Shulou(Shulou.com)06/01 Report--

This article mainly introduces "how Facebook analyzes big data". In daily operation, I believe many people have doubts about how Facebook analyzes big data. The editor consulted all kinds of materials and sorted out simple and easy-to-use methods of operation. I hope it will be helpful for you to answer the doubts of "how Facebook analyzes big data!" Next, please follow the editor to study!

First of all, Facebook is big data.

I used to have a point of view, or dissenting opinions, which was very difficult to communicate with my Internet counterparts. They always think that Facebook is first of all SNS;, but I think Facebook is first of all data. I pointed out: "the domestic industry always treats Facebook as a SNS. In fact, this is a relatively amateur view. Facebook is indeed a SNS, but its real core competitiveness is in the data core business."

My unusual view rarely resonates with project managers, especially those who boast the concept of SNS. This time, I finally met my bosom friend with the investors. I think the view of the capital man is closer to the reality of Facebook than that of the project manager.

First of all, capital people have a broader vision than project managers and are longer at looking at Facebook in the context of the times, rather than the simple business point of view of project managers.

Capital people put Facebook into two big contexts:

1. Realize in the judgment of the times that the value mining of social network (SNS) leads the Internet into the era of big data and promotes the development of "big data" industry.

2. In the judgment of the industrial chain, we realize that the social network represented by Facebook is the first to enter the era of big data, which will further lead the big data application in other Internet fields, and the mining of user value will drive the development of the "big data" industrial chain.

Although the project manager noticed Facebook's SNS, he didn't think much about what SNS would do on a larger scale, so he took SNS as the goal itself.

Secondly, capitalists are longer than project managers to see the nature of data through the phenomenon of SNS.

Capital people put the "special" of Facebook in the "general":

1. Capital people realize that SNS is only a special case of big data on the collection side: big data refers to "massive data + complex data types", while social network (SNS) is generating huge amounts of unstructured data (text, applications, location information, pictures, music, video, etc.) every second, which is a typical "big data" system.

2. SNS is only an application of big data: the core of "big data" lies in the multi-faceted value produced by data mining and application. Social networking (SNS) value mining itself is an important application of "big data" and business intelligence applications.

3. Facebook represents the transformation of one-to-one consumption-driven model: Facebook user data contains huge business value. Users' comments, uploaded pictures, music and videos are all typical unstructured data, in which users' propensity to consume. Mining and analysis of "data" can greatly improve the accurate effect of advertising, help Facebook to develop applications that are more attractive to users, and can predict the development trend of many industries through user behavior, which contains huge commercial value.

In contrast, the project manager's understanding of Facebook focuses too much on functional details that have nothing to do with the transformation, and less on the meaning of the Facebook business model in the transformation.

For Zuckerberg, even the capitalist can't read his mind. Capital only interprets what Facebook is at the level of value. Zuckerberg has repeatedly stressed that the purpose of Facebook is not to become a company, but to understand the mission of Facebook on a meaningful level.

It's not what Facebook will be after SNS.

Facebook is far from being a gift of SNS,SNS, which can be said to be a gift of Facebook. For endowments, SNS is even one of many possible endowments. Even business analysts can see that. John Battelle of Federated Media sees that "some of the transformation that Facebook is making is different from what we had predicted in the past. The company is trying to redefine itself from being a social networking site in a narrow sense, and that is exactly what the outside world understands about it."

1. Become the operating system for people's lives.

One of Bartley's favorite new directions is the living operating system.

Mr Bartley says all companies are scrambling to become the operating system in which people live, and they want to be a central place where users participate and store all their data, and then they need to take advantage of those hubs for everything they do. Zuckerberg himself said that the world's information infrastructure should be similar to the social graph.

Zuckerberg's vision has gone beyond socialization, from being socially inspired to a "world information infrastructure" with a similar map. This statement is not as good as Bartley's "becoming the operating system of people's lives". In fact, Zuckerberg's original meaning may be closer to that, because he says: the more people share, the more information they can get about products and services through people they trust. They can find the best products more easily and improve their quality of life and efficiency. Clearly emphasizing the meaning of life.

"World Information Infrastructure" and "living operating system" may be regarded as the prototype architecture behind big data. In comparison, the former focuses more on grasping the overall structure of data from the object, while the latter focuses more on the overall structure of data grasped by the subject. For big data, what is the living operating system? This refers to the reconstruction of life with meaning, and data is just the material used to reconstruct the meaning. As soon as the meaning of the subject is focused, the data can be divided into good and bad: the data eventually tend to be meaningful and become wise, and finally deviate from meaning and become stupid. Therefore, whether it is a smart earth or a smart city, it is not a big accumulation of data, but it is more meaningful because it is oriented to life.

The structure of industrial society does not focus on meaning, but on value. The relationship between value and meaning is the relationship between means and ends. But having value is not necessarily meaningful, for example, having money is valuable, happiness is meaning, but having money is not necessarily happy. In other words, master the means to achieve happiness, but can not achieve the goal of happiness. The human infrastructure of industrial society, related to value, is fully socialized and extremely professional, but those related to meaning are in a state of small-scale production and extremely amateur. This makes an industrial society imperfect and makes it easy to become a society that systematically forgets its purpose and purpose for the sake of highly developed means.

Big data's mission, not from the perspective of technology, but from a human point of view, is to establish a professional and socialized focus system between means and ends, so as to systemically make things not deviate from its purpose. so that the human meaning system, which is at the level of small farmers under the condition of industrialization, has become a highly developed overall social structure.

What will be the relationship between SNS and living operating systems? SNS is just a fishing net. It is the intention to build a life operating system that focuses on the meaning of life and catches the net of meaning of life. Now people who imitate Facebook are attracted by the fishing net, which is a data collector for daily life. They build some identical fish nets and imitate the action of Facebook throwing the net, but they do not know that the action is fishing. As a result, one net does not fish, the second net does not fish, and finally only meets some small tail fish. As everyone knows, in addition to fishing with fish nets, there are also various means of fishing, such as forking, fishing, and electric fish. there are many data collectors like SNS, such as LBS, O2O, payment, and even offline POS machines. If Facebook is not SNS one day, it must be found that there are other fishing methods that can catch more fish. Fishing, here is a metaphor for meaning or the core value of the enterprise; fishing means, here is a metaphor for endowment (usually said, what do you do, which business). If the enterprise wants the foundation industry to be evergreen, it is necessary to let the endowment change with the change of the environment while maintaining the core value.

2. "reshaping the structure": an inverted customer-centered economy

Zuckerberg proposes an important principle of "reshaping architecture": we want to reshape the way information is transmitted and consumed by helping people build relationships. We believe that the world information infrastructure should be similar to the social graph-it is a bottom-up peer-to-peer network rather than the current top-down monolithic structure. In addition, letting people decide what to share is a basic principle of reshaping the architecture.

This basic principle of reshaping the structure is actually of guiding significance to big data's structure.

Without the guidance of principles, big data is likely to be structurally opposite: he still uses the traditional structure of transmitting value from producers to consumers, but serves the old traditions with new technologies. Although this kind of service is also necessary, it is not the positioning of big data by Facebook.

We read the following meaning from Zuckerberg's words:

1) reshaping the structure means that big data will invert the economic structure

First, "We hope to reshape the way information is disseminated and consumed by helping people build relationships." Reshaping is inversion, and the so-called inversion is to reverse the direction of value generation, originally from producer to consumer, and now from consumer to producer. This is the first meaning from top to bottom and from bottom to top.

In the inverted economic structure of SNS and search engines, value is generated not from producers to consumers, but vice versa, from consumers to producers. First of all, consumers expose (actively "produce") consumption intention information in SNS and search engines, and enter the exchange to form abstract consumption value; second, big data processes and increases the value of consumption information, which is equivalent to capitalizing consumption, so that consumer sovereignty can get the same residual treatment as capital.

Second, "it is a bottom-up peer-to-peer network, not the current top-down monomer structure."

Traditional economics and economics have an important asymmetry in the relationship between production and consumption. To create value in the order of "from production to consumption", producers first change the specific value into abstract value (exchange value) in the link of commodity and exchange; second, the general value is magnified through the capital mechanism. However, consumers do not have such rights and power, first, they can not transform the specific value of consumers into abstract value, and second, they can not add value to this value, that is, there is no capitalization process of consumption.

As soon as big data walks out of the original stage of talking about technology (2012-2014) and enters the medieval stage of big data's integration with the economy (about after 2015), people will find that bottom-up involves not only the transmission of information, but also the transformation of the way of value generation. Into the process of empowering consumers through big data. I suggest you read the two books on empowerment in the opposite direction, Public Trends and Innovation promoters, to understand the changes in economic life caused by such empowerment.

Third, "let people decide what to share".

An important concept is mentioned here: "autonomy". Under the industrial economic structure, the most important step for people to lose their autonomy is to alienate independent labor into labor force. therefore, in informatization, the first mention of people's autonomy through information is to return the factors of human nature to the labor force. to form the effect of "human nature + labor = independent labor".

Due to historical limitations, Zuckerberg only vaguely feels that he is reshaping consumption patterns, and the boys and girls who will replace Facebook in the future will need to deepen this information-sharing process into a consumption capitalization process. This will be the subject of big data's next stage (after 2018).

At that stage, people will begin to think generally about the future problem that Haier solved a few years ago: to solve the problem of data generated by consumers and inversely determine the strategic structure of production, especially capital, through the "single person in one" direct economic model. Now the American Institute of Management Accountants (IMA) secretly keeps an eye on Haier's strategic income statement, which is named ZEUS (Zeus), which may reveal big data's strategic secret channel during the capitalization period. I also discussed its decisive impact on BI with Hu Jiansheng, the pioneer of information enterprises, yesterday. At that stage of Facebook, if there is no further progress, his life will be worrying.

2) reshaping the structure means the inversion of value and meaning.

The value structure of the industrial society is from value to meaning, people first produce around the means, and then use the purpose to correct the means; the value structure of the information society is from meaning to value, locate the meaning through SNS and search engine, and then do something valuable according to the meaning.

In the past, in industrial society, the grasp of meaning depended on the way of small farmers. Big data wants to expand the meaning into a system with data structure. In the study of meaning, the task to be accomplished by this structure is called "interpretation of meaning". This is a kind of mind reading. Big data's architecture, from the perspective of subject meaning, should be a mind-reading system, a large-scale system that professionally solves the riddle of the human Sphinx. Through such a living operating system, human beings can be promoted from just valuable to not only valuable, but also meaningful. So that human beings can get higher satisfaction because of meaning.

For enterprises, the same is true. With regard to the direction from meaning to value, Zuckerberg pointed out that in the process, companies gain the benefit of being able to make better products-people-oriented personalized products. In addition to making better products, a more open world will encourage companies to interact directly and reliably with customers.

Here, people-oriented and personalization all refer to the meaning; what is emphasized is to go beyond the intermediate link of value to realize the "direct and reliable interaction" between producers and consumers. In the words of Zhang Ruimin, it is the unity of a single person.

Meaning needs to be explained, and interpretation must be realized through the cycle of meaning.

In Zuckerberg's words: it is a bottom-up peer-to-peer network, not the current top-down monolithic structure. In addition, let people decide what to share. Meaning is not given by the producer (equivalent to the author), but by the consumer (equivalent to the reader, that is, the recipient of the product). Big data system realizes the unity of value and meaning and the unity of means and ends through the cycle of meaning between producers and consumers.

On the other hand, there is another aspect of big data's structuralization, which is to get through the three aspects of meaning: form, semantics and pragmatics. The significance of the mystery of the Sphinx is understandable and indescribable. Through the data collection mechanism such as SNS, the language industry of meaning is formed; then the semantic industry, that is, the processing industry chain of unstructured data, is formed; finally, the pragmatic industry is formed, which anchors the data mining with the specific situation of a person by means of LBS, payment and so on. Only in this way can we decipher the meaning of individuation and experience at the language level. From an artificial intelligence point of view, Facebook's data computing model has unique advantages. It is human computing, rather than man-machine computing like Google (Weibo). Human computing, which is equivalent to in dialogue, people are search engines for each other, forming ecological computing power. There is still great potential for development in this area.

3. The internal exploration of big data's productivity engine.

Big data, as the productivity engine of the new era, studying its productivity characteristics is a basic lesson for understanding the future business frenzy. In big data's era, people who had no sense of technology were likely to become corpses dragged around by galloping productivity chariots.

Even capital people who know nothing about technology have noticed typical problems such as "massive data + complex data types" and unstructured data in the structure of Facebook big data. In fact, this does not involve many fundamental issues such as Hadoop, NoSQL, data analysis and mining, data warehouse, business intelligence and open source cloud computing architecture.

Big data's general technical process is to use SNS, search engine, POS and other collectors to collect massive data into the data warehouse, and then use the distributed technical framework (Hadoop) to carry out heterogeneous processing (NoSQL) of non-relational data, and develop one-to-one business intelligence through data analysis and mining. As the problem of big data is quite complicated, I have some personal ideas now, but I will not mislead you until I consider it mature. Let's first follow the practice and knowledge of Facebook and sum up from the bottom up.

Facebook is also one of the prominent protagonists in big data's business. It is praised by big data insiders in terms of low-cost integration of huge amounts of data. But at present, in my opinion, the big data strategy of Facebook has not been completely finalized, and it mainly focuses on internal data management.

Data analysis: structure before mining

Is it valuable to collect a large amount of data? People are bullish on Facebook because it doesn't just bring users together. As some scholars have judged: "the efforts of Facebook in the past few years have established connections and ties for nearly 1 billion digital immigrants, and the boundaries of the world still need to expand, and the next more important step is to consider how to make the huge amount of data generated by relationships more valuable."

Now, Facebook will collect 500+TB data every day, if there is no data analysis, Facebook in the global big data industry chain, at best is a miner, coal digger. Only with the key step-data analysis, can it really realize the transformation to the role of "ore processing", classify and refine the collected data, and give full play to the real value of the data.

To deal with this data information, the first hurdle that Facebook has to go through is classification. The fragmented and unstructured data such as user comments, uploaded pictures, music and videos are analyzed in a waterfall to aggregate and classify them into structured data information. Form identity data (basic information for user registration), demand data (explicit information with "like" button, status information, mood information), relational data (through people and fans followed by the user, judge the relationship with other social network users) and other data modules. So when users share, listen to music, click the ubiquitous "like" button on Facebook, and change their status to "engaged", the first thing Facebook needs to do is to sort and structure their clutter.

The second level of data analysis will be more difficult, which is to interpret these structured data and delve deep into the potential meaning behind the data. Every time a user logs in, Facebook,Cookie will always reside in the user's browser, and its browsing behavior and keywords on the page will be recorded. Through the continuous analysis of keywords and uploaded information, Facebook can easily find out the user's long-term interests and recent needs. Coupled with the analysis of your circle of friends, you can get your education, job, income, geographical location and so on. This kind of excavation and interpretation is often more comprehensive and true than the information you fill in voluntarily.

Guo Wei, vice president of Kaixin, expressed his admiration for the ability of Facebook data mining: "data mining according to the habits of a large number of users, and then 'portrait' of users is one of the most powerful functions of social networks. Compared with other social networks, Facebook's ability to portray users is very strong, which will make it more accurate to grasp the needs of users and advertisers. If you take a sketch as an example, the domestic SNS website may draw a rough picture, but Facebook may be very detailed, including how long the eyelashes are, whether the eyes are gray or blue, what the hairstyle looks like, and then wear a shirt, tie, suit, and beard. "

It is based on such meticulous data mining that Facebook has brought advertisers unimaginable precision: Ms. Boston, who has just announced her "engagement" on her home page, received a push ad for a wedding photo, while a bride-to-be in Mumbai, India, received an ad to marry Sally. When two friends on Facebook are talking about planning to travel to Europe sometime in the future, Facebook will scroll a travel company ad in their right advertising area, which will show the ticket price and departure time of the trip to Europe.

Data application: advertisement, product and user

The application of data in the big data strategy of Facebook has not been completely finalized, mainly focused on advertising marketing, product service and user management.

Self-service Advertising order system based on data Mining

Before introducing Facebook's self-service advertising order system, we first need to understand Facebook's advertising model.

We know that the keyword advertising model of Google's Adword search engine is like this: users are searching for keywords, and if this keyword matches the word that advertisers bid for, its ads will appear. Facebook's model is different. Instead of using keywords to find target consumers, it uses users' basic attributes, fans, and interests to identify potential user groups. The feasibility of this advertising model inevitably requires a strong data system in the background as a support.

Therefore, based on this advertising model, Facebook's advertising order system is basically self-service. Advertising advertisers start with a custom audience, and Facebook will lead customers to set a series of parameters step by step, mainly in three ways: first, screening according to demographic characteristics, that is, the basic attributes of the audience, there are a total of 11 dimensions, including location, age, gender, temperament object, emotional status, language, education, school, workplace and so on. Second, according to the fan page to filter, that is, specific to what kind of relationship. Third, according to interest screening, each user can set their own interests when opening Facebook, including religion, favorite things (travel, movies, reading, etc.), favorite brands and so on.

Next, advertisers need to submit the total budget and daily budget for the advertising campaign. The system will calculate the number of target audience groups according to the audience conditions set by advertisers, and then give the range of suggested fees according to the advertising method (CPM/CPC) chosen by advertisers.

Take LadyGaGa as an example to make a simple explanation: in California, there are about 150000 women who like LadyGaGa, but after joining the UK, the number of women who like LadyGaGa has greatly increased to nearly 2 million. In this way, if advertisers want to target audiences who like performances or singers like LadyGaGa, they will have a rough idea of the size of the audience. Assuming that the product is an energy drink (something like Red Bull), it can be further targeted at Facebook users who like sports (# sport) and adventure (# adventure), and even set the fan pages of energy drinks such as Red Bull as audience conditions.

Because it is connected to the backend data in real time, advertisers can learn about the number of new / reduced fans every day, where they come from, and the basic information of fans on the advertising order system. You can also see the number of people, click-through rate and conversion rate that can be accessed for each ad launch, so that the strategy can be changed at any time. Under the support of the database, this full self-service system has obvious advantages: first, it provides more small and medium-sized enterprise customers who lack advertising agencies with tools to make their own advertising budgets and audience groups; second, through meticulous index selection, it brings advertisers professional and accurate delivery experience; third, it improves the efficiency of advertising operation and saves advertising operating expenses. Fourth, it also avoids artificial service mistakes to keep some customers out of the door.

Using data to optimize product design

The data mining and application of Facebook not only has a strong attraction to advertisers, but also helps the product design team to optimize the content of the website, master the user's use pattern, and optimize the interface interaction and operation.

A small example:

Julie is a member of the Facebook product design team. Through data analysis, we have learned how Facebook makes use of the various functions provided by the website.

Take the image upload function as an example: Julie got the following sets of data: 87% of users selected a type in the album album name prompt box on the first screen, 57% of users turned on the file selection function to select the image they wanted to upload, 52% of users clicked the upload button, and 48% of users waited for the upload progress to be completed.

The data clearly tell the story: less than 50% of users can upload pictures successfully. Therefore, in order to improve the rate of successful uploads of images, Facebook changed the Java/flash facebook file selection function to the browser native file selection function, resulting in an 11% increase in uploads. After the problem had been running for some time, the team analyzed the data and found that 85% of the users who uploaded the picture uploaded only one picture. The team also observed that users did not know how to use shift to select multiple images for upload, so they added a prompt to upload multiple images before the upload started. As a result, the data dropped from 85% to 40%. 

At this point, the study of "how Facebook analyzes big data" is over. I hope to be able to solve your doubts. The collocation of theory and practice can better help you learn, go and try it! If you want to continue to learn more related knowledge, please continue to follow the website, the editor will continue to work hard to bring you more practical articles!

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.