The Necessary Data Analysis Methodology for Product Managers
Data is a tool that is absolutely objective and can evaluate the success of product improvement, so you must develop habits to analyze data and master data analysis methodology. In the process of iterative product development, it is driven by data to ensure that the product develops in a better direction.
As the controller of the product iteration direction, every time a product manager makes a decision, he should avoid “I think” such a subjective decision-making method, and instead use data as an argument.
Data Analysis Perspective
Based on User Path
The idea based on the user’s path is to analyze the user’s operation behavior, mainly based on each user’s click behavior log in the App or website, analyze the user’s circulation rules and characteristics of each module, and tap the user’s access or click mode. Then achieve some specific business purposes.
Such as the arrival rate of core modules, mainstream paths and browsing characteristics of specific user groups, and optimization and revision of App product design. By analyzing the user’s path behavior, we can get the typical path, so as to optimize it.
In addition, user path analysis is a good way to define user portrait tags. For example, for a social e-commerce app, we can divide users by user operation data, such as create active users who actively post, interactive users who are keen to like comments, and diving users who silently read posts without feedback.
Based on Product Node
The idea based on the product node is to analyze the conversion rate or data proportion of a key node, for example, for an e-commerce app, the data conversion rate analysis of adding a shopping cart to the order is successful, and the analysis of coupon usage rate.
Data analysis based on key nodes can help to optimize by adding functions, for example, the increase of payment conversion rate can be added to countdown, etc., to prompt the completion of payment as soon as possible.
Steps for Data Analysis
Before doing data analysis, we must be clear about the problem that the data analysis is aimed at. Are you trying to figure out the arrival rate of a certain page? Or do you want to know the overall conversion rate of the user’s behavior path?
For the problem of data analysis, the data indicators are determined and split. For example, the conversion rate of order users is defined as order users / all users, then the indicator of order user conversion rate is split into order users and all users.
The analysis of data indicators is to make products or businesses to develop better. Before doing data analysis, you should think about the purpose of data analysis and determine the scope of our analysis based on the purpose. Only when the scope is determined clearly, the results of data analysis can more accurately guide product improvement and solve practical problems.
The data collection methods are generally the following:
(1) Questionnaire survey: generally used for preliminary user surveys or user subjective experience acquisition, but the accuracy is poor and the sample is small;
(2) Client data: It is generally used to record the user’s browsing path, and the indicators such as the ease of use of the product can be analyzed through indicators such as user behavior and page stay time;
(3) Server-side data and historical logs: The data output by the server is more accurate and in-depth. For some data with higher accuracy requirements, it is recommended to use the server-side logs as the original data;
After the data is collected, some dirty data needs to be processed.
The collection of data is only preliminary work. Which method you choose to analyze data is the key point.
Here are several methods of data analysis:
The analytic hierarchy process, or AHP for short, refers to the decision-making method that decomposes the elements that are always related to decision-making into goals, criteria, plans and other levels, and performs qualitative and quantitative analysis on this basis.
Taking the analysis of user loyalty as an example, loyalty is a skewed indicator, and we need to measure it with a quantitative value. Then we can use AHP to analyze, select four quantifiable values of user usage frequency, recent usage time, average usage duration, and an average number of pages used to measure. The product manager defines the weight of these four values and finally can obtain the loyalty value of each user so that quantitative comparison and analysis can be performed.
However, AHP will be affected by people’s subjective judgments. When different people have different weight distributions, the results may be quite different. This analysis method has certain non-objectivity.
The core idea of DuPont analysis is to decompose the problem layer by layer until it reflects the most fundamental problem.
Taking the e-commerce industry as an example, GMV (Gross Merchandise Volume) is the most intuitive indicator for evaluating performance. When GMV declines YoY or MoM, it is necessary to find factors that affect GMV and dismantle them one by one. If the decrease in GMV is caused by the decrease in the number of users who placed an order, is it because the number of visitors (traffic) has decreased, or has the conversion rate decreased? If the number of visitors is reduced, is it because of a decrease in natural traffic, or is it because of insufficient marketing traffic?
Using the DuPont analysis method will allow us to clearly find the reasons that affect the results, especially for data that will be affected by several factors. It is a very effective analysis method.
Funnel analysis is a set of process-based data analysis, which can scientifically reflect the user’s behavior and the important analysis model of user conversion rate from the beginning to the end of each stage. Now all Internet products and data analysis are inseparable from the funnel analysis. Whether it is a registration conversion funnel or an e-commerce order funnel, there are two points to pay attention to. The first is to pay attention to which step loses the most, and the second is to pay attention to the behavior of the lost users.
Taking the purchase behavior of e-commerce as an example, the user can browse the goods to complete the transaction can be divided into the following five steps:
Through funnel analysis, you can get the following figure:
We can see that the conversion rate before the shopping cart is high, but in the process of placing an order, the conversion rate drops sharply to 35.4%, and this may be the place that needs improvement. After identifying the key node of the problem, we can conduct a detailed analysis of the user behavior of the node, such as the user’s stay time, specific events on the order confirmation page, and further analysis.
Improvement and Tracking
It is the purpose of data analysis to find problems based on data and find the optimal solution; then verify the effect of the program through the effect of the later period and the comparison of the data before and after.
Data analysis is an important part of the product manager’s work. Adjusting products based on data changes is an effective way of working. How to conduct data analysis, and formulate improvement plans based on the data is a skill that every product manager should master.