User operation: How to conduct user behavior analysis?

Many students are most afraid of doing open-ended questions. For example, “You do a user behavior analysis/business analysis/sales analysis…” and then nothing else happens.

When I receive this kind of topic with no beginning or end, my mind often goes blank and I wonder: " What exactly should I analyze? " This is especially true for user behavior analysis, because there are so many user behaviors. After finishing the work, people either complained that the analysis was not thorough or that there was no focus. What was the point of the analysis?

What should we do? Today, let me explain the system

1 Common Mistakes in User Behavior Analysis

▌Error 1: Randomly placing indicators and mixing everything up. The most typical example is to put basic user information such as gender, age, occupation, height and weight at the top. Please note that when analyzing user behavior, it is the behavior, not the basic information. Too many irrelevant indicators will only interfere with your vision and make things even more confusing.

▌Error 2: Listing data without making any judgment. The most typical example is the listing of a large number of data such as the number of user logins, number of clicks, and page jumps. What does this actually indicate? No conclusion. This kind of thing cannot be called "analysis" at all, it is just a basic data display. Since it is an analysis, there must be conclusions, questions and answers.

▌Error 3: Taking words literally and jumping to conclusions. The most common:

● Fewer users are logging in, so we need to increase

● This product is popular among users, so we should sell more

● This content has a lot of users clicking on it, so continue to publish it.

Basically, if the data is low, raise it, and if it is high, maintain it. The conclusion was so stupid that the business department cried...

The above chaos mainly comes from the lack of understanding of the user behavior focus of different departments . If you don't know the key points, you can just piece together the data and ignore how to draw conclusions from the data, which will end up being superfluous. If you want to break the impasse, you have to think seriously: What can the business see from looking at user behavior?

2What is User Behavior?

A user ID and any recordable actions generated in an enterprise's internal system can be called user behavior.

These elements are expressed differently on different platforms (as shown below)

The way of collecting user behavior is different on different system platforms.

In short, this is why the user behavior-related indicator data appear to be numerous, complex, and chaotic: there are many kinds of user behaviors, and if they are not combined with specific business needs, they cannot be explained clearly.

3. Different business needs

There are four situations in which the business side pays attention to user behavior.

▌Situation 1: I know nothing, let’s wait and see.

In this case, it is better to be rough rather than fine, and complete rather than precise . First give an overall overview to allow leaders/business colleagues to get a feel for it, and then when there are specific topics, conduct in-depth analysis (as shown below). Otherwise, if you start with a bunch of trivial things, it is very likely to make people dizzy and wonder "What on earth is all this talking about?"

▌Situation 2: Have something in mind and focus on the results. This situation usually occurs after a specific business process, product function, or content is released. The business side’s goal is clear: to see how this thing performs.

At this point, we can no longer talk about it, but focus on the functional points that the business is concerned about, and display the data from large to small (as shown below)

Notice! There is a big pitfall in user behavior analysis, which is: more user behavior does not mean good performance. For example, in the e-commerce business, the operator enthusiastically launched an activity where users could get discounts by watering and planting trees, in an attempt to increase the number of active users. However, it was discovered that all the users were playing games and waiting for discounts, and the number of people placing orders was decreasing!

At this time, you can use matrix method, before-and-after comparison method, behavioral relationship analysis and other methods to specifically look at the impact of this behavior on performance (as shown in the figure below).

▌Situation three: Performance pressure, overwhelmed. In this case, a specific business process is usually evaluated, and this process is a core process. For example, new user registration, participation in large-scale events, transaction processes, complaints about key issues, etc.

This kind of user behavior analysis with clear goals can be said to be the simplest and easiest. The core idea is the following four modules.

It should be noted here that many students will directly insert conversion process analysis. The data presented in this way is too detailed and can easily blur the overall judgment. Good / bad judgment always comes first . If you even misjudge “good” and “bad”, then all the subsequent analysis of the causes will be wrong. Therefore, the first thing to do is to judge the overall situation and see whether it is acceptable.

Another point is that the analysis of remedial measures is often overlooked by many students. The second biggest pitfall of user behavior analysis is that it is an analysis of “ knowing what it is, but not why it is” . User behavior is the result of various factors. In actual business practice, it is impossible to conduct controlled variable research on every project like in a laboratory. Even if AB testing is done in advance, there will be various differences when it is actually launched due to time and location.

Therefore, when a problem really occurs, it is very likely that the cause cannot be analyzed in a short period of time, or even if the cause is roughly known, there is no way to stop the activity/change the channel. The idea at this point is not to worry about whether the user doesn’t like the copy or the product, but rather what else we can do to save the relationship.

Therefore, remedial measures analysis must not be omitted. This is much more valuable than just shouting out in isolation: "This process doesn't work!" This is also the reason why even though many data clearly provide users with conversion paths, the business side still says: "Not constructive." No one likes the bird of mourning that cries, “It’s over! It’s over!” People want to hear, “ Try this! Try this!”

▌Situation 4: The situation is unclear and people are suspicious. This situation usually occurs when a certain business is not doing well and the business party has no clear assumptions. I was thinking: "Can we dig deeper into user behavior? Find the reasons?" As for what to dig and what reasons to dig out, they may not know themselves...

In short, it is easier to find inspiration for solving problems from extreme situations.

For example, regarding points redemption, the business side just feels that this business is not doing well, but they can’t tell exactly what is wrong. At this point, you can look at the data in two ways as shown below (as shown below)

If you find that high-value users have a clear preference for certain gift redemptions, you can design corresponding gift plans to attract high-value users. If it is found that heavy users are clearly suspected of taking advantage of the system, the reward rules can be modified accordingly. In short, as long as the behavioral differences between user groups are large enough, strategies can be generated.

4 Summary

From the above four situations, we can see that even the same data can be presented in different ways in different situations. This requires students to carefully understand business needs in their work.

Many students would say: Why not just ask the business directly? The problem is that among the four situations, except for situation three which clearly has KPI pressure, the other three situations are very vague, and the final demand expressed verbally is: "Do a user behavior analysis and see."

This requires students who work with data to have a certain level of judgment. The above four situations are progressive, and their logical relationship is shown in the figure below. Students can guide the business like peeling an onion, find the issues that they really care about, and make valuable analysis.

-END-

Author: Down-to-earth Academy

Source: Down-to-earth School (gh_ff21afe83da7)