Shulou(Shulou.com)06/02 Report--

Everyone is talking about data-driven, such as data-driven management, data-driven operation, or data-driven testing. There is a lot of discussion about data-driven applications, but there is less discussion about the basic principles of data-driven. This paper attempts to trace back to the source. Talk about the basic principles of data-driven: how does data drive? What can be driven?

Before we talk about the principle, let's talk about my personal experience.

Once when my wife came home from Terminal T3 of the Capital Airport, I called a Didi for her. Later, I saw that the bill showed that it was more than 80 yuan and the distance was more than 20 kilometers, while my home was only 7-8 kilometers away from the airport. Didi Premier Taxi's fee was usually more than 40 yuan. Obviously, the driver took a detour. Didi's APP also showed a message, which roughly means: the fare is abnormal. Do you need to appeal?

I clicked on "need to complain", and Didi APP immediately popped up an interface to the effect that you have a good reputation record and accept your complaint. The charge is calculated at 42 yuan (forget the specific figure, anyway, it is at the normal charging level).

At that time, I thought, wow, the function of Didi is awesome, and the experience for customers is great!

Just imagine, taking a taxi before Didi is a completely different situation:

You may not even know that the driver took a detour.

You later found that the driver had made a detour, but you forgot to ask for the ticket when you got off the bus, so you couldn't prove it.

At that time, you found that the driver made a detour and asked the driver for the ticket, but there was only mileage and time on the ticket, but there was no place of departure, destination and route, so you could not prove that the driver made a detour.

On the other hand, Didi has completely changed this situation: take the initiative to remind you whether you need to complain, and immediately after you lodge a complaint, you will make a satisfactory handling!

How did Didi handle complaints so thoughtfully and intelligently? If we have some basic understanding of data and the relationship between data and information, knowledge and artificial intelligence, we can understand the operation mechanism of Didi complaint handling.

The data pyramid can help us understand the relationship between data and information, knowledge and artificial intelligence.

Data itself is meaningless if it cannot be transformed into information and knowledge, but if there is no data, or if there is a lack of data, the generation of information and knowledge will become a source of no water.

If you experience something, write it down.

If you record something, upload it.

If you upload something, share it.

This means that everyone becomes a data collector, processor and sharer.

In the case of Didi, Didi apparently did this: all passengers' car experiences were recorded, uploaded and shared by the system.

Two problems existing in the data level of Enterprises

1. The data is missing:

For example, in an Internet company I work for, they do not have data on customer recommendations (how many customers have recommended products to others). This is a very small example, but the lack of data is a common phenomenon in enterprises, because at present, the data owned by enterprises mainly come from various business systems, such as CRM and ERP, while business systems are designed to complete specific business, and the data are only by-products. It inevitably leads to the lack of some data needed for decision-making.

two。 Invalidity of data collection:

Traditional enterprises attach great importance to data collection, for example, they will require store staff to record information about receiving customers, but the quality of the data recorded is not high and the use is very limited, for two reasons:

First, using pen and paper or Excel to record, the process is more troublesome.

Second, they simply collect, process and share data, but the application of data basically has nothing to do with them, employees just deal with data collection, lack of internal motivation.

Internet companies also have ineffective data collection. For example, an Internet company communicates with potential customers through QQ, and the customer demand information they know is recorded in QQ. If you want to follow up a customer after a period of time, it is often difficult to find relevant information about that customer, and the data recorded in QQ cannot actually be used by the business.

In order to solve these two problems in the data layer, enterprises need to make a unified plan for the data based on the needs of business decision-making: what kind of data do they need? How to collect it? How do you record it?

For example, if the Internet companies mentioned above have plans on what data support potential customers need to develop, it is possible to design a structured requirements communication tool that can effectively record customer demand information. it is also conducive to later data analysis.

Without a unified plan for the data, the enterprise's data is likely to be in a "ROT junk state", that is, Redundant, Obsolete and Trivial.

Information: is organized data, is to process and establish internal relevance to the data for a specific purpose, so as to make the data meaningful. It can answer the questions of who (who), what (what), where (where), and when (when). For business management, the role of information lies in process management and performance evaluation.

In the above example, Didi's system integrates time, departure, destination, route, membership and other information to form a complete passenger ride information, thus realizing the monitoring and management of the driver service process.

In the stage of converting data into information, enterprises have two problems:

1. Lack of effective data analysis tools:

A few large enterprises (such as banks and telecom companies) generally have BI systems that can integrate data from different sources and support online analytical processing and reporting, but many enterprises still rely on Excel for analysis and reporting, such as a medium-sized jewelry company with hundreds of stores. The boss attaches great importance to data and emphasizes speaking with data, and every time he holds a business analysis meeting. The regional manager has to stay up late to use Excel to do all kinds of analysis reports, which is very inefficient and annoys the regional manager: I lead the war, but I need to do so much desk work!

two。 Lack of analytical ability to convert data into information:

Enterprises of a certain size now have a large amount of data. For example, we derive tens of millions of lines of data from various business systems of an Internet company. How to see the connection between the data from these data and organize them into meaningful information is undoubtedly a challenge. Ordinary enterprises do not have talents who know both business and data analysis.

As a result, only a small part of the data owned by the enterprise has been effectively processed and turned into valuable information, while most of the data remains in its original state: just a meaningless objective existence.

Knowledge: the summary and refinement of information. Is based on the relationship between information, summed up the law and methodology, mainly used to answer why (why) and how to do (how) questions, the application in the enterprise includes problem diagnosis, prediction and best practices.

For example, Beijing is hot and rainy in summer. The temperature in August is between 20 and 36 degrees, with an average of 12 days of precipitation. This is the law of Beijing's climate based on years of data. This knowledge has three functions:

1. Problem diagnosis (answer why), for example, this knowledge explains why it rained so much in Beijing in August this year.

2. Forecast: in August next year, the temperature in Beijing is likely to be between 20 and 36 degrees, with an average of 12 days of precipitation.

Best practice: travel to Beijing in August to wear short-sleeved clothes, the weak should wear long sleeves, it is best to bring an umbrella.

Didi should have a knowledge base on how to deal with driver detours in Didi's system, otherwise it would not have handled complaints so intelligently.

Most enterprises do poorly at the knowledge level. Although some enterprises have established knowledge management systems, they do not have an effective mechanism for knowledge generation, application and updating. More enterprises do not have a knowledge management system, these enterprises have a lot of tacit knowledge, for example, there are always some sales experts in enterprises, they can achieve outstanding sales performance based on intuition or experience, they may not be clear, but they know how to select potential customers, know when to follow up, know when to promote, this is the so-called "tacit knowledge", tacit knowledge seems so mysterious. As a result, many managers think that sales experts are born and cannot be copied.

An effective data-driven mechanism will make tacit knowledge explicit and explicit knowledge structured, so that everyone in the enterprise can acquire relevant knowledge for business operations anytime and anywhere.

In the case of sales, it may be impossible to clone a sales master, but from a data point of view, if the sales behavior data of salespeople can be fully recorded and effectively sorted out and summarized, the best practices of sales can be abstracted, so that every salesperson can use these practices in sales. (for those interested in this topic, take a look at Jenny Dearborn's Revolution in sales, which tells stories about how data can help sales managers diagnose problems, predict problems, and summarize sales best practices.)

Artificial intelligence: the autonomous application of information and knowledge by machines

Artificial intelligence is the thinking ability similar to human brain (including learning, reasoning, decision-making, etc.) based on data, information and knowledge.

At the level of information and knowledge, data provides decision support, while in the stage of artificial intelligence, the system imitates human application of information and knowledge to make independent decisions.

The example of a Didi driver detour I experienced was that Didi's system completed a complaint handling process by the system rather than Didi employees based on my ride information and the knowledge base in the Didi system.

In fact, such applications have long been common, and Amazon's famous recommendation mechanism is similar. The system replaces employees and takes the trouble to recommend products that might be of interest to customers.

Many people think that the top of the data pyramid is wisdom rather than artificial intelligence. I have different views on this: wisdom is not based on knowledge, knowledge is not a necessary condition for wisdom, and many eminent monks are very wise, but do not have a lot of knowledge. for example, Liuzu Huineng can not read from childhood, but can hear the sutra and interpret meaning.

The reason is that knowledge comes from experience (data), from human observation of the three-dimensional world, and wisdom can be obtained without experience, but by establishing connections with higher dimensions. (professor Liu Feng of Peking University has a lecture. The name is "Open your high-dimensional wisdom", you can refer to)

On the other hand, artificial intelligence must be based on data, and there is no use for expert algorithms or deep learning without data. with data, it is possible for computers to form knowledge through expert algorithms or deep learning. and then have the thinking ability similar to the human mind.

In this sense, artificial intelligence can never surpass human intelligence. From this we can also see the limitations of data: it can exert human rationality to the extreme, but it can only imitate but cannot create, it can not replace human sensibility and intuition, and it is this sensibility and intuition that makes life more interesting and soft, and real creation takes place!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.