Former LinkedIn boss Zhang Ximeng: All business models must ultimately be more efficient!

Focus on user experience; this link will generate a lot of data;

Pay attention to operational data; find growth points for performance and revenue through analysis of operational data;

Focus on monetization; further promote the iteration and update of user experience based on monetization requirements. Behind LinkedIn is a very large and sophisticated data operation platform, which is driving the development of LinkedIn;

Further reading: Reid Hoffman’s classic book, The Alliance: Talent Transformation in the Internet Age, states on the first page that “lifelong loyalty to a company has become history.” Since reality is so cruel, how can we, as employers, better recruit and retain employees? Reid Hoffman, founder of LinkedIn, recommended two effective methods in "Alliance" in Silicon Valley technology companies:

➀Three terms (rotation period, transition period, basic period)

The company and employees set different goals for different periods. Before the expiration of the goals, the completion degree of the goals will be assessed and whether to renew them will be considered. If renewed, goals for the next period will be set. It is a moral constraint and has no legal effect. The goal setting of both parties is based on full trust and fairness. Therefore, in theory, there is no situation where the leader gives an unreliable KPI indicator, the employees complain while working, and are fired or resign voluntarily when they fail to meet the target by the deadline.

Of course, this method places particularly high comprehensive demands on leaders. When setting goals for the term, they must not only meet the demands of employees, but also ensure that the company's goals can be achieved. From this, it can be seen that in the current fast-paced Internet companies, the previous simple and crude model of distributing company KPIs downwards, or the leader's simple and crude model of requiring a function to be launched within XX days, is gradually becoming unworkable!

➁Company employee contact network

Since lifelong employment is not possible, one will have to leave sooner or later. All good things must come to an end, but people shouldn’t leave and the tea gets cold. Companies should proactively invest in and operate a network of colleagues (similar to a university alumni association). The book cites the examples of former employee alliances (or networks) such as PayPal, Procter & Gamble, and LinkedIn.

The founders of LinkedIn, Tesla , YouTube, Yelp, and Yammer all worked at one company: PayPal. LinkedIn now has more than 118,000 company colleague groups, covering 98% of the Fortune 500 companies, while P&G's employee group is completely independent of P&G and now has more than 25,000 members, as well as a charitable foundation and a speaker group. Tencent and Alibaba, which are well-known in China, will permanently retain an employee's work number, corporate email address, etc. within the company.

➂The benefits of employees establishing connections and forming alliances:

• Former employees may come back, or they may recommend excellent employees to join the company, which can significantly save recruitment costs.
• When current employees encounter problems, they can quickly find reliable former employees for help
• Create a word-of-mouth effect, everyone says this company is good, then it will be more attractive when recruiting!

EPSON is a manufacturing company, or an Internet company. It may be the earliest company to use the "Internet model" in its business model. When selling printers, it basically does not make money and can only break even. The profitable business is ink, with a profit margin of more than 90%. When doing marketing activities, it is necessary to make a forecast of the income of this activity. Each activity must be calculated through statistical models plus a large amount of manual data entry. The final result can be obtained that the deviation between actual business and forecast is no more than 5%. This is already a good example of refined operation.

Basic principles of data analysis : Noise and distortion are bound to occur during data collection, and the final result cannot 100% reflect the business process, so this requires people with business experience and business intuition to make judgments; the data may not be accurate, but it needs to be sustainable, and cannot be accurate sometimes and inaccurate sometimes, otherwise it will not be analyzable.

In the world of mobile Internet , the Matthew effect is very obvious, that is, 20 apps account for 71% of users' usage time; all the other millions of apps and websites have to compete fiercely for 29% of users' time, which requires start-ups to run faster, more efficiently, and have better business models. In the United States, its overall user growth is currently in single digits. In a world where "traffic is king", the world is about growth, but now all competition is based on competition in the existing market, which is a competition of speed and efficiency.

User is king VS traffic is king: In the process of competition for existing resources, all business models need to return to "user is king", take users as the core, products as the performance, and data operations as the compass, "twisting the towel" in the existing world, and ultimately achieve higher efficiency.

Growth is king! If everyone in a company needs to pay attention to one thing, what should it be? It's growth! Why should we pay attention to growth? There are three main factors:

• Valuation: Only rapid growth can guarantee market performance;
• Only growth can continue to provide high-quality services to more and more customers;
• Society needs companies to continue to grow and create social value.

I recommend the book "The Four-Step Entrepreneurship Method". You can also check out the past content in "Tong Jilong's Notes": "Customer Development Methodology, a Must-Read for Product Marketing Managers - Notes on the Four-Step Entrepreneurship Method." 》

The “Pirate Law” of User Growth A AR RR:

Later, I searched for some of Simon’s speeches online and found some of their valuable insights on data analysis, which I would like to share with you below:

Data is a connection. It connects the four most basic quadrants: time, place, task, and event. Why do we say that data will be the most important indicator of the next wave of technological revolution? Let's take a look. According to reports from several top research institutions in the United States (Gartner, IDC, etc.), in the next five years, we will have 4 billion people generating various data through the Internet, which will create a 4 trillion US dollar market. There will be 25 million types of software connected, 25 billion various devices connected to various data systems, and 500 trillion GB of data generated.

American data analysis framework and methodology: In the United States, a very systematic data analysis methodology has been formed. This methodology began to be used during World War II and has been applied to various aspects such as military, science and technology, and people's livelihood. Data analysis can be broken down into several steps, and you can still see that each link from the beginning to the end is based on the condition of continuously increasing value.

First and foremost, correct data collection and implementation of data labeling methods will exponentially promote the rapid production of results from future data analysis. This is also the part that is missing or seriously neglected by some companies.

Second, the engineering architecture of big data, data warehouse, and distributed computing level. Today's distributed computing systems are very different from the overall architecture of previous data warehouses. This requires our IT departments to keep up with the pace and implement and deploy new open source-based distributed data technologies, such as the relatively mature Hadoop. This technology has been used in the United States for nearly 10 years and has gradually become mainstream in Internet companies.

Third, responsiveness analysis. This is probably what most companies do the most, which is to constantly use data to answer various questions raised by the business side, create simple reports, business intelligence, BI, etc.

Fourth, diagnostic analysis. For example, multi-dimensional attribution, implementation of score cards, etc.

Fifth, strategic analysis. Competition trends, price elasticity, judgment of corporate financial revenue, etc. In the past, high-level strategic analysis of enterprises has always been dominated by companies such as BCG and McKinsey. But why are they not at the top of the pyramid now? It is because of the emergence of big data.

Sixth, predictive analysis is the analysis of future business based on statistical models, machine learning, and various large-scale simulations and optimizations.

Seventh, we can return to the full data automatic analysis and decision-making we just talked about.

The current state of enterprise data analytics in real-world situations. Let's take a look at how most companies accomplish this. Are all of you industry leaders here, especially CTOs who focus on technology, familiar with this diagram?

Isn’t this the data flow diagram within the enterprise? What would you think if I told you that this diagram is a flow chart of the sewage treatment of Hamilton River in the United States? This picture shows the process of turning a polluted river in the United States into clean water, which is very similar to the data analysis process today.

The basic process of data analysis within many companies is just like this: a lot of dirty water flows in, and we need manpower to monitor it in various ways and put it in a pool for sedimentation. Then we do various cleaning, aggregation, re-cleaning, re-disinfection, and then transmission, step by step. The United States has conducted a study that actually shows that the process from data collection to the final generation of useful commercial value of the data takes three to five weeks. Just think about it, making a simple decision takes three to five weeks. What a painful process it is. If future companies want to succeed in their data strategies, they must be able to quickly turn sewage into clean water.

Let’s take a look again. The real value is generated at the top of this pyramid. According to a research report by DJ Patil, chief data scientist at the White House, 90% of data engineers and analysts' time is spent on data collection and cleaning, and only about 10% of resources are spent on work that can generate a lot of business value.

Traditionally, the entire data analysis is processed sequentially by several departments, which makes the performance very slow. When we talk about big data, data starts with the collection of data labels, which is generally implemented by front-end engineering personnel, and then the data transmission is managed by IT. ETL is generally handled by the data warehouse or data platform team within the enterprise. The BI (business intelligence) department is in the analysis department or exists in the business department, and then we have various business analysts and statisticians involved. This operating framework system has a lot of people involved in various departments and the process is very long, which greatly reduces the efficiency.

Internet companies, in particular, which are at the forefront of technology, have made various attempts. For example, if they want to break the deadlock, they have to integrate various functional departments. However, because functional departments require employees with different abilities and experiences, it is difficult for departments that understand the business to truly understand the technology, and departments that understand the technology do not have enough energy to fully understand the diverse needs of the business departments. This results in slowness and inefficiency in several decision-making links. In order to meet the ever-increasing demands, enterprises need to build and customize various IT systems. This customization has resulted in the formation of several data micro-islands in various departments within the enterprise. Several enterprise data islands have further increased the workload of the IT department, which requires data integration of various internal customized systems to make various unified data decisions.

In the short term, this customized data integration seems to solve the company's information decision-making problems, but in the long run it will even further slow down the company's decision-making speed. Please look at this data analysis pyramid diagram. In the past few years, we have found that the real value generated by big data analysis is the above 10% of investment time, which will generate more than 90% or even more than 90% of value. But if he does not have the time and resources to do the following 90% of the work, he will not be able to create any value. Including sales management is also a digitally driven operation.

As China is developing rapidly today, we have to ask ourselves whether each of our companies needs to build a "sewage treatment plant" or repeatedly develop and deploy so many types of software to provide corporate analysis services. The opportunity before us today is how to effectively adopt advanced methods to cross various technological and management gaps and make our companies more efficient. Moreover, as the demographic dividend decreases, the most important thing we need to do is for companies to increase efficiency.

Building a data-driven closed loop: How to improve the scale and efficiency of data analysis and operational decision-making?

The main approach is to simplify the data analysis process of existing businesses on a large scale, so as to achieve end-to-end integration and make the decision analysis system a closed loop. The speed of this data analysis closed loop is basically equivalent to the speed of enterprise decision-making. The closed loop of enterprise big data analysis must have at least two components: the first part: the participation of the business side, and the second part: the implementation of the technical side.

The more external participation there is on the business side of this decision-making loop and the less and faster the internal implementation on the technical side, the higher the efficiency will be. How to understand it? The latest research materials from authoritative institutions in the United States mentioned the concept of shadow CTO in the next generation of data revolution, that is, the IT department should become an external manager of enterprise software rather than an internal executor. Moreover, the cloud-based SaaS software in the United States puts the information decision-making function in the cloud, thus transcending some of the lengthy IT processes and technical gaps.

This has been well demonstrated in several trend-setting companies in Silicon Valley. For example, various departments of first-class companies such as Salesforce, LinkedIn, Facebook, Uber and Airbnb are increasingly adopting various SaaS-based solutions instead of building them all by themselves.