Operators, have you built your data analysis framework?

Data analysis, as a core skill that operations personnel must possess, plays a particularly important role in career development. This article will share several basic data analysis frameworks and methods based on business scenarios to conduct data analysis in a systematic manner.

Event: The process of efficiently tracking user behavior or business through tracking points. Registration, startup, login, click, etc. are all common events. For example, the following figure shows the registration events on a certain day.

• How many clicks did the promotion page receive during a certain period of time, and how much did it increase compared to yesterday?
• What is the cumulative number of registrations for a certain channel ? What are the top ten registration channels in the first quarter?
• What is the intraday trend of UV for a certain activity page and the proportion of Android and iOS?

2. Funnel analysis

Funnel model : Analyze the conversion and loss of each step in a multi-step process. Taking Internet finance - financial management as an example, new users will go through the following steps when making their first investment :

• Successful registration
• Real-name authentication
• Recharge successful
• Investment success

We can use the funnel to analyze the overall conversion situation, as well as the conversion volume, loss volume, and conversion/loss rate of each step.

By clarifying the three basic concepts in the funnel model, you can conduct in-depth analysis with the help of powerful filtering and grouping functions.

• Step: refers to user behavior, consisting of events plus filter conditions
• Time range: The time range in which the first step of the funnel occurs
• Conversion cycle: The time limit for users to complete the funnel. The funnel only counts the user's conversion from the first step to the last step within this time range.

The difference between funnel analysis and event analysis here is that funnel analysis is based on users, or people, to count the behaviors of a group of users. It will not be affected by the events of users who have browsed the page historically, and can more accurately expose the problems that exist in the product in a certain period of time.

Discover problems in a timely manner through the funnel model: We established a registration conversion funnel to measure the conversion rate of each step and the overall registration conversion rate, and monitor the trends of each step and the overall conversion rate through the time dimension.

For example: On April 12, we found that the conversion rate of entering the graphic verification code was obviously abnormal, so we urgently notified our technical colleagues to investigate and found that the graphic verification code function was invalid, causing a large number of users to be unable to display it. After the emergency fix, the conversion rate returned to its previous level. Therefore, by monitoring and analyzing the conversion rate of each step of the funnel, problems can be discovered in a timely manner and losses can be stopped in time.

(Shumei Note: We have launched a more powerful funnel function, which can also directly drill down into groups of users who have not successfully converted in the funnel and push targeted recalls!)

3. Retention Analysis

Retained users: If a user performs the target behavior some time after performing the initial behavior, the user is identified as a retained user.

Retention behavior: After a target user completes the initial behavior, if he completes a specific retention behavior on a subsequent date, the number of retained users will be +1.

Retention rate : refers to the ratio of “users with retained behavior” to “users with initial behavior”. Common indicators include next-day retention rate, seven-day retention rate, next-month retention rate, etc.

Retention table: The retention table gives the retention details of the target users, mainly including the following information:

• Target users: The target number of users who complete the initial behavior every day is the base number of retained users;
• Retained users: the number of retained users and the retention rate

Taking e-commerce as an example, we observe whether operational strategy optimization/product revision will affect user purchasing behavior. At this point we can divide user behavior into:

• Initial Action: Registration
• Retention behavior: paid order

Then, we grouped customers by week based on their registration time to obtain the same cohort, made a retention curve, and observed the 30-day retention of users in this group after they made a purchase. By comparing different cohorts, you can see whether the new user purchase rate indicator is improving.

Taking e-commerce business as an example, the following figure builds the entire context of user life cycle operations and the key indicators that need to be paid attention to at each node based on the AARRR model:

In the user acquisition stage, we hope to attract more potential users to our products and expose our promotional pages through the following basic channels:

• Paid acquisition: media advertising, SMS, EDM, traffic trading/exchange
• Search Marketing : Search Engine Optimization ( SEO ), Search Engine Marketing ( SEM )
• Word-of-mouth communication: user-to-user invitation activities, viral H5 communication, etc.

After users visit the page, they can learn about our products through navigation, active search, and algorithm recommendations. Users who meet current needs will register, which is the first real meeting with the users.

At this time, we should focus on important indicators such as promotion page UV, click-through rate , registration volume, registration rate , and customer acquisition cost.

2. Activation

Do users learn more about our products after registration? This involves product functionality, design, copywriting , incentives, credibility, etc. We need to continuously optimize and guide users to take the next step, so that new users can become long-term active users:

We can improve conversion rates in the user activation process through interface/copy optimization, new user guidance, preferential incentives, and other means. Monitor the funnel conversion of browsing product pages, adding to shopping carts, submitting orders, and completing orders.

In this process, we should focus on activity. If we define adding to the shopping cart as an active user, then we need to observe the conversion rate of the registration to adding to the shopping cart funnel, split it by dimension, analyze the common characteristics/operation strategies of high-quality conversion funnels, improve strategy coverage, and optimize the overall conversion effect.

3. Retention

Once users complete the initial purchase process, will they continue to use the product? Can lost users continue to come back and use our product?

Lack of product stickiness will lead to rapid loss of users. We can build a life cycle node marketing plan to remind users to continue using our products through push , SMS, subscription accounts , emails, customer service follow-ups and other appropriate methods. And on this basis, through the points/level system, we can cultivate user loyalty and improve user stickiness.

Focus on indicators such as retention rate, repurchase rate, average number of purchases per person, and recall rate.

4. Revenue

How much does it cost us on average to acquire each user? How much value can each user contribute to us on average? Can we make money from user behavior or even in other ways?

The foundation of e-commerce business should focus on customer acquisition cost CAC and customer lifetime value. On this basis, users should be encouraged to make purchases through operational activities, thereby increasing user unit price, frequency, and ultimately increasing the user's life cycle contribution value.

Focus on indicators such as customer acquisition cost, customer lifetime value, and marketing activity ROI.

5. Referral Viral Spread

Will users promote our products spontaneously? Can we get more loyal users to promote our products through incentives?

In today's highly developed social network , we can promote our products through various novel ways: user invitation activities, vertical community operations , H5 marketing communications, allowing old users to promote our products and attract more potential users.

Focus on the number of invitation initiators, the number of new users in each viral transmission cycle, the invitation conversion rate, the transmission coefficient, etc.

At any stage in any type of product, there is a single number that needs to be found and placed above all else. When analyzing data, you can capture a lot of data, but you must focus on the most critical things. It is also a key trait in "growth hacking": focus on the goal.

1. Sticky Growth Engine

The sticky growth engine uses retention as the OMTM to drive growth

A typical example is gaming apps. Facebook has proposed the “40-20-10” rule for gaming, which means that if you want your game’s DAU to exceed 1 million, the next-day retention rate of new users should be greater than 40%, the 7-day retention rate should be greater than 20%, and the 30-day retention rate should be greater than 10%.

The next-day retention is far different when not using any operational incentives and when using retention incentives.

For example, common game play methods such as check-in activities, login rewards, and time rewards are all based on the purpose of improving user retention .

A good retention rate is different for different products. We will not expand on the classification of user retention rates here. The correct approach is to find the most suitable retention indicator for different types of products and user stickiness.

2. Paid growth engine

The paid growth engine uses Revenue as the OMTM to drive growth.

Simply put, revenue growth can be driven continuously as long as the value contributed by customers to the product is greater than the cost of acquiring paying customers.

• CAC (Customer Aqusition Cost)
• CLV (Customer Lifetime Value)

For example: In a certain month, the cost of marketing investment is 20,000 yuan, and 100 new investment users are added. Then the cost of acquiring each investment user is 200 yuan. If the average investment per person is 50,000 yuan and the profit margin is 2%, the customer lifetime value (CLV) is 1,000 yuan per person.

When CLV>CAC, ignoring other costs, the engine has been driven to run normally. The next step is to think about how to provide more exposure, expand the top funnel, and shorten the customer's break-even time as much as possible.

3. Explosive growth engine

Explosive growth engine uses referral communication as OMTM to drive growth

Typical case: Sharing based on social scenarios, constantly reaching potential users through sharing red envelopes, bargaining, group buying, flash sales and other gameplay.

In the explosive growth engine, we need to focus on the viral coefficient K = I x Conv:

• I: Invitation, which is the number of invitations sent by each user, reflecting the distribution density.
• Conv: Conversion rate, that is, the probability of each invitation being successful, reflects the infection intensity.

So how to increase the viral transmission coefficient? Here are three common solutions in the above activities:

• Focus on improving the acceptance rate: lower the acceptance threshold, and try to control the acceptance steps within the social scene to avoid two jumps that reduce conversions.
• Shorten the lifecycle of a single invitation process: By using a time-limited approach, you can speed up the growth process while increasing the sense of urgency.
• Try to persuade users to invite more people: The first few invited users can bargain down a lot of money, which will encourage more forwarding after users have tasted the sweetness.

Whenever a new business problem arises, systematic thinking through a framework plays an especially important role in solving the problem.

Data analysis is the basic skill of Internet products and operations. The author is still in the stage of continuous improvement in the direction of data analysis. The above are some cases and experience sharing from recent study and work. I hope it can bring some ideas to newcomers in learning!