How to achieve the “closed loop of data analysis” required by major companies?

We always say that we need to achieve a closed loop of data analysis, but when we see the results of data analysis, we don’t know how to achieve it. How can we achieve the “closed loop of data analysis” required by large companies? This article systematically explains the relevant content, let’s take a look.

Students often ask: "I always hear that we need to achieve a closed loop of data analysis and see the results of data analysis", but how can we achieve it? Why do we often send data to the business but it falls into oblivion in our daily work? How can we meet the requirements of large companies? Today, I will explain it systematically.

1. What is a data closed loop?

A complete data closed loop should be: use data to monitor business → discover problems in business → analyze the causes of problems → select solutions → continue to monitor business trends , a complete link. If a closed loop can be achieved, it must be truly data-driven, which can not only reflect the value of data analysis, but also promote scientific business operations and avoid snap decisions, which is the best state.

However, in reality, many companies do not have such links. There are four common problems (as shown below):

So how can we optimize and achieve a closed data loop? If the enterprise has a very low degree of digitalization, lacks data collection, or even has no data, then there is really nothing we can do. Therefore, except for problem 1 which is restricted by data collection, there are corresponding solutions for problems 2, 3, and 4. We will introduce them one by one in detail.

2. Key Point 1: Propose Business Hypothesis

First of all, we should be clear that if the index falls, it is not necessarily a problem. The real problem is that the index does not fall as expected. Therefore, don't keep repeating: the year-on-year and month-on-month increase or decrease.

Instead, do basic data analysis first:

1. What is the rhythm of the changes in the off-season/peak season of the business?

2. How long is the growth cycle of the new business itself?

3. Have you made any proactive adjustments to your business recently?

In this way, normal fluctuations can be eliminated and real problems can be found. In fact, many times, the reports sent out are criticized by the business department: "I knew it long ago." This is because they lack the above work and recite "year-on-year and month-on-month" every day. You will be annoyed when you read such reports.

The decline in indicators is only a symptom of the problem. What people really want to hear is:

1. Is this an internal or external problem?

2. Is the internal problem due to poor strategy or poor execution?

3. External problems, coming from competitors or the overall bad environment, are everyone failing?

These are collectively called "business assumptions". Only when data is combined with business assumptions can the business be truly driven to take action and solve problems. Otherwise, if we only say: there are fewer customers, the average order value has dropped, and the conversion rate has dropped, what exactly is the business assumption? Too many reasons may affect this result, and the business cannot make judgments and take actions.

Therefore, when discovering the problem, it is important to put forward the business hypothesis. There are two common ways to put it:

Method 1: From the data, we find that the problem is more concentrated in XX place, so we assume that...

Fangfa 2: The business has its own predictions. First, record the business description and then convert it into data problems

The two methods are shown in the figure below. Either way, they can promote the next step of work.

3. Point 2: Establishing analysis logic

There may be a lot of business assumptions to be verified for the same problem. At this time, it is important to establish analytical logic to determine which one to answer first and which one to answer later. Otherwise, various factors are intertwined, and it is impossible to determine the main cause and promote implementation.

In real work, it is often impossible to exhaust all possibilities due to limited data collection. Therefore, elimination method is a good way to focus on the core problem more quickly by eliminating interference items. Here are three classic principles of elimination method:

Principle 1: External causes should be excluded first

Because too many people like to use "bad environment" as an excuse to pass the buck. If you don't plug this loophole first, there will always be excuses to mess up. In fact, if you really encounter a big environment problem, it must be affected in all business lines. As long as you can find enough benchmarks (as shown in the figure below), you can plug the excuses and find a way out for the business.

Principle 2: Execution issues should be eliminated first

If the execution is not in place, even if the strategy is good, it will not be effective. Therefore, whenever you suspect that there is a problem with the execution, you must first check the process. Have you sent out the necessary publicity? Have you done the necessary training? Have you prepared the necessary goods? Are the functions that should be launched running stably? Have you visited the customers that should be visited? If the execution is in place but it is not effective, then you can review: Is there room for optimization in the strategy? What else can be changed?

Principle 3: Prioritize those with backup plans

Attention! Enterprises are not scientific research institutions. Finding out the reasons is far less important than recovering KPI indicators. Therefore, if there are backup plans in the business, use them if you can. For example, if you see that the performance is not good, and there are backup expenses, then prioritize calculating whether the investment can be pulled up. For example, if you see that the communication is not good, and there are backup materials, then prioritize analyzing which materials may be effective after being put out. In short, be problem-solving oriented, rather than asking like an old pedant, "Is the analysis clear?"

After screening and sorting, a clear set of solutions can be formed for specific business problems. It is even possible to form detailed analysis and solutions for different cities, different teams, and different users, thereby promoting problem solving (as shown below):

4. Key Point 3: Evaluate Feasible Solutions

Many students get scared when they hear "feasible suggestions". They don't know how to make them, and they don't know how detailed they should be. In fact, in the short term, there are very few optimization actions that can be taken by the business, so there are not so many fantastic ideas. Moreover, ideas that are not backed up by data are likely to be bad ideas, and the better the transformation.

Therefore, if you want to make high-quality and feasible suggestions, the best way is to organize the results of past evaluations around business issues. For example, if you want to make suggestions for the last promotion, you can use the data on the effects of various types of promotions in the past and the investment-output ratio. First, see which one is better with the same investment and choose a major category. For example, if you want to suggest optimizing promotional materials, then what was the effect of the promotional materials in the past in bringing in goods? First make a list and provide it to the business for selection.

This makes it very convenient for the business to use. Once the business has an idea, they can immediately see the results of similar actions in history, and naturally they can make their own judgments.

The only situation that cannot be dealt with is when the business has a brand new idea. In this case, there is no historical data for reference, so you can do a test. The data analyst helps the business to clearly list: the test objectives, the key influencing factors of the test, the interference factors to be eliminated in this test (very important! Otherwise, the results will be difficult to explain), and the estimated time for the test to take effect. This can also provide effective implementation suggestions.

V. Practical Remedial Measures

Of course, in real work, there will always be problems such as company leaders not paying attention, business not being active, too little data, etc. How to remedy this? You can't expect to work in a 100% perfect company, so I suggest that you hold the spirit of "if you can't do something, look for the reason in yourself" and do your own work well first.

Including but not limited to:

1. Stop being a pedant and learn more about the business

2. Think about problems from a business perspective and proactively propose business hypotheses

3. When you encounter a problem that is entangled, take the initiative to list the analysis logic tree

4. Don’t stop at “what is a habit”, ask more: “what is the reason behind the habit”

5. Keep good records of assessments and organize the assessment results around the business to keep as a reserve

Author: Down-to-earth Teacher Chen WeChat public account: Down-to-earth Teacher Chen