How to give valuable suggestions when writing data analysis reports

When writing a data analysis report, making valuable suggestions is a key step, but it is also a difficult part for many analysts. This article will provide you with a practical methodology to help you accurately give practical and insightful suggestions based on the understanding and needs of different business departments for data.

What are you most afraid of when writing a data analysis report?

——Suggestion section! Either I don’t know how to write suggestions, or I simply write two sentences like “GMV has dropped, I suggest you increase it!” After writing it, the boss still despises me: “You need to give valuable suggestions! You need to be insightful!”

What should we do? Let’s talk about this today.

1. Key points to solve the problem

The key to solving the problem is to figure out what the business really wants. Not all business departments have mastered data thinking. Some departments are not concerned about data, while others pay close attention to data and may have done a lot of analysis themselves.

If we want to serve the business department well, the first thing we need to consider is: "How much do these people know about data?" Then we can prescribe the right medicine. In contrast, people who know more about data will expect us to give more specific suggestions, so we should discuss this situation first.

2. Soothe the Nervous

Among those who understand data, there is a type of person that I particularly hate: neurotic people. If the market goes up 2% today, they will ask you to "analyze it in depth"; if it goes down 1% tomorrow, they will ask you to "analyze it in depth" as well. The worst thing is that before you finish writing your analysis report on the decline, the indicator has already rebounded, and you have to turn around and write an analysis report on the rise...

In essence, such people neither understand the regular trend of data nor have the habit of looking at the big trend first and then the details. Therefore, they are unable to distinguish regular fluctuations nor recognize real abnormal trends. That is why they get entangled in trivial issues.

At this time, I really want to give them a suggestion of "stop messing around". Of course, we are very polite and our suggestion is: "It is recommended to understand the basic data rules and pay attention to the development trend. We will continue to monitor this indicator." This not only comforts them, but also helps to find the real problem (as shown below).

3. Those who lack guidance

There is a type of people who can look at data, but they only look at KPI data and result data. When they see that the KPI indicators have dropped year-on-year and month-on-month, and the total volume is not up to standard, they are very anxious, but there is nothing they can do. Pay attention! This type of people hate the saying "We need to increase it!" the most. They know that they need to increase it, but they just don't know how to do it.

At this time, the analysis needs to be done in more detail. It is recommended to use benchmark analysis or process diagnosis to help them see the gap between themselves and the benchmark, and see the problems in their execution process, which makes it easier to form ideas.

Finally, suggestions are given in two directions: learning from benchmarks and improving weaknesses: "It is recommended to learn from XX benchmarks and improve XXX practices" and "It is recommended to improve XX problems, and the overall performance is expected to increase by x%" (as shown below)

4. Verify the hypothesized object

There is a type of people who, when looking at data, will do conventional cross-analysis, comparative analysis, and process analysis, and directly form a hypothesis about the problem. At this time, if you talk to him about "learning benchmarks/changing problems", he will say: "I already knew it, can we talk about some key issues?"

At this time, it is best to directly get his analysis assumptions, and then verify and directly answer his questions. This is the most effective advice. The difficulty here is that his analysis assumptions may be too complex and difficult to quantify directly. At this time, you can disassemble the assumptions and analyze them from some quantifiable angles (as shown in the figure below).

The final advice can be directly stated: "The XX issue that the business department is concerned about has been verified by me and is true. The influencing factors do exist and it is recommended that it be dealt with as soon as possible."

There is one thing we should pay attention to here, that is, the assumptions of the business department are all about the external environment, and at this time, they are likely to want to pass the buck. As a data analyst, we naturally cannot help others to clean up the mess without any moral integrity. We can collect some data on the external environment, and as for the suggestions, we can just state the facts.

5. Assist in testing implementation

There is another type of people who have already done the hypothesis verification themselves. They have a clear plan in their mind and just need to try it out. At this time, it is best to directly support them to do the test, and you can even skip the trouble of verifying their plan. They have been thinking hard for so long, and you can easily start a quarrel if you just pick on them. There is no point in arguing about something that has not happened.

At this time, the suggestion can be directly written as "It is recommended to conduct a test, and the effect is recommended." One thing to note here is that before testing, the assumptions must be clearly listed, especially whether there are strong features that affect the results, and whether these features have been considered in the test process. This must be stated in advance. Otherwise, when interpreting the test results, you will be tortured and peeled off. The suggestion at this time can be upgraded to: "Considering that XX factor has a greater impact on the results, it is recommended to do XX treatment when designing the experiment" (as shown below)

6. Dealing with People Who Don’t Understand Data

The above types of people understand data and can discuss data directly. However, the following three types of people do not understand data at all, and even think that data is useless. When facing these people who do not understand data, giving advice can save trouble, and even keeping silent is also a good strategy.

Some business people are not sensitive to data, purely because of their low quality and inability to understand complex reports. The more data you throw at them, the more they will not understand. What you need to do is to provide the main KPI monitoring data, and then remind them when there is a problem with the KPI: "It is recommended to pay attention to the changes in the KPI trend, there may be problems." If they are interested, then discuss in depth.

Some business people are not sensitive to data because their KPI indicators perform well, so they are happy and think that they don’t need to consider anything else. At this time, no matter what you say, they will say: "Oh, I heard it" or "The analysis is not in-depth enough, take a look again." For these people, you don’t need to give a lot of suggestions every time, but grasp the problems hidden under the KPI. If you can point out the problems, not only can you attract their attention, but also make the leaders recognize our capabilities (as shown in the figure below).

Some businesses are insensitive to data because of empiricism. They believe from the bottom of their hearts that "data is useless!", and their experience is the most useful. Even if they encounter poor performance, they will confidently say: "In my experience, just issue coupons", "In my experience, just launch new products", "In my experience, nothing works, it's the environment"

It is very common to be harassed by this kind of person, because he is blindly arrogant, so don't provoke him easily without being prepared. You can be patient temporarily and wait for the right time. When his old experience doesn't work and the boss loses patience with him, stab him hard: "This method has been implemented 3 times, and the ROI is getting worse and worse. It is recommended to change the measures." "Not all people can't do it. There are business lines that can do it. It is recommended for reference."

VII. Summary

After being criticized for "the analysis suggestions being too simple", many students always try hard to find "high-end" suggestions. They especially hope to have a standard suggestion template, or a gold medal suggestion from a big company. As long as they take out the gold medal, the business will bow to them.

In fact, because different people have different abilities, it is common to see "one man's honey is another man's poison" at work. Therefore, it is a better solution to communicate carefully, understand business needs, and provide suggestions according to different situations. I have classified the situations in this article as above for your reference.

Of course, if some students want to see more specific cases of data analysis supporting business, they can check out my Knowledge Planet!