How should operations build a data analysis system?

A classmate asked: I often hear requirements such as "building an operation analysis system, building a performance monitoring system, building a product analysis system", etc. But what exactly is a data analysis system ? It seems that what we often see are only the five letters AARRR, and the meaning is vague. What exactly does it mean to build a system? Today we will give a systematic answer.

Building a data analysis system is an essential step in the development from primary data analysis to advanced data analysis, so pay attention to it.

1. Common Mistakes in Building a Data Analysis System

1. Listing indicators without focus

Many articles discuss data analysis systems and lay out a large number of indicators. There is no explanation as to which one to look at first and which one to look at later; it takes half a day just to understand the hundreds of indicators, and you don't have to do anything else in your business. You can just look at the numbers here every day.

2. Getting caught up in details and not having a goal

Many students habitually list the indicators and then start to split them up by time, channel, region, and user level. After splitting them up again and again, they mark a lot of increases and decreases. The problem is that there is no specific standard.

I struggle every day: Is a 1% change a problem? What percentage is the problem?

3. Not distinguishing responsibilities and seeking perfection

Many articles have titles like "E-commerce Index System", "Operation Index System" or even "Internet Index System", but in reality, for a company with more than a dozen BUs and dozens of business lines like BATT, do they all look at the same set of indicators? Simply speaking, operations are divided into: users, products, data, new media, communities, activities, commodities, channels... There are dozens of types of operations, and we also look at a set of indicators? These big and comprehensive summaries always seem to make sense, but are actually not very useful.

The ultimate consequence is self-congratulatory data reports. It seems that listing hundreds of indicators and breaking them down into dozens of dimensions makes the daily updates very tiring, but when you look at it, the report opening rate is less than 10%. When operations, product, and sales encounter problems, they still file temporary data collection orders. Every day, they run temporary data collection until their fingers are broken...

2. What is a data analysis system?

As the name implies, the data analysis system includes two points:

• Data analysis: It means not just listing data, but interpreting the data, explaining the business meaning behind the data, and finding points that are useful to the business.
• System: It means that data cannot be laid out without logic, but should be presented in a rhythmic, prioritized, and sequential manner; this will support the business more efficiently instead of being buried in endless SQL codes, and will also allow for better accumulation of analytical experience.

A true data analysis system is one that strings together data reports and special reports, presents them in a hierarchical manner, and applies them to the business.

3. Basic ideas for building a data analysis system

The essence of data analysis is to serve the business. The service purpose is to help business work as much as possible and waste less business time.

Therefore, when building a data analysis system, you should first ask yourself:

• Who am I serving?
• What are the job responsibilities of each of them?
• What kind of data can better help them do their work?
• When can I offer help so that it is less intrusive to them?

This is the basic idea of ​​building a data analysis system.

1. The first step: identify the service object

The company has departmental divisions of labor, so the first step is to identify: which department I am serving, this is very critical! Because even for the same problem, different departments will have different focuses - for the same sales problem, if it is looked at by the sales department, the focus will be on the completion rate, progress, and quality of each sales team; if it is looked at by the supply chain, the focus will be on the total volume, the quantity of each product, and the peak demand period; if it is looked at by risk control, the focus will be on collection, bad debts, and arbitrage. Knowing the department will help you understand the real needs.

Secondly, there are different levels of positions within the department, and it is necessary to make a specific distinction: who needs to see the report, and what are his responsibilities and concerns. In sales, department leaders focus on the deployment of subordinate teams, which areas to focus on, and which products to promote; every salesperson focuses on which customer to follow up, which step to follow up, and what to say to people. Generally speaking, the higher the level of management, the more they focus on strategic issues, while the lower the level of management, the more they focus on execution issues.

Even if there are some things that seem to be able to be done by one person, there is also division of labor and cooperation in the enterprise; for example, publishing articles on a public account seems to be something that can be written by one person, but in the corporate scenario, there is a professional name for it called: new media operations, and there is also a detailed division of labor.

2. Step 2: Clarify work objectives

After clarifying the people, it is necessary to recognize everyone's work goals. Quantifying goals is the soul of data analysis; subsequent evaluation of the quality of work, judging whether business trends are normal or abnormal, and exploring solutions to problems all start with calculating the gap between goals and current conditions; this is very, very important. Many students who work with data get bogged down in details and the reports they make are difficult to understand. This is because they simply don’t know what the correct value is.

Business goals are not all as simple and crude as “a small goal of 100 million yuan”. Under the subdivision, there can be many types, such as the common ones:

• By time of achievement: year, quarter, month
• By the form of assignment: long-term task/temporary task
• By service object: Own/other departments
• By service object: leader/team leader/employee
• By process position: result-oriented goals/process-oriented goals

Continuing with the example of new media operations, a team may have multiple goals at the same time:

Note: There is a logical relationship between different goals; for example, the annual task of increasing fans may be composed of multiple forms such as increasing fans through promotional activities, increasing fans through fission, increasing fans through popular copywriting, and natural growth. One big goal corresponds to multiple small goals; by sorting out the various goals according to size and time sequence, you will have the basic framework of the analysis system.

Later, we can follow this framework to track the completion of goals and diagnose operational effectiveness, which will lead to the next step.

3. Step 3: Track business trends

With clear responsibilities and goals, you can track business trends. When tracking, the first thing to focus on is: the achievement of the goal; the supervision of the goal achievement rate involves a series of subsequent action judgments. When encountering something, first judge the importance and urgency, and then look at the details (as shown in the figure below).

It should be noted that people at different levels have different focuses. Taking new media as an example, the guy who is responsible for the content may be responsible for every article; the guy who is responsible for the delivery is responsible for the effect of each delivery; if a single execution is not good, it is necessary to review and summarize the problems. But as the leader of the operation group, you may pay more attention to the overall KPI achievement. If one article is not good, it can be made up by other articles.

Many data analysis systems based on traditional enterprise scenarios end here. Please note that achieving this step can only be regarded as completing the construction of the "data supervision system"; because just looking at the number of targets and the completion rate is a state of knowing the facts but not the reasons. We cannot answer questions like "Why is it not done well? What should be improved?"; if we want to answer in more detail, we have to go deep into the business process and understand the specific actions. (Traditional companies stop here more because the traditional store and salesperson sales model lacks data records, which does not mean they do not want to go deeper).

4. Step 4: Understand business actions

If you want to improve a business, you must understand it. Most businesses are more complicated than we think. For example, new media operations. Students who don’t do it may take it for granted and think: Isn’t it just writing an article? I just look at the reading and forwarding data... But in fact, if you look closely, an article may have many business details (as shown below):

Understanding business actions and breaking down business details is to "find points where data can help"; data is not omnipotent. For example, when a new media guy writes an article, data cannot just tell him how to write it.

But when it comes to business details, data can provide a lot of reference, as shown in the following figure:

This step is the key to improving the quality of data analysis. By breaking down business actions and finding helpful points in the data, we can further analyze the problem while tracking progress, which pushes us to the next step.

5. Step 5: Review the results of the action

If you have a good understanding of the details of business actions, you can review the results of the actions and summarize the experience; the advantage of data is not to directly produce superhuman creativity, but to summarize general experience afterwards; excellent business capabilities are always scarce resources and cannot be replicated, but through data analysis and review, you can summarize obvious self-defeating behaviors and prevent ordinary people from making mistakes.

Just like writing copy, it is impossible to expect every creator to become a Half-Buddha Immortal, but we can summarize:

• The conversion rate of current affairs topics is 50% lower than that of emotional topics, so they should not be used for conversion;
• On Thursdays and Saturdays, push notifications are 40% lower than other times, so no push notifications will be made.
• If the link jumps more than 3 steps, the conversion rate will drop by 30%. Control the length of the article...

Having analytical conclusions can help operations avoid many pitfalls. Even if I fail occasionally, I know clearly: "There is no other way, I have to publish the article at this time. If I lose a little more readers, I will lose a little more." When doing business, we are never afraid of failure; what we are afraid of is failing without any clear reason. If data can be accumulated over a long period of time, the business side will become more and more experienced and their thinking when encountering problems will become clearer and clearer, and the role of data will be truly brought into play. However, the problem is not static, so the data analysis system must also be continuously iterated and upgraded.

4. Iterative Upgrade of Data Analysis System

Keep this standard in mind: stick to your goals, iterate your methods, and accumulate experience. This is the basic method, bottom line, and highest requirement for building a data analysis system.

Under this principle, the iterative upgrade route of the data analysis system is shown in the figure below:

• After setting the goal, keep track of the goal completion rate by monthly, weekly, and daily reports.
• When problems arise in achieving your goals, first determine the priorities and then look at the details.
• Provide temporary support for key issues, explore causes and solve problems.
• Guide follow-up work based on experience, precipitate effective methods, and guide future goal setting

The operation of such a system is also very easy for the business department: usually they only need to look at the achievement rates of a few core KPIs and don’t have to worry if everything is safe; they can receive timely warnings when the trend turns bad.

If you want ideas, you will have enough materials to use, and the user experience is very pleasant. As for the data analysts themselves, fixed KPIs and business support are made into data products, and case analysis is used as special topics. If you have done more products and topics, it will reflect your personal achievements; it is much better than writing SQL endlessly and not knowing what you are doing with it.

V. Summary

Building a data analysis system is essentially a matter of "coming from the business and going back to the business", which requires everyone to spend more time and effort internally.

However, many novices are too focused on theories, methods, and models, and ignore, disregard, and despise the business; they think that other people's work has no technical content, "it's just publishing an article" or "it's just fooling customers", and only their own algorithms are truly awesome.

When encountering problems, they will not communicate with the business in detail, but will only go to various data analysis WeChat groups to ask: "Is there an XX indicator system, preferably an authoritative, standard, BAT-recognized version"; this is going in the wrong direction, and in the end they will only get one reply: This does not suit our company's situation.

A good data analyst should be like an ophthalmologist. There may be many professional methods and tools for fitting glasses, but in the fitting process, the doctor is not concerned with his own theories, but whether the user can see clearly, and constantly asks the user "Is this okay? Is this clearer? How about trying this way?"

Using professional methods to serve personalized needs is what professionals do. Let’s share this with everyone.

Author: Down-to-earth Teacher Chen

Source: Down-to-earth Academy