product sales

Different assessment methods determine different main indicators of the indicator system, which will of course affect which sub-indicators are examined. Therefore, it is important to understand the intentions of management.

In the churn analysis scenario in this article, the second challenge is that the business side is not clear about what to do . That’s why they ask questions like “Is it worth doing?” and “Should it be done?”

At this point, you can do some auxiliary work:

1. How much contribution did these users make before they churned → In theory, the greater the contribution before churn, the more they should be recalled

2. How long these users are in the churn state → In theory, the shorter the churn state, the easier it is to recall them

3. Do these users show signs of natural return? → In theory, if there are signs of return, it is easier to recall them.

You can sort out the data first and let the business side take a look at it to get a feel for it, so that it will be easier to set goals.

It is important to help the business side set goals. Because goals of different difficulty mean different means to be adopted, and soft business means will not work. For example, if you want to recall lost users, if you just send a text message without giving any benefits, then there is no hope that users will be willing to come back to consume.

At this time, the business asks you to "analyze in depth why the recall is useless". What is there to analyze? As a user, you are happy to give money to the company after seeing a text message. You would not do that either.

If you really can't set a goal in the early stage, you can set aside some time for observation. For example, set aside three months and a budget to test the effects of different recall methods to understand what effects can be achieved with a certain budget support. With the support of test data, it will be easier to implement the goals later.

3. Sorting out data flow

Sorting out the data flow means clarifying whether the business objectives have been quantified and recorded, and whether the business operation process has been quantified and recorded. This step is to ensure that the content sorted out in the first two steps can be implemented into data tables and reports, rather than hanging in the air. What is tested is the effort of data collection.

For example, in the sales process, the biggest problem of traditional enterprises is their lack of digital capabilities. Except for the last step of signing the contract, there is no data on the previous process. This is a complete failure. There is no point in building an indicator system or guiding business work.

Back to the example of user churn, in addition to the lone indicator of the number of churned users, it is best to record the content of the above two steps (as shown below):

Note! The data indicator system is not only about indicators, but also very important labels:

1. User consumption before churn: high consumption, medium consumption, low consumption

2. User preferences before churn: category preference/discount preference

3. User status before churn: with complaints, without complaints

4. Response to recall: Yes, No

5. Recall user materials: discounts, new products, and seasonal changes

6. Recall user discount intensity: high, medium, low

These need to be labeled to facilitate subsequent in-depth analysis. When you need to explain questions such as "Why is the recall effect not good?" and "Why is the churn increasing?", you can directly use label comparison to get preliminary answers, thereby greatly improving the usability of the data indicator system.

4. Ideal effect

Ideally, a good data indicator system can guide the business to achieve a closed loop of work. In layman's terms, it means: what I want to do → who I want to target → how I want to do it → whether I have done it or not. The whole process is monitored by data. In this way, under the guidance of data, the business side continues to find specific user churn cases. The entire indicator system operates as shown in the figure below:

In this way, it is not just an isolated indicator of "churn rate of 35%", but you can follow the clues to find solutions to the problem and track the effects. It can be said that you can achieve data-driven business.

Author: Down-to-earth Teacher Chen

WeChat public account: Down-to-earth Teacher Chen (ID: gh_abf29df6ada8)