Overview of series articles:

7. 1 correlation and correlation coefficient analysis

7.2 the logical relationship between things and scientific laws

7.3 Causality and causality, the development logic of intangible things

7.4 The complexity and scientific abstraction of the development law of things

7.5 Causality and Regression Analysis

7.6 logistic regression

7.7 Relevance and * * * Life-Exploration of Phenomena and Laws

The core purpose of data analysis is to grasp the development law of things. Only by mastering the development law of things can we better control things and make them develop in the direction we expect. We have mastered the law of the development of things from the data, so we can create the future according to this law and make things happen and develop in the direction we expect.

No data analysis method itself can directly tell us the causal relationship of things, and most causal relationships are obtained through the logical judgment of the human brain. Scientific experiments are also revealing the relationship between things, and then human beings can get the causal relationship through the logical judgment of the brain. The rigor of scientific experiments makes the judgment of causality more rigorous, and many data analysis in the fields of economics and management reveal the relationship. Because sociology, economics and management are not complicated, it is difficult to test them repeatedly in most cases.

In most cases, the thinking mode we adopt in the field of data analysis is causal thinking, that is, according to the results of things' development and changes, we look for the reasons that may affect the results, and then use data to verify or quantify this influence relationship. Causal thinking is the basic method and thinking mode for data analysts to construct mathematical models.

When building a mathematical model, we need to find out the reasons that affect things. For example, sales are the result, and the factors that affect sales may include advertising investment, brand influence, product quality and consumption experience, distribution rate, competitor strength, promotion activities, pricing and so on. Quantifying these factors will form a sales model, then collect relevant data, establish a mathematical model, and then constantly test whether the mathematical model is reasonable in practice, or optimize relevant data indicators and coefficients, so that we can have a more rational grasp of future sales.

Y sales =f(X 1, advertising investment, X2, brand influence, X3, product quality and consumption experience, X4, distribution rate, X5, competitor strength, X6, promotion, X7, product relative pricing).

Among them,

X 1, advertising investment: this is a relative concept. On the one hand, it reflects our company's investment, on the other hand, we should also consider the advertising investment of other competitors in the industry. For example, although we have increased our advertising efforts this year, if our competitors put in more and more effectively than us, then our advertising efforts are still relatively weak.

X2, brand influence: Also known as brand equity, different companies have their own mathematical models to evaluate brand power, and some market research companies also have their own models for customers to use. Different categories will have different influencing factors, which need to be explored by enterprises themselves. Generally speaking, we should consider popularity (user's mention rate and first mention rate), reputation (user's preference), loyalty (whether users will try other brands) and recommendation (whether users will recommend them to friends).

X3, product quality and consumer experience: You can refer to consumer survey data or use product competitiveness model to score. Different companies will have different research schemes, and each company can design relevant models according to its own ability and actual situation.

X4. Distribution rate: that is, how many stores of your product have entered the shelves in the target market area, which can be expressed as a percentage. Of course, there are also problems about the quality of goods and shelves, which need to be accounted for with a certain correction coefficient. This correction coefficient is evaluated and corrected according to the shelf quality evaluation standard and the store quality evaluation standard, combined with the shopping characteristics of the target audience.

X5. Competitor's strength: This is a relative concept, which can be evaluated by factors such as competitor's market concentration and brand influence.

X6. Promotion strength: including discount rate, rebate rate, participation, etc. If the user participation of an activity is low and the coverage is small, the promotion activity will be weak.

X7。 Relative pricing of products: Based on the pricing of similar products of our company, if Coca-Cola is priced in 2 yuan, then the price of Coca-Cola is 1.8 yuan is too high, because its brand influence cannot support this pricing level. Relative pricing gives consumers the feeling that the price is high, low or moderate, and this data can also be obtained through market research.

To sum up, you will find that through causal thinking, we can explore the "causal relationship" of things by constructing mathematical models and quantifying the factors that affect the results with data, thus guiding enterprise practice and making enterprise management more rational, scientific and data-based. The sense of heaviness and strength given by an enterprise also comes from the precipitation of its history. Some enterprises have history but no history because they have not accumulated data in the development process, and there will be no data-based analysis and data-based business experience precipitation. Even if there is precipitation, it will be deposited in the brains of employees. If the mathematical model and data management are not regarded as the company's management know-how, and the past history cannot provide the basis for future management decisions, then it is not a historical enterprise.

The longer an enterprise does it, the more refined it should be to quantify its sales volume, which is the enduring logic of the enterprise. If we don't manage these business know-how through digitalization, then we will never understand the driving factors behind sales and the relationship between these factors. P&G has a history of nearly 1, 80 years, and it is still firmly in the position of the global leader in FMCG industry, which is inseparable from its powerful data analysis team. During my five years working in this company, I have been studying consumer demand, constructing the logical relationship among consumer demand, sales volume and profit, constantly improving this relationship, and monitoring the influence of changes in consumption trends and the development of competitors on us. This is the secret that an enterprise can master the process of market changes and always stay ahead in this process.

The development of everything has its regularity, and accidental phenomenon is the inevitable result. We think it is an accident or a random event because we know too little about the details of the laws of things. We regard all unexplained events as accidents. The more laws we have, the more data and facts we have, the more clearly we can know how things should happen.

When we know more laws and master more data, random events will be reduced, errors will be reduced, and accidental events will be more controllable. When we master the regular relationship between advertising and sales and more relevant data, we can accurately predict sales and calculate profits, so as to judge whether business management activities can make money. If not, it is necessary to stop the project in time and use the time and funds in the most effective place.

The more excellent the enterprise, the stricter the control of laws and data, and they regard data as a "belief" in management or in systems and processes. "If you don't record the data, your work is equivalent to not doing it!" This is the idea that the author was instilled in P&G work. After employees finish their work, they need to sort out relevant forms, data and records before the end of the work. In the process of providing consulting services for domestic enterprises, the author found that these enterprises have a general consciousness: "When things are done, they are finished, so hurry to do the next thing, and all records and data are a waste of time". This concept exists not only in managers at all levels, but also in the minds of bosses. The boss doesn't want employees to waste time, but wants them to concentrate on doing effective things. If bosses don't know the value of data and data analysis, then enterprises will not accumulate data, will not precipitate analysis models, and will not form "precipitation", so there will be no "history".

The development of things in the real world is very complicated and influenced by various factors. When we begin to know things, we often feel at a loss because we know too little. For entrepreneurship, why do some people succeed and others fail? For a marketing activity, why are some very effective and some simply ineffective? For these problems, it is because our cognition and summary are not enough. Many people rely on cleverness and cleverness to control more factors, which makes a marketing activity popular. And some people don't sum up and accumulate relevant experience and data, even if they succeed this time, they may not succeed next time.

Things are very complicated, even if you master many laws, you can only improve the probability of success, but you can't guarantee success. P&G has accumulated 180 years of experience, and the success rate of a new product is less than 70%. But for the fast-moving consumer goods industry, the success rate of 70% is an astronomical figure, the success rate of conventional enterprises is generally lower than 20%, and the success rate of clothing products is even lower.

Now that we have entered the era of big data, we have more powerful data acquisition and data processing capabilities, and more intelligent terminal devices can automatically collect data. With the mobile internet and wearable devices, we have enough conditions to use data to understand the laws of things and control the development of things. The world is becoming more and more complex, and for those who have data and can use it, the world will become simpler and simpler.

Internet has changed the way of information dissemination, and also brought new ways of transaction and resource utilization. And big data is the way of information processing, which will change our way of thinking, thus creating more wisdom to understand things, so we should master data, use data, apply data, and grasp the law of things' development.

Scientific abstraction is the second core idea of data analysis (one of the core ideas is the causal thinking introduced earlier). When we study things, we need to abstract the development of things and study things with scientific thinking methods, so as to establish a mathematical model. Instead of considering that things are too complicated and making yourself an obstacle to action. One of the most basic qualities that a data analyst must possess is to consider "possibility" and abstract things scientifically. For example, abstract the causal relationship of things into a functional relationship, get this functional relationship through data collection and calculation, and then verify the effectiveness of this functional relationship.

The essential difference between sociology, economics, management and natural science is that when studying things, more mistakes, more deviations and more "unknowns" will be tolerated. The relationship between advertising and sales will never be accurate, and there will never be a stable mathematical relationship. There must be other unknown factors, but this is not an excuse for us not to study the relationship between advertising and sales, but the driving force for us to study the relationship constantly.

The full text is taken from Zhao Xingfeng's Business Data Analysis-Ideas, Methods, Applications and Tools.

The next issue is more practical!