From 0 to 1, building an Internet operation analysis system

There are indicators but no system

Numbers, no analysis

There are charts, but no conclusions

It is the biggest pain point for students who work with data in actual work. Today, I will take operations as an example and systematically explain how to break the impasse. There are eight branches of operations. Today we will take content operations as an example. Because content operation is the best way to reflect: there are a lot of indicators, but no analysis (as shown below).

1. What is the problem?

When it comes to content operation, many students instinctively think of official accounts, Weibo, and Douyin. So the data indicators came out of his mouth: number of fans, number of new fans, number of regular fans, number of readings, open rate, forwarding rate... He became more and more excited as he spoke, and the words in his throat: "The number of readings today is low, I need to increase it" were almost about to burst out. However, operations colleagues can silence the data with just one sentence: “Try to record a high-quality video or write an high-quality article!” If they add another sentence: “ I already know these routine data, what’s the use!”, the data will most likely be at a loss.

What's the problem?

The problem is (emphasis added):

1. The indicator itself is just a data measurement, not an explanation of the problem

2. The problem itself does not come with a solution, so you need to design a solution

3. The plan itself cannot prove its effectiveness, and needs to be supported by evidence

4. Operations require problem prompts, solution assistance, and effect verification, not just one or a few isolated numbers.

Therefore, to get from a bunch of numbers to useful conclusions, we need to work step by step, using data to observe the current situation, derive solutions, and test the effects. Instead of simply hoping to calculate a super magical number that will win everything.

2. Step 1: Explain the problem

Data itself does not explain the problem, only data + standards can explain the problem . Where do standards come from? Of course, from the perspective of business goals, being able to achieve the goals means doing a good job . So the first step is to ask three questions to understand what the goal is:

1. What is the difference between Internet content operations and traditional enterprises?

2. What are the tasks of Internet content operations?

3. What is the current task for our company?

Among them, questions 1 and 2 are business common sense and you need to do your homework. Question 3 is a conclusion derived based on "current situation of the company + leadership requirements". To put it simply, it can be summarized in one sentence: VS traditional enterprises, Internet content operations involve the process of raising fish, so three major goals are differentiated: dissemination, fan base, and conversion.

Once you understand your mission, you can set specific goals. Please note that the working model of Internet content operation determines that it will not pursue a single goal, nor will it only look at one indicator. When setting goals, there is often an overall reading goal, which is then allocated to each content release, using a main goal + an assessment criterion method (as shown below).

This step is very important. Because in actual work, operations always like to go to extremes:

They overemphasize single indicators and like to make boasts like “gaining 1 million followers for free” or “selling 100 million yuan worth of goods with one article”. It doesn’t matter if other indicators collapse.

Various indicators are mixed together . They will talk about the number of readings, the number of reposts, the conversion rate, etc. at the same time, and then speculatively report whichever indicator is better this time. The excuse is: Although I didn’t achieve XX, my YY indicator performed very well.

This kind of sneaky practice is a huge destruction to data-based operations, scientific management, and data analysis. Because it confuses standards and confuses right and wrong. If even the judgment of “right/wrong” is vague, how can we summarize experience and improve results ? Therefore, if you want to get things done, you must firmly promote the 1 main + 1 deputy evaluation model. Each task focuses on whether the main goal is achieved. Bad is bad. Only by admitting mistakes can we make progress.

3. Step 2: Deduction of the plan

With the first step completed, we can judge the quality of the operation. But knowing good/bad is not enough and cannot guide the details of the work. If you want to guide the details of the work, you must first understand what the operations are doing, which involves sorting out the work flow. Many students think that operations work is very simple, but upon closer inspection, there is a lot to it (as shown in the picture below).

Understanding the workflow is the key to avoiding the problem of "wanting to do it too high" . When you find that there are so many factors to consider when operating an article, you will never dare to say "I want to do it well" lightly again. There are so many details to consider, and one wrong move could lead to a complete loss.

But another problem arises: the writing of the article itself is too creative, and many of the hot topics are only effective for hype at the time and become invalid after expiration. In such a complex environment, how can we use data to assist? Let’s make one thing clear first: data itself represents rational, objective, and logical thinking, but content creation is likely to be a product of sentimentality, subjectivity, and emotion. Therefore, data does not replace creation, but provides opportunities for creation and helps creation avoid risks.

To achieve this, you need to do three things:

1. Label the content and extract quantifiable labels

2. Based on labels, test results and accumulate experience

3. Collect external data based on tags to indicate opportunities

To give a simple example, on August 5, 2020, an operations editor was writing an article and found that "Nothing But Thirty" was very popular at the time, so he wanted to take advantage of the popularity and spread it. If data is used to assist, it can be done from the following three angles (as shown in the figure below).

This can greatly improve the efficiency of the operations editor. And to be honest, the creative ability of most operations editors is not strong enough to be imaginative, and they are more likely to copy others. So if content tags are really established, many editors will probably just add or subtract tags:

1. For communication-related topics, write about your personal reading experience and make up a story!

2. For increasing followers, distributing materials is effective, and the PDF package is ready!

3. Conversion type, good at triggering gender conflict, start to criticize straight male chauvinism!

Strictly speaking, we do not recommend mindlessly copying, as this devalues ​​the value of operations work. It would be better to just let data analysts write the article. But I can’t resist its fragrance! At this time, it is necessary to establish a continuous monitoring system for operational results, and when a certain routine fails, promptly remind operations to change tactics (as shown in the figure below).

4. Step 3: Verify the effect

After designing the content, you can observe the effect of the delivery. This is something many students do, so I won’t go into details. What’s interesting is that, looking back at the beginning, the following data that the students mentioned casually: number of fans, number of new fans, number of regular fans, number of readings, open rate, forwarding rate...are actually all generated in this step. These are all outcome indicators. Only result indicators cannot be used for in-depth analysis. As far as content operation is concerned, we must at least have clear classification goals and a content labeling system to determine the effect and assist the plan.

Some students would say: There’s no need to go through so much trouble, I can just ask the business directly. Asking is a good communication habit, but the prerequisite is that we ourselves have clear business common sense and judgment . Otherwise, if you just ask directly and naively, what if the operator himself is confused? What if the operation is very speculative? What if the operations meeting used tricks at the beginning to fool everyone? What if the operation puts the entire blame on the data: “We don’t have artificial intelligence and big data methods”, so the operation capabilities are not good enough? If you have basic knowledge, you won't be fooled. It's the same everywhere.

5. Summary

The process of establishing a content operation analysis system:

1. Understand work goals and processes

2. Establish results observation indicators

3. Establish evaluation criteria

4. Set up content tags

5. Evaluate content dissemination/increase in followers/conversion effects

6. Accumulate questions and effective labels

7. Continuous iteration to improve analysis accuracy

Many students would say: Our company’s content operations write their own data reports, so I don’t need to participate so much. Yes, this is the main reason why data analysis cannot be implemented at present.

You know, as a user, the first hurdle when we deal with a business is content. If the topic of communication is not good and the content is not attractive, there will be no subsequent conversion. Data analysts have their hands dirty, don’t understand content, are not familiar with products, don’t study jump processes, don’t pay attention to user feedback, and sit in front of the screen while enjoying the air conditioning and thinking hard: system, closed loop, link. In the end, all I can say is: Let’s make it higher.

Author: Down-to-earth Teacher Chen

Source: Down-to-earth Academy