The common concern of hundreds of thousands of Internet practitioners!

Ximeng Zhang is founder and CEO of GrowingIO and former senior director of business analytics at LinkedIn. Zhang Ximeng has worked for EPSON, eBay, LinkedIn and other Silicon Valley star enterprises. He has 14 years of experience in Data analysis and user growth, and was named as “The World’s Top 10 Frontier Data Scientists” by The Us Data Science Central. This article is edited according to Zhang Ximeng’s speech. Originally posted on GrowingIO blog and official account, reproduced with permission.

Download the complete product Manager’s Data Analysis Manual now and upgrade your data analysis skills.

Since returning to China to start my business more than two years ago, I have been constantly asked questions about data analysis by friends, customers and partners. I find that people’s curiosity and thirst for knowledge about data analysis is very strong, which is a very good trend. However, the problems are also very prominent: people’s cognition and understanding of data analysis is fragmented, and there is a lack of an overall and systematic thinking framework; People’s horizons are more limited to data reports, BI systems, advertising monitoring and other fields, and they actually lack deep insight into data and data analysis.

I would like to take this opportunity to share with you my experience of data analysis in the past ten years. I hope that after reading this article, you will have a systematic understanding and improvement of the ability required by data analysis.

I. Three levels of data analysis capability system

Using Chinese culture to define the ability required by data analysis, I summarized it into three levels of “Tao”, “skill” and “instrument”.

Figure 1: Data analysis capability pyramid

First, what is tao? As the basis of the whole system, Tao refers to people’s understanding and cognition of data and data analysis, and is a concept at the level of values.

Second, what is “shu”? Technique refers to method (theory). To do data analysis, we need to master a set of scientific methods, which I will introduce in detail below.

Third, what is an “instrument”? Tool refers to tools, without tools, data analysis can only be empty talk, unable to land.

2. Tao: Values of data analysis

As I mentioned above, “Tao” refers to values, how data analytics are valued. To truly understand this, we must make breakthroughs in value identity, job positioning and business model.

Figure 2: Values of data analytics

(I) Value identification of data analysis

To do a good job in data analysis, we must first identify with the meaning and value of data. A person who disagrees with data and lacks an understanding of what data analysis means is not going to do a good job.

In an enterprise, the CEO and management must attach great importance to and understand the value of data analysis. If the boss doesn’t recognize the value of data analytics, how can data-related projects work in the organization? Then there is the need for a data-driven culture within the organization. If people would rather make a decision than believe the data analyst’s advice, then the data analysis is often twice the result, go through a formality, otherwise it is twice the result with half the effort.

(2) Work orientation of data analysis

To do a good job in data analysis, the value of data analysis should be clearly positioned. Don’t deify data analytics as the master key; Do not easily deny the significance of data analysis, abandoned. Data analysis should have practical guiding significance to the business, and should not become a mere formality, reduced to “taking numbers”, “making tables”, “writing reports”.

Over the years at LinkedIn, we’ve made it clear what analytics is all about: using it to make fast, quality, and efficient decisions for all professionals, providing guiding insights and scale-up solutions.

Figure 3: EOI framework for data analysis

At that time, we also adopted an EOI analytical framework, which clearly identified the value of data analysis for different businesses. For core task, strategic task and risk task, we think data analysis should play three roles of Empower, Optimize and Innovate respectively.

(3) Business model of data analysis

Do a good job of data analysis, to understand the business model of the enterprise. The ultimate goal of data analysis is to serve the growth goals of the enterprise, so it is important to have a deep understanding of the industry background, business implications, products and users.

Figure 4: Where data fits into LinkedIn’s overall strategy

Taking LinkedIn as an example, as an important link of enterprise growth, LinkedIn gave priority to the value model of data at the beginning of product design. The first is the growth, use and activity of users, then the generation of a large amount of data, and finally according to the data for business realization (enterprise advertising, enterprise recruitment, advanced account, etc.) and user growth, so as to continue a virtuous cycle.

“Tao” is the basis of the entire data analysis capability system. Only by recognizing the value of analysis, clarifying the job orientation and knowing the business model well, can data analysis be on the right track.

2. Technique: methodology of data analysis

If the level of “Tao” is a little little, then “shu” is often discussed. The “technique” I’m talking about here is the method, including the process framework of data analysis, macro methodology and specific analysis methods.

(I) Framework of data analysis

In the whole data analysis framework, the user is the source of data and the object that data analysis ultimately serves. The whole analysis framework can be divided into four levels: data planning, data acquisition, data analysis and data decision-making. From users, business systems, to data acquisition platforms, ETL, data warehouses, to analytics, BI, DM, AI, insight, to decision, behavior, value, and finally back to users.

Figure 5: Framework for data analysis

In the overall analysis framework above, the lower level takes more time and effort, while the top level takes less time. From the perspective of value generated, the lower the value generated by the bottom layer, the higher the value generated by the top layer. If you think about it, you will understand that most of the time in the process of data analysis is spent on data collection, cleaning, conversion and other dirty work, while the most valuable part of analysis and decision-making often takes little time.

Therefore, you should focus on the most valuable analysis and decision levels of data analysis, and use tools to automate the underlying operations whenever possible.

(2) Methodology of data analysis

Data analytics should help us constantly optimize our marketing, operations, products, and engineering to drive business and user growth, not analytics for analytics sake. Here I introduce two methodologies, one AARRR model for business and the other learning engine for analysis.

Figure 6: AARRR model

AARRR is known as the pirate law for Growth hackers, Acquisition, Activation, Retention, Revenue and Referral, respectively, cover the user’s life cycle.

When conducting data analysis, we should consider which part of AARRR model the user is in, what are the key data indicators, and what are the corresponding analysis methods?

Figure 7: Learning engine model

The “Learning engine” is a lean operation advocated in the book “The Lean Startup” and widely adopted by both large and small companies in Silicon Valley.

When we have an idea, we can Build it in a minimalist viable product (MVP) fashion. Once the product goes live, we need to Measure the reaction of users and the market. By analyzing the data we collect, we can test or overturn our previous ideas, and Learn and optimize.

(3) Specific methods for data analysis

The purpose of this article is not to introduce specific analysis methods, but to let everyone have a systematic understanding of the whole data analysis ability system, so I will not elaborate on each method in detail.

Figure 8: Specific methods of data analysis

It is one thing to understand the principles of each approach; it is quite another to apply them flexibly in business. Taking product managers as an example, they can integrate “user behavior – data analysis – product design & optimization” into one and master the essence and significance of various analysis methods through continuous practice and application. As for the specific analysis method, I recommend you to read an article written by my colleague Justin, which gives a detailed explanation.

Data comes from users, and the ultimate purpose of data analysis is to serve enterprises and users. Before data analysis, business objectives and data indicators must be clear, scientific analysis methods should be selected, and data should be used to guide product and user growth.

Third, tools: data analysis tools

“Big data, big data, the most important thing is data. But where is the data? What is lacking most now is a unified data acquisition platform!” That was my opinion in an interview a long time ago, and still is!

(1) Why are tools so important?

When I started GrowingIO, I was pulled into a networking group with the ceos of various Silicon Valley startups. I found the group discussion interesting in two things: one was about the methodology of entrepreneurial growth, and the other was about tools. “If you want to do a good job, you must sharpen your tools first.” This is the truth.

Figure 9: Data analysis pyramid

I’ve been talking to people about this a long time ago, and the bottom part of the data framework may take 80% of your time and effort, but produce less than 20% of your value. Everyone is building data acquisition platforms, writing code burial sites, doing ETL, and building BI systems, where there is more time and manpower to do Analytics and Insight.

When I worked at eBay and LinkedIn, there were no good data analytics tools out there. We had to deploy a lot of systems, build a lot of mechanics, and even hire three or four teams to do one thing. There are a lot of good tools on the market today to help us with data analysis. In order to save time and resources (especially for growing enterprises), there is no need to build a set of analysis system internally. You should use good tools to help you do data analysis.

(2) Choose appropriate analysis tools

What kind of analysis tool you choose depends on your job position and analysis scenario. There are several tools to choose from for each scenario, and some tools can be used for multiple analysis scenarios, depending on your familiarity and understanding of the tools.

Figure 10: Data analysis tools summary

Excel is by far the most basic and common data analysis tool available. For small amounts of data, it can be used for data processing, data visualization, and some statistical analysis. Once the amount of data is large, it needs a large database to support it.

Marketers need to analyze advertising data, and monitoring website traffic is the focus of their attention. Product and operations focus on user behavior and product usage, and user behavior data analysis tools are their first choice. People used to focus on business data, but these results didn’t tell them what was happening and why; Now people are paying more and more attention to fine operation and have higher demand for user behavior data, which is also the reason why I came back to China to set up GrowingIO.

If you have some KNOWLEDGE of R and Python, and have developed in data modeling, statistical analysis, and data science, you will be a better analytics player.

“Tao”, “technique” and “instrument” constitute the ability system of data analysis. Only by recognizing the value of data analysis, mastering the methods of data analysis and flexibly applying data analysis tools can we do data analysis well.

This article is excerpted from GrowingIO’s third ebook, Data Analysis Handbook for Product Managers.

Download the ebook now and upgrade your data analysis skills.