If your boss doesn’t understand data, are you still data-driven?

Regardless of the industry, data-driven work has become a trend. So if we work in an environment with a small amount of data, or if the business leader does not understand data-driven, how can we develop data-driven business? Let's take a look at this article, hoping it can help you with your data analysis work.

I recently received a consultation from a reader. The boss of his company asked him to do data-driven development. I have also done several consultations on data-driven development in the past year and found some common problems that were not included in the previous content, so this article was written.

If you are currently working in an environment where there is very little data, or you also want to develop data-driven operations in your company, after all, team building is a basic capability requirement proposed by Professor Yu Jun. As a bystander and eyewitness, these are all issues you need to think about. As the industry gradually matures, relying on data-intensive operations is an inevitable result, and data-driven operations are the inevitable fate of every working person.

Background: The boss communicated with my readers and felt that the team’s current work lacked data knowledge, and most of the judgments were based on the intuition of the business leaders. He hoped that the team could use data to make business and operational decisions, so he found me.

1. Reviewing the question is the most important, the most important, the most important

In my own research, I found that people who promote data-driven development have a natural tendency to believe that analytical ability is the most important, and that a team cannot perform data analysis because it does not have the ability to do so. I still focus on data analysis. However, from a business perspective, it is not a serious problem for a team to not be data-driven. The most serious business problem is still a fatal problem in the business, such as no growth. Whether it is revenue or business growth, this is the most painful problem for the boss.

In fact, it is very good for a business person to maintain high growth even if his/her quality is not high (usually this situation does not happen). If this happens, I think the boss will most likely be very happy. The high growth even if the quality of the staff is not high means that the staff is highly replaceable and the cost is low, which is a good business.

So this brings us to the first point, “reviewing the question”. Reviewing the question is very important. What does the boss want when he says he wants data-driven?

When most bosses say they want their teams to be data-driven, they actually just want business growth. Moreover, most bosses themselves do not understand data and may not have experienced a data-driven business. They believe that a clear understanding of the data can guide the business and they overestimate the value of data. The core problem still lies in the business, so those who need to implement the establishment of a business team should check whether the team has the foundation for data-driven development.

Find out what your boss wants and how he views data-driven development. Does he see the team's data-driven development as capability building or as a business strategy? In an organization without data capabilities, data-driven development is more about capability building at the beginning.

Because the business is promoted by the organization that supports the operation of the business. If the organization does not have data analysis capabilities or does not recognize this capability, it is impossible to drive the business data. It is impossible for the business to be turned around just because the boss has data analysis capabilities or the business analysis department has analysis capabilities.

1. Data-driven is the job of the CEO

Usually, to promote a team to be data-driven, it is necessary to drive the team from multiple directions such as data infrastructure, tools, workflow, organizational division of labor, and personnel performance, so that the entire team can enter the way of data-based daily work, because a single department cannot schedule all links, and any link that is not done well will affect whether the team can be data-driven in the end. This is why data-driven is the responsibility of the top leader, the CEO.

2. Common Mistakes Bosses Make When Making Teams Data-Driven

Usually these problems will lead to the failure of data-driven development. This is why it is difficult to drive data in a team. These problems can become a checklist for data-driven development within the company or when you want to do data-driven development. If you encounter problems here, you should either solve them or leave. Because these problems will lead to the failure of the final result.

First of all, we need to think about whether the business has the conditions for team data-driven. Usually, a business is either in a growth phase or a relatively mature phase, which is more suitable for data-driven. That is, the business itself is growing and the team is in a running stage. If the business is currently slowing down, it is most likely not suitable for data-driven.

Because bosses and CEOs will focus more on data-generated business strategies due to business pressure, but usually when a business enters a period of no growth or slow growth, it is not a business strategy problem. Since it is something that a top leader needs to do, we will list the core mistakes that top leaders or bosses often make.

1. The CEO has no data analysis skills and has too high expectations for data

The CEO’s lack of ability to translate business into indicators and the lack of knowledge of the relationship between indicators and business actions will lead to:

Problem 1: Formalize data communication for business.

He doesn't use data to communicate with the team. Why do data analysis become a formality in many companies? Most of them are because the top leaders are unable to analyze and have no sense of data. This will lead to him not knowing the value of data and indicators, and not knowing how to operate and explore the direction of operation through data. Gradually, the reporting of data will become a formality. Because it has no value, this way of working is difficult to maintain. It's like a person who exercises to lose weight. If he doesn't lose weight after exercising, it's difficult to stick to it. There must be positive feedback.

Question 2: I always feel that the data is not enough.

Because he can't get any insights, he thinks he needs to expand the search area and thinks the current data is not enough. Because he doesn't know which data can provide insights into the content, it becomes a purely business-based data decomposition. You should know that the best state is to use the minimum cost to obtain enough data, just enough is enough. Especially in the early stage of the business, using WeChat and Douyin to operate the business, these tools themselves provide data to support their own business judgment.

Whether the data is clear enough is like the circle of justice that Professor Luo Xiang talked about. There is no absolute justice, and there is no absolute clarity of data. Clear data itself has development costs and time costs. We all need to make judgments under limited conditions.

2. CEO management issues in data-driven

Question 1: Usually, for short-term benefits or due to business pressure, the boss will directly push the business to the business division or data science team, which will cause several problems.

First, it is the business side that actually implements the business operations, which is equivalent to the business side operating the business completely independently of the business. Second, it will cause the business division or data science team to be caught in the middle, with the boss's business pressure on one side and the business side generally not involved. Third, this will directly cause the business division to change from a supporting business role to the opposite of the business.

Let me explain this picture in detail: When it comes to team data-driven things, it is the CEO and the business side that really drive the business data. People who really do the front-line work must start to understand data and have data analysis capabilities.

The CEO does not know which numbers to look at, nor does he know what the data represents, and it is difficult for him to communicate with various business parties. The core is still the CEO and the team to operate the business, and the data analysis team assists in business decision-making, provides data, and reduces the cost of data acquisition.

To put it bluntly, the business side understands the business, and the data side understands the data. The business side has the right to implement the business and they need to cooperate with each other. A good process is like the one on the left, and a bad process is like the one on the right. Many bosses will put the business pressure directly on the business side, which is a typical unequal rights, responsibilities and interests.

The second point is that data-driven is harmful to the business side in the short term. Who doesn’t like to slack off and eat from the same pot? If you suddenly make all business status and operating actions transparent, it will not be easy to fill in the gaps.

The heads of various departments do not have data capabilities for organizational change. They do not have strong motivation for reform, but in the long run, it is a good thing both from a business perspective and from a personal perspective. Therefore, data-driven development should be "quietly introduced" and gradually strengthened. Even if the CEO directly promotes it, the intensity should be gradually increased, otherwise it is easy to cause confrontation with the business side.

In many consulting cases, the boss has the motivation to change, but the stakeholders do not. However, the boss cannot push it hard for various reasons, and the people who resist cannot get around it. They are worried that if they push too hard, the team will fall apart, which leads to the project being shelved. This is why many people say that data-driven is something that a CEO should do. Only when the CEO drives the front-line business personnel to participate in this through work processes and organizational processes can data be integrated into daily work.

3. The team lacks data capabilities and cannot advance the process

There are two main reasons why the team cannot be driven by data. The first is the lack of a personnel capability model. The reason why the business is not data-driven is that the people who do the business are not data-driven. The business is done by people. The second reason is that the business side lacks methods and strategies to improve data capabilities. Therefore, simple requirements can easily lead to confrontation between the team and the leader.

4. Rashly trying to implement data-driven development through special tools

Don’t try to solve the problem through special projects. If there is a problem with the main line, modify the main line. It is useful to cram for a while, but it cannot be sustained in the main business. Tools without ideas cannot generate analytical ideas. It is like a good brush that cannot draw, not a tool. Don’t try to provide an analytical framework through tools. This is unrealistic. The complexity of real business is much higher. If a tool framework can solve analytical problems, then these companies that provide data analysis tools will not be acquired.

5. Wrong division of labor, business units are subordinate to various business lines

The team of data analysts and business analysts must be independent of the business line and report to the CEO so that they can tell the truth. Otherwise, they will find evidence to justify their points. I have encountered many companies where the business was growing as soon as the team reported, but after summarizing the results, it was found that the overall market was not moving.

6. CEOs don’t take their time

Since it is a capacity building issue and involves the coordination of many departments, the CEO must ask and follow up in person, and spend time and energy on it. Capacity building is complex, involving organization, work business processes, middle office architecture and other aspects. You cannot skip steps and need to break them down step by step. It is not realistic to want to give everything to a coordinator, as he cannot do it in terms of authority and responsibility. Since it is important to look at the time spent by the CEO, time is the most important resource for the CEO. Finally, let's talk about what data-driven is and how to do it.

3. What is data-driven?

1. What is data-driven?

First of all, what exactly is data-driven? I tried to find a definition of "data-driven". The explanation on Wikipedia says that it is a thing where the entire process is driven by data, rather than relying on experience and intuition. When we say data-driven, its opposite is more of an experience, or a direct drive.

2. What is a data scientist or business analyst?

A data scientist is someone who knows more about statistics than a software engineer, but knows more about programming than a statistical analyst. In other words, the basic qualities of a data scientist require certain programming skills.

Another part requires understanding of statistics, which is of course a relatively narrow perspective. In fact, due to the rapid rise of the Internet in China, positions such as data sicentist and business intelligence have just become popular. Compared with products and operations, data analysis is a relatively new position with unclear work content, scope and responsibilities.

I think this picture can better represent what data science is.

The most important thing is the business logic and the ability to understand the business. First of all, the company is a business operation, and any action is inseparable from the business operation. This means that data scientists must first go deep into the business, otherwise the analysis will be aimless and there will be no results if it is separated from the business.

The second is that the lower right corner requires certain programming language and data operation capabilities, which means being able to extract the required data from complex data and then do some analysis.

The last one is scientific methodology, which includes some statistical methods and business analysis methods. This can at least make our analysis scientific rather than blind analysis. The intersection of these three fields is actually what is now called data science, or being a business analyst.

3. Reasons why it is difficult for teams to be data-driven

Many places say that it is difficult for a team to be data-driven. Apart from what I said above that the CEO is the most important driving force, there are usually problems.

The first is data. Data collection is often incomplete and inaccurate. For example, when certain indicators are measured, different people may give different results. Or after an experiment is completed, the results may be different at different times. In any case, the data itself is not very complete or accurate.

Second, it is human nature to make decisions based on experience and intuition, whether we are aware of it or not, this is our brain. It is designed for this purpose. In Thinking, Fast and Slow, people subconsciously make judgments based on intuition. So, in essence, data-driven means based on data and scientific evidence. It is not based on experience and intuition, which is counter-intuitive in itself and requires energy.

The third is to establish a data-driven model. It actually takes time to cultivate a normal thinking mode.

The fourth is the environment. To do this, you need a certain level of trust and an atmosphere that encourages people to tell the truth. To draw conclusions from data, you need an environment where you can speak them out, rather than first determining the conclusion and then looking for data to support it.

The fifth one is actually balance. In many cases, data analysis is actually a balance between speed and implementation cost. For example, when we make a decision, should we do AB testing? The answer is not certain. Testing has its benefits, but it has costs. In many cases, AB experiments will slow down the pace of business, but its benefit is that it can get some optimal solution conclusions to a certain extent. So this is a question of the optimal solution between speed and cost.

4. What does a good data-driven approach look like?

What we hope a good data-driven approach will look like is improved team capabilities and business growth.

It can become a growing function. For example, our work is very rewarding and can gain a strong sense of identity, which is very important. The result is that capable employees will be more willing to stay in this function and may work harder and more proactively to bring about a very strong business impact. The business impact will bring about stronger cognitive recognition, and it will become a positive cycle.

We hope that the entire function in a company is in such a state, but to achieve such a state is not the responsibility of this functional department alone. It requires the cooperation of multiple departments. It does not exist in isolation, that is, it works together with other functions. This actually requires the support and help of various functions. Because the business does not operate in isolation.

5. Why data-driven is a capacity building issue

The essential problem is that building data-driven capabilities is a long process, and it will gradually improve the business during the construction process.

Let’s say we look at China’s large talent market, data analysis, or data science, this function is compared to other functions, I mean engineers or product managers or UI designers.

Data science is still a very young role. In fact, if you take a look, engineers have been around for decades. The role of product manager has at least 20 years of experience in the technology industry, and data science has only been around for the last 5 or 6 years. The term has not been around for more than 10 years at most, so it is a relatively young role.

This means that in the entire business environment, data science is actually catching up with other functions at a fast pace. This is particularly evident in China, where data science is still less recognized than Silicon Valley. I think it is lagging behind by a lot, but this is not just a problem in China. Even in Silicon Valley itself, Facebook or Google are constantly exploring what data science does in a company, and there is no definitive answer.

That is to say, there is a huge difference between today’s data analysts and those four years ago. Chinese companies are also slowly improving our own cognition and gradually exploring the commercial value of data analysts.

Secondly, it takes time to build a data-driven team. This is essentially a question of mindset. This is essentially about building the entire company's understanding of the data, and then the team works hard little by little. This is not something that can be changed in one or two days. It takes time.

The third is that breaking the status quo requires resources, efforts and patience, and requires cooperation from all parties. In essence, promoting data-driven is a cultural construction, which is actually a process of cognitive upgrading. Many people think that data-driven or data capabilities are a capability building. In fact, it is not that simple. The most important thing is actually the cognitive upgrade of the business participants.

6. Four stages of data-driven

How can we achieve data-driven?

Generally speaking, let's put forward a framework. Data-driven development has four stages. The most basic is the first stage, which means that data must first be able to present what happened, objectively. In other words, for example, what happened to the business yesterday or last month, how many orders were there, how much extra money was earned, what about gross profit? How is our experience? How are the indicators? Is our business making money or losing money? First of all, we must be able to know what happened to us.

Let’s not talk about insights first. Let’s first be able to objectively reflect what is happening in reality. This is the first thing. It is actually quite difficult to do this. Many companies’ data indicators are still incomplete, or the data indicators are not unified, and the caliber is not unified. So it is often said that if you ask something, everyone may give a different conclusion.

The second stage is to be able to passively support the business's decision-making. My business here is just a broad concept, which includes products and operations. It more refers to the role that our business can passively support in this regard.

For example, if a user buys three items and one item, with or without a coupon, will there be an impact on future retention? What is the impact? Questions like these are objective business issues, but if this question is raised, as a data analyst, being able to passively answer such a question can at least lead to the second step, which is that I can answer the question.

The third step is very advanced. If you can achieve the third step, you are already very good at it. Data analysis must also be able to proactively define problems. This includes the concept of discovery, which is how to guide the business side to proactively define problems.

That is to say, because the problem exists, it is no longer called a definition. Many times we need to take the initiative to find and discover business opportunities. For example, if no one asks, we can ask ourselves some questions like data analysts, such as what harm a bad shopping experience will have on long-term retention. If this is not an opportunity for business insight, but a question we ask ourselves, we may consider how to analyze this matter and may have some analysis frameworks. It is more about what the problem itself represents and how we can convert a bad shopping experience into business retention and increase business revenue in the future.

The third step actually requires a completely different capability model. This requires the ability to frame a problem. Product managers may have a deeper understanding of this, that is, to a large extent, what products do is to frame a problem, and the same is true for data analysis.

The fourth step is to make data go deeper, that is, data can be used to assist every aspect of our daily thinking and work. I believe that by this time, companies will no longer talk about data-driven, because this thing is no longer necessary to talk about, just like it goes deep into our capillaries. At this time, everyone will ignore its existence because they have already been aware of it at all times. I have not seen any company reach this step yet.

I think this is more of a framework for everyone to think about how we think about the advanced process of data.

7. Data Scientist or Business Analyst Job Description

What should data scientists do? I have listed two dimensions here, the upper and lower dimensions. The upper one is insight and the lower one is presentation. You can imagine that the lower one is more about what happened, and the upper one is about why it happened, that is, what happened and what positioning means. One is more about display, and the other is about what happens after the display. This is one dimension. The left and right dimensions are easy to understand, which is the difference between active and passive.

If you look at this quadrant in the lower left corner, it is said to be a passive presentation of things. In fact, it is basically collecting data. The simplest thing is collecting data, and there are some basic data reports.

The quadrant in the lower right corner is actually actively presenting some things, such as monitoring product operations by indicator monitoring changes. Our products are relatively proactive in terms of indicators and dashboards, but it is still a presentation-level issue, and it is still at the level of "what happened" rather than "why it happened".

The upper left part is passive, but provides insights. For example, in the four stages of data-driven development, the second stage is about answering questions, but passive questions, including answering business questions, including user stratification and classification, generally belong to this quadrant. The top right quadrant is the most advanced part, which is about actively providing insights.

First, we need to frame different and unclear issues, identify strategic opportunities, and create some sustainable solutions. This is more difficult. The more we move to the upper right, the more long-term impacts we have, and the more we move to the lower left, the more short-term needs we have. We hope to pull it to the upper right, but this does not mean data analysis. We only want to do things in the upper right quadrant. In fact, we need to do things in all four quadrants. This is a comprehensive thing. We need a certain balance, but we cannot just do things in the lower left quadrant.

8. The value of a data scientist or business analyst to the business

This is a critical point. We always talk about what data analysts should do, what insights they should generate, and what they should do. In essence, we want to increase our influence. Let’s go back to the core question of what influence is.

For analysts, whether it is analysis reports or insights, it is actually the same. Or when you do some product work or operational work, no matter what it is, what impact does it have?

The impact usually occurs in four aspects. Often, we do some analysis and then affect a certain indicator. This is the most common phenomenon. For example, we study retention and eventually improve the retention rate. The second impact on product form is that the analysis we do directly affects the product manager's design and positioning of the product.

The third is the impact on the operational process. This is often more common and is more related to people, such as the operation of its coupon allocation strategy and its coupon allocation model. These are also impacts. For example, we have made some system operation tools that have caused it to be different from before.

The fourth is to create scalable solutions, which is more about efficiency. For example, when we make models, whether statistical models or predictive models, they are essentially the same, and this model can make the business more efficient.

What can’t be called an impact? For example, completing a difficult analysis is a pitfall that many analysts fall into. They tend to pursue the difficulty and depth of the analysis. This is not a problem or a mistake, but it cannot be called an impact at this point. The key is whether the analysis is difficult, sophisticated, or grand, but the key is what happens afterwards.

The second one is participating in a successful project. If the value of an analyst is determined by the success of a project, many people will like to pick projects and will be willing to work on a project that is more likely to succeed. In essence, this problem means that there are certain problems with our incentive mechanism, or there are certain problems with my understanding of it.

Whether a project is successful or not does not actually indicate your influence. The key is that if you are the only one in a very failed project, the project will not fail so badly, which is actually an influence. But if you are in a very successful project, you cannot explain what you did to make the project more successful. In fact, this is not an influence either. In other words, we still need to produce something, and we need to consider the issue of increment.

In a nutshell, we believe that analysis itself has no intrinsic value unless its results affect something else and, most importantly, change the business.

9. How do data scientists or business analysts integrate into the business?

How do analysts participate in the workflow? They participate in every step. The upper left corner shows some analysis requests. Before we decide to do a project, DS often starts to participate.

For example, discovering user problems, from a product perspective, means discovering user pain points, discovering user value, defining opportunity points, and roughly seeing how big the ceiling of this thing is, whether there is a lot of room for improvement, and then whether it is worth our time to start the next effort, what kind of solutions we should make, and then convincing the business party or the partner that we can do this.

Then participate in the whole process and participate in tracking in time. Analysts should participate in tracking at the very beginning, because they are the users of the data in the end. Therefore, analysts should participate in the upstream and downstream cooperation of the data.

The second thing is to consider our test plant. What kind of tests do we do? How do we release the test? What are the standards? What is a good result? What is a bad result? How do we explain if it is not good? For example, the entire product cycle or any project cycle, analysts need to participate in the whole process from the beginning to the end, and it is not just doing AB testing or giving an online result.

10. Qualities of a Data Analyst

Many teams that start data analysis will be very obsessed with tools. I personally think that this thing is not that important. All the drawings are about how to do it. Essentially, what is important in data analysis is the ability to think, to think about problems, and to promote it.

For business opportunity insights, why should we do it and why not do it? Because all decisions rely on data, the probability of success will be much higher. In terms of characteristics, you have a strong sense of ownership, so you will take the initiative to look for business opportunities.

Secondly, we have a good ability to frame problems. In fact, many businesses do not have clear problems. That is to say, many times we ask some in-depth questions ourselves and frame the business problems as one problem. This ability is actually difficult to obtain, but it is a very core ability.

The third is the ability to acquire data. I won’t go into details about this.

Fourth, mathematics and statistics require some statistical analysis background to ensure that your analysis is significant and scientific.

The fifth one is very important. You need to be able to tell a good story, because analysis itself has no intrinsic value. You need to be able to push the business side to explain to others why we are doing this, why it is useful, and why they should listen to me. This ability is very important.

The sixth is the driving force. Many people find that there is no effect or follow-up after two months after doing the analysis. People who do analysis actually have to follow up with the business side. You are responsible for the results. Analysts are not the ones who analyze, but what has changed after the analysis. You are responsible for this.

So I did an analysis to find out whether this analysis is useful for the business visits I want to track. If it is useless, I did something wrong, I did not do enough, or I went in the wrong direction. It doesn’t matter if it has no impact on the business, but I want to learn something from it and try to avoid this kind of thing in the future, or why the business side doesn’t take the next step.

Speaking of opportunities, why do we have to care about the business side when we have been working for a month and it will be two months, and there is no movement. So this kind of driving force is very important. Common backgrounds, such as mathematics, statistics, physics and computer science are more suitable, but it does not necessarily have to be a statistics background or something like that. Everyone's background can be varied, but the key is to have sufficient thinking ability, a sense of ownership, and a spirit of driving.

Logical analysis and divergent thinking are abilities that are relatively difficult to acquire, while the technology itself is relatively easy to acquire.

11. Common Misunderstandings in Data Analysis

A common mistake or trap is that some people think data can solve all problems.

This is actually wrong. Data only provides a perspective, a basis for decision-making. It is not only a data perspective, nor is it a data-only perspective. For example, products have product perspectives, design has design perspectives, engineering has engineering perspectives, and data has data perspectives. When we ultimately make a product or a function, we need to summarize and analyze all the information. Data is a very important part, but it does not mean that data should override any other aspects. This is the first thing that needs to be made clear.

The second point is also very important. Seeing what happened with data does not necessarily explain why it happened, nor does it necessarily provide a strategy. From seeing the data to providing an effective strategy, it is a process. Don’t over-mythologize the power of data. Many problems are not actually data problems .

For example, the supply and demand problem. The infrastructure problem is not a business problem that can be solved through data insights. In fact, dialectics itself is not necessarily a data problem. Data can help frame the problem, but the decision itself is actually a matter of feeling. In today's situation, do we think we should focus on supply first or demand first, and how much to invest in infrastructure or supply chain? Data can help us understand and see clearly, but this itself is not a data decision.

The third one is that AB testing must be done . Needless to say, not everything can be AB tested. Secondly, AB testing will reduce the speed of business iteration. You have to wait for significance, which takes time.

Fourth, for example, if the negative effect is not statistically significant, then we should push the function in full. In fact, it is not the case. In many cases, it actually has a negative impact, but our sample size or test method has not clearly measured it. Statistical insignificance does not mean that there is no impact. This must be confirmed first. On the contrary, if it is statistically significant, then we should push the function in full, but this is not necessarily the case.

For example, if we want to improve something, such as GMV, we may do something and then GM increases by, say, 0.1%. This increase is very small, but statistically significant, and it may not be worth investing more resources in. Not all problems have a data role, but data can help you frame the problem and show what happened.

Hopefully this helps you understand how to be data driven.

Author: Arun's Growth Research Institute

WeChat public account: Arungrowth365 (ID: arungrowth365)