How did this analyst go from making decisions based on intuition to supporting them with data?

Why can analysts make data-supported decisions? This article will help you find out the reasons. This article is also recommended for data analysts.

Data-driven decision making is a buzzword that everyone talks about every day. But how is data driven? Few students have actually seen the whole process. Some students are always confused: "I am being chased by others to count, and I feel like I am being driven." Today, I will explain it systematically, and refuse to follow the trend.

1. The most original decision-making process

First of all, do you need data to do things? The answer is: absolutely not. In theory, you only need two horizontal lines and one vertical line to do things: just do it! So the simplest and brainless way to do things is (as shown below)

But everyone knows that such empty slogans are useless except for driving employees to death. Because it is too brainless. No one knows what to do, how to do it, and what the result will be.

2. The embryonic form of scientific decision-making

Therefore, this original decision quickly enters the second stage: the three-shot stage (as shown below)

This three-stage decision-making has already formed the prototype of scientific management, and it became popular with the popularization of the contract system in the late 1980s and early 1990s. So many leaders born in the 1950s and 1960s, as well as those born in the 1970s and 1980s, still like to use the three-stage statement: what to do, how to do it, and how to do it. It is also common in articles and books.

However, this is not scientific management in the true sense. It is too crude. In particular, leaders often make decisions based on their own ideas. Their favorite catchphrase is: "First, achieve a small goal and earn 100 million yuan." As for why it is 100 million yuan, why it is necessary to make money instead of occupying the market, there has never been an in-depth analysis and explanation. The result of making decisions based on one's own ideas is to pat one's chest in front of others and pat one's thigh when something goes wrong. Therefore, there is the nickname "three pats".

3. From extensive to fine

If you want to improve the refinement of decision-making, you have to introduce data measurement and data analysis. It can be said that data analysis is born to serve scientific management. With the support of data, a lot of refined management can be done.

Before decision making:

• Quantitatively evaluate the current operating status, income, expenditure, and profit.
• Assess opportunities and threats based on market, competitor, and consumer data.
• Based on past performance trends, we can discover the laws of natural change and set more reasonable goals.

In decision making:

• Quantify the time, manpower, and material resources required to evaluate alternative solutions.
• Based on past performance, quantitatively evaluate the feasibility of the plan and assess the expected completion rate.
• Based on past performance trends, we can discover the laws of natural change and set more reasonable goals.

After decision:

• Quantitatively monitor the execution process and identify execution problems.
• Analyze implementation methods and evaluate optimization and adjustment measures.
• Review the execution results, check the actual completion of goals, and summarize the experience.

At this stage, data-driven decision-making and data-based management have been achieved. The most classic theory at this stage is the PDCA theory. It divides the decision-making process into four stages:

• Plan
• Do
• Check
• Act Processing

Through iterative cycles, we ensure that our goals are achieved and that the quality is gradually improved (see the figure below).

It sounds like data-driven decision-making has reached its limit here. In principle, this is true. Many classic management theories are established at this stage. The subsequent evolution is mainly reflected in technology. Because obtaining data is a very difficult task that requires a lot of technical support. Therefore, the level of technical means directly determines whether management theories can be implemented and whether they can be innovative.

Many classic management models, such as AIDMA, PSM, and double-blind testing (ABtest), are based on survey questionnaires. Although survey questionnaires can theoretically collect all the data, surveys have inherent limitations:

• Human memory is limited, and data accuracy is not high
• People have limited energy, so we can’t have too many questions or too complicated options.
• People are lazy, option A is always chosen the most, and the proportion of 3 points and 5 points is always the highest.

Due to the above limitations, good questionnaires are costly, have a long cycle, and the speed of obtaining data is extremely slow.

Therefore, subsequent improvements in management methods are more accompanied by improvements in data collection methods, which become more and more sophisticated.

4. More sophisticated data-driven processes

More sophisticated data-driven methods are all technology-driven:

• With OMS/CRM system, you can collect transaction + user data more immediately
• With APP/Mini Program + data embedding, user behavior data can be collected
• With rich data, we can improve user portraits and make prediction models
• Through tools such as CDP/ECRM, data can be directly pushed to business execution

With technical support, management details are richer (as shown below)

Submitted in the previous stage, the main added contents include:

• Decompose the overall goal into sub-goals for each department and each link, and implement them step by step
• Clearly distinguish indicators and judgment criteria, and introduce comprehensive evaluation methods
• Added CDP (select old method) and ABtest (test new method), making it easier to select a good method

The overall process can be described in detail in the following figure:

The most popular method at this stage is the OSM method, which promotes decision-making by decomposing/quantifying indicators step by step (as shown below)

Note: To achieve the driving effect, you need to configure appropriate data tools at each stage of the decision-making process and play a role separately. It is a combination of tools, not a super powerful model or formula to calculate super powerful results.

Throughout the entire work process, the biggest technical difficulty lies in data collection. It is very troublesome to ensure high-quality, multi-dimensional data collection without delaying the overall progress of the project. The biggest difficulty in business is consensus. How to avoid arguments and gain recognition is the key (as shown below)

That’s all for now. But some students must be curious: This management theory doesn’t seem complicated, why can’t I see it in reality? That’s because there is always a gap between theory and reality. When it comes to enterprises, all kinds of weird people and things emerge in an endless stream.

5. Why you don’t feel data-driven decision making

Problem 1: Backward people/systems/institutions.

• Quite a number of companies do not attach importance to data construction, especially data collection and process standardization.
• Quite a number of corporate systems are still in the primitive stage, working quickly and aggressively.
• Quite a number of business leaders are still using barbaric management methods.

Times have changed, but people/institutions/systems remain the same, and nothing will change.

Question 2: Being too selfish and deliberately distorting data.

Many leaders only talk about digital-driven development, but in fact they just use data as a facade, write more good-looking numbers, and make ugly numbers look good in various ways. If it doesn’t work, they will blame it on “our company lacks artificial intelligence and big data analysis capabilities”. This will definitely not achieve data-driven development, but rather play a numbers game.

Question 3: Blindly believing in artificial intelligence and big data.

Note: From the evolution of data-driven development, we can see that to achieve data-driven development, we need the division of labor in the early, middle and late stages, and the coordination of reports, management models, algorithm models, test platforms and special analysis. It is not a "super-powerful intelligent model" that can be modeled with a "Duang!" sound. However, some people do not believe it, and some people think that there is an omniscient "model" in the computer that can determine the outcome of the world with just one model, which naturally ends in tragedy.

Question 4: Too much emphasis on indicators and neglect of standard establishment.

This is a common problem for data analysts. When talking about analysis, they often mention dozens of data indicators, but which one is the main indicator, which one is the secondary indicator, and which one is the reference indicator. Which indicators should be combined to look at, and what indicator value is considered good and what is considered bad. There is no clear standard, no consensus with the business. In the end, they only know how to list the data, but cannot make any conclusions.

Question 5: Out of touch with the business and lack of data accumulation for business processes.

This is a common problem for data analysts. Every day they only focus on GMV, traffic, DAU, MAU, conversion rate and other indicators, but they know nothing about business processes and have no observation and accumulation of the effects of different business methods. In the end, apart from repeating a few indicators over and over again, they only say: we must increase them, we must maintain them. This is all nonsense and cannot drive decision-making.

VI. Summary

Data-driven decision-making requires close coordination between business processes and data, and the participation of leadership to achieve it. Data-driven decision-making has never been a matter of a hermit with ingenious calculations who can just chant the spell "Mommy, Mommy, boom" and come up with a shocking conclusion. This is basic common sense.

Outdated people, outdated systems, and outdated processes will make data superficial. Therefore, even if you don’t see results in a short period of time, don’t lose confidence. These outdated people and things will eventually be eliminated in history. As practitioners, we need to exercise our own abilities so that we have the opportunity to join better platforms and create better projects. I encourage everyone.

Author: Down-to-earth Teacher Chen

Source public account: Down-to-earth Teacher Chen (ID: 773891)