11 commonly used data analysis methods for product operations!

There are many methods of data analysis . Today we will start from the well-known 5W2H and extend to group analysis, basically covering various methods that may be used in work and life. Let’s talk about them one by one.

It is to think about problems using English words starting with 5 Ws and 2 Hs. It is easy to understand and suitable for solving simple problems. However, when facing complex business problems, other methods are needed to assist.

Proposed by Fermi, it is mainly used to turn complex problems into simple problems, gradually spreading out like branches, breaking down problems, and turning a complex problem into simple sub-problems.

Common questions in interviews: such as estimating how many product managers there are in Shenzhen, how many piano tuners there are in Chicago, etc. We call this type of estimation problem the Fermi problem . When solving the Fermi problem, the focus is usually not on really calculating how many product managers there are in Shenzhen. The focus is on your analysis method, that is, your ability to analyze problems using logic trees.

When it is necessary to analyze industry issues and formulate development plans, industry analysis should be conducted, and the PEST analysis method is the first choice.

The multi-dimensional decomposition method is dimension + decomposition , thinking about problems from multiple angles.

So from which dimensions can we break down the problem?

• Disassembly from the indicator composition
• Disassemble from business process

A common question in interviews is " How do I analyze a 5% drop in the next-day user retention rate ?" This type of question can be solved using the dimensional decomposition method.

By breaking down the data in multiple dimensions, we discovered a phenomenon where we get opposite conclusions when examining the data as a whole and its different parts, which is called Simpson's paradox.

When we conduct group research on two variables, the one that has an advantage in both groups becomes the losing side in the overall evaluation.

The most famous example of gender discrimination is the 1973 University of California, Berkeley, where the male admission rate was 44% and the female admission rate was 35%. Based on this data, some people think that the school has a tendency to discriminate on the basis of gender. However, if we look at the admission rate of each department separately, we can find that the admission rate of female students in the four departments of ABDF is higher than that of male students. This paradox tells us that a simple statistical number cannot fully describe the complex meaning behind it. Therefore, it is wrong to only see the data as a whole and ignore the differences between each part of the data.

When conducting a comparative analysis, there are two main issues to consider: who to compare with and how to compare.

Who to compare with

• Compare with yourself : year-on-year, month-on-month, fixed ratio, comparison with target value, vertical comparison, horizontal comparison, comparison in a specific period
• Compared with the industry average

How to compare

• The size of the data as a whole : mean, median
• The overall fluctuation of the data : coefficient of variation
• Trend changes : line chart, year-on-year, month-on-month

Note: Comparability is possible only when the objects of comparison are of the same size.

A/B testing is the application of comparative analysis

Analyzing the causes of problems, also called attribution analysis, "why" questions, problems with declining indicators

A method of studying the relationship between two or more data. If one indicator changes with another indicator, it means they are correlated . If one indicator changes first and then causes the change of another indicator, it means they are causally related.

It should be noted that correlation is not causation, and in real life, 100% causal relationships are difficult to find. How to determine whether it is correlation or causation? The answer is: single variable control method , control other factors unchanged, change only one factor, and then observe the impact of this factor on the result.

It is also called cohort analysis , which means comparing the data after grouping.

For example, analyzing retention rates over time is aimed at finding groups with low retention rates and then further analyzing these groups.

There are also lost user analysis, financial overdue analysis, etc.

RFM analysis is used to classify users by value , from important value users to general retention users, identify valuable users, conduct refined operations, and continuously convert users into important value users.

The RFMs here correspond to:

• R -Last consumption interval
• F -Consumption frequency
• M- Consumption amount

For example, credit card membership service is an example of operation based on RFM analysis. The same operation strategy cannot be adopted for all users, otherwise it may lead to loss.

Note: The RFM value should be used flexibly according to different businesses

The AARRR model is used to analyze user behavior, make decisions for product operations , and achieve user growth .
Corresponding to the 5 important links of product operation:

• Acquisition -How do users find us?
• Activation : What is the user's first experience like?
• Retention- Improving retention : Will users come back?
• Revenue- Increase your income : How to make more money
• Refer- Recommendation : Will the user recommend it to others?

In the user acquisition stage, we are more concerned about the following indicators:

• Channel exposure
• Channel conversion rate
• Number of new users per day
• Daily app downloads
• Customer acquisition cost

During the user activation stage, you need to find the "aha moment", which is the moment when users can't help but like the product highlights and express admiration.

In the retention stage, the core goal is to make users develop usage habits, focusing on the retention rate indicator

In the income increase phase, the main focus is on:

• Total volume related indicators , such as total transaction amount and transaction quantity
• Per capita indicators , such as ARPU/ARPPU and average visit time
• Payment indicators , such as payment rate and repurchase rate

The referral phase, also called viral marketing or self-propagation, focuses on:

• Forwarding rate
• Conversion rate
• K Factor

Funnel analysis is an analytical method to measure the conversion rate of each step of the business process. It has corresponding applications in various industries, such as user conversion analysis, user churn analysis, traffic monitoring, etc. The purpose is to locate the problem node and find out where the problematic link is.

Read a book carefully with Coke and make a little progress every day.

Author: Coke's Journey into Data Analysis

Source: The Way of Data Cola