How to understand the meaning of "data analysis" in Internet industry?

Internet companies have a large amount of online data, and the amount of data is still growing rapidly. In addition to using big data to enhance business, Internet companies have also begun to realize data services and use big data to discover new business value.

Take Alibaba as an example, it is not only constantly strengthening personalized recommendation, a consumer-oriented "thousands of people" big data application, but also trying to use big data for intelligent customer service. This application scenario will gradually extend from internal applications to call centers of many external enterprises.

In the big data application for merchants, taking "business personnel" as an example, more than 6 million merchants are using "business personnel" to improve the operation level of their e-commerce stores. In addition to facing its own ecology, Alibaba's data business is also accelerating. Sesame Credit, an application of personal credit evaluation based on collected personal data, has made great progress, and its application scenarios have extended from Alibaba to more and more external scenarios, such as car rental, hotels and visas.

Because all the behaviors of customers will leave traces on the Internet platform, Internet companies can easily obtain a lot of customer behavior information. The information generated by the Internet commerce platform is generally true and definite. Analyzing these data through big data technology can help enterprises to formulate targeted service strategies and gain greater benefits. Practice in recent years has proved that the rational use of big data technology can improve the business efficiency of e-commerce by more than 60%.

In the past few years, big data has changed the face of e-commerce. Specifically, the application of big data in the e-commerce industry has the following aspects: precision marketing, personalized service, and personalized product recommendation.

1. Precision Marketing

Internet companies use big data technology to collect all kinds of data of customers, establish a "user portrait" through big data analysis, and describe a user's information in an abstract way, thus personalized recommendation, precise marketing and advertising for users.

When the user logs on to the website, the system can predict why the user came today, and then find out the appropriate products from the commodity library and recommend them to him. Figure 1 shows what basic information and characteristics of the user will be included in the user portrait.

Figure 1 user portrait

The core of marketing supported by big data is to push the business of the enterprise to the users who need this business most at the right time, through the right carrier and in the right way.

First of all, big data marketing has a strong timeliness. In the Internet age, users' consumption behavior can easily change in a short time. Big data marketing can implement marketing strategies in time when users' demand is strongest.

Secondly, personalized and differentiated marketing can be implemented. Big data marketing can realize one-to-one marketing of segmented users according to users' hobbies and needs at a certain point in time, so that the marketing of merchants can be targeted and the marketing strategy can be adjusted in time according to real-time effect feedback.

Finally, big data marketing can analyze the relevance of target user information. Big data can carry out multidimensional correlation analysis on all kinds of information of users, and find interesting associations and related connections between data sets from a large number of data.

For example, by discovering the relationship between different commodities in the user's shopping basket, other consumption habits of users can be analyzed. By knowing which products users frequently buy at the same time, it helps marketers to discover other consumption rules from a consumer's consumption habits, so as to formulate marketing strategies for related products for this user. Figure 2 shows that the website will recommend different products for different customers according to the user's portrait.

Figure 2 Precision Marketing

For example, an e-commerce platform grasps customers' consumption patterns through their online browsing records and purchase records, so as to analyze and classify customers' consumption-related characteristics. Such as income, family characteristics, buying habits, etc. Finally, grasp the characteristics of customers and judge the products and services they may be concerned about according to these characteristics.

Since consumers entered the website, the website has deployed five recommendation columns with different algorithms on four pages, such as list page, single item page and shopping cart page, to recommend the products they are interested in, thus improving the product exposure and promoting cross-selling and up-selling. After comprehensive optimization of the website from multiple angles, the conversion rate of shopping mall orders increased by 66.7%, the conversion rate of shopping mall orders increased by 18%, and the total sales volume increased by 46%.

In Wal-Mart stores in the United States, after the cashier scans the goods purchased by customers, some additional information will be displayed on the POS machine, and then the salesperson will remind customers what other goods can be purchased according to this information. The "consulting marketing" system supported by Wal-Mart's big data system can establish a forecasting model. For example, if a customer has a lot of beer, red wine and salad in his shopping cart, 80% of them may need to buy side dishes with wine and seasoning.

2. Personalized service

E-commerce has the inherent advantage of providing personalized service, which can obtain users' online records in real time through technical support and provide customized services in time.

Many e-commerce companies try to provide users with comprehensive personalized product recommendations on their homepages by relying on data analysis. Haier and Tmall provide users with the function of customizing TV through the Internet. Customers can choose the attributes such as size, border, clarity, energy consumption, color and interface before TV production, and then the manufacturer will organize the production and deliver it to the door. This personalized service has been widely welcomed.

Similar customized services are now used in industries such as air conditioning and clothing. By meeting individual needs, these industries enable customers to obtain more satisfactory products and services, thus shortening the cycle of design, production, transportation and sales and improving the efficiency of commercial operation.

In order to provide users with ideal personalized service, enterprises should first fully understand the user's personality through data, and then reasonably control and design the service personality. Understanding users' personality is the basis of providing users with the products and services they want. Enterprises need to find the most valuable data in a huge database, then cluster users through data mining, and then design targeted services according to the characteristics of user types.

Personalized decentralization units can be large or small, as large as a customer group with the same needs, and as small as each user is a personalized demand unit. Enterprises must master the granularity of personalized service. Individualized service that is too scattered will increase the service cost and management complexity of enterprises, and the increased personalized cost is directly proportional to the actual income demand.

Figure 3 provides a personalized travel service.

Ctrip's big data application analyzes the data of all Ctrip users from the user's perspective, including the data generated by users' behaviors before and after a series of trips, such as inquiries, browsing, reservations, trips, comments, etc. Ctrip ensures the authenticity of the data left by users while eliminating invalid data, and then screens, sorts and reorganizes a large number of data in real time, and applies them to the personalized needs of users before, during and after travel, as shown in Figure 3.

To be personalized, it is very important to clarify the user's target needs, not only to look at the order, but also to care about what the user cares about. For example, when booking a five-star hotel, some users are very sensitive to hotel facilities, some value hotel location, and some care more about hotel services. For this, Ctrip will recommend different hotels according to the needs of users.

Target department store in the United States set up a baby bath registration form to model and analyze the customer's consumption data in the registration form. They found that many pregnant women will buy a lot of odorless hand cream in the early stage of their second pregnancy, and buy a lot of health products such as calcium and zinc supplements in the first 20 weeks of pregnancy.

Target finally selected the consumption data of 25 typical commodities and constructed the "pregnancy prediction index". Through this prediction index, Target can predict the pregnancy situation of customers within a small error range, so as to send preferential advertisements for pregnant women to customers at the right time.

"Nike running shoes or wristband sensors" have gradually made Nike an innovative company in big data marketing. As long as athletes wear Nike running shoes to exercise, the related iPod can store and display data such as exercise date, time, distance and calorie consumption value.

Nike has mastered the database of the best running routes in major cities through the running routes uploaded by runners, and better organized the running activities in various cities. At present, Nike's sports online community has more than 5 million active users uploading data every day, and Nike has established an unprecedented strong relationship with consumers. At the same time, massive data has also played an irreplaceable role for Nike to understand user habits, product improvement, precise delivery and precise marketing. Nike even mastered what songs runners like to listen to best. Personalized service is inseparable from the active participation and sharing of customers, and the data from customers can also serve customers more accurately.

The rapid development of "Three Squirrels" in recent years, on the one hand, relies on brand promotion, on the other hand, on the basis of data analysis, constantly improves the details, including personalized names, cartoon images of "Three Squirrels", differentiation of gifts, classification of different customer labels and user experience. "Three Squirrels" can know all customers' purchase records in the mall through ERP system, and can accurately capture user comments through CRM system. Some casual messages and ratings will reflect their needs.

By analyzing customers' previous buying habits in shopping malls and users' purchasing evaluation, we can judge which products of different tastes sell best in which area and which products are most acceptable to consumers, so as to make more targeted product home page recommendations. At the same time, they will carry out personalized and humanized label classification and detailed analysis on customers, so as to push different product types according to these classifications. For example, the products purchased by the wife-loving customers are mainly eaten by the wife. The "three squirrels" will put letters in the package and write a letter to their wives in the voice of "squirrels".

3. Personalized recommendation of goods

With the continuous expansion of the scale of e-commerce and the rapid growth of the number and types of goods, it takes customers a lot of time to find the goods they want to buy.

Personalized recommendation system analyzes users' behaviors, including feedback, purchase records and social data. , analyze and mine the correlation between customers and products, so as to find the personalized needs and interests of users, and then recommend the information and products that users are interested in.

Personalized recommendation system can recommend products according to users' characteristics and interests, which can effectively improve the service ability of e-commerce system and retain customers.

1) e-commerce website

With the vigorous development of e-commerce, the dominant position of recommendation system in the Internet is becoming more and more obvious.

Internationally, the recommendation algorithm adopted by Amazon platform is considered to be very successful. In China, relatively large e-commerce platform websites include Taobao (including Tmall Mall), JD.COM Mall, Dangdang and Suning.cn.

In these e-commerce platforms, the number of goods provided by the website is countless, and the scale of users of the website is also very large. According to incomplete statistics, the number of goods in Tmall Mall has exceeded 40 million.

In such a huge e-commerce website, users will get many similar results after entering keyword queries according to their purchase intentions. It is also difficult for users to distinguish the similarities and differences between these results, and it is difficult to choose a suitable project. The recommendation system can recommend some products that users are interested in according to their interests. E-commerce websites use recommendation system to recommend products for users, which is convenient for users and thus improves the sales of websites.

2) Movie and video websites

Personalized recommendation system is also widely used in movie and video websites, which can help users find their interested videos in the vast video library. One company that successfully uses recommendation system in this field is Netflix.

Netflix was originally a DVD rental website, and later began to set foot in online video business. Netflix attaches great importance to personalized recommendation technology, and has held the famous NetflixPrize recommendation system competition since 2006, hoping that researchers can improve the prediction accuracy of Netflix recommendation algorithm by 10%.

The contest has played an important role in promoting the development of recommendation system: on the one hand, the contest provided a large-scale user behavior data set in the actual system for the academic community (400,000 users rated hundreds of millions of records of 20,000 movies); On the other hand, in the three-year competition, the contestants put forward many recommendation algorithms, which greatly reduced the prediction error of the recommendation system.

Figure 4 is the movie recommendation interface of Netflix, including the movie title and poster, user feedback and recommendation reasons. Netflix uses a project-based recommendation algorithm, that is, it recommends movies similar to those they once liked to users. Netflix claims that 60% of users find interesting movies and videos through its recommendation system.

Figure 4Netflix movie recommendation

As the largest video website in the United States, YouTube has a large number of video content uploaded by users. In order to solve the problem of information overload in video library, YouTube has also conducted in-depth research in the field of personalized recommendation, and now it also uses project-based recommendation algorithm. Experiments show that the click rate of YouTube personalized recommendation is twice as high as that of popular videos.

3) Internet radio station

Personalized Internet radio stations are also suitable for personalized recommendation. First of all, there is a lot of music, so it is impossible for users to listen to all the music before deciding what they like to listen to. Moreover, new songs are increasing rapidly every year, and users are undoubtedly facing the problem of information overload. Secondly, when people listen to music, they usually listen to music as background music, and few people have to listen to a certain song. For ordinary users, you can listen to any song as long as it is in line with their mood at that time. Therefore, personalized music network radio is a product that is very in line with personalized recommendation technology.

At present, there are many well-known personalized music network stations. And Pandora and? Last.fm | Playing music, finding songs, discoverer, Douban Radio is the representative of China. These three personalized network radio stations do not allow users to order songs, but give users several feedback methods: like, dislike and skip. After a certain period of user feedback, the radio station can obtain the user's interest model from the user's historical behavior, so that the user's playlist is more and more in line with the user's interest in songs.

Pandora's algorithm is mainly based on content. Its musicians and researchers personally listen to tens of thousands of songs by different singers, and then mark the different characteristics of the songs (such as melody, rhythm, arrangement and lyrics, etc.). ). These tags are called music genes. Then, Pandora will calculate the similarity of songs according to the genes marked by experts, and recommend other music similar to his previous favorite music genes to users.

Last.fm | Play music, find songs and discover artists? Record all users' listening records and users' feedback on songs, and calculate the similarity of different users' preferences for songs on this basis, so as to recommend songs to other users with similar listening hobbies. Meanwhile, last.fm | plays music, finds songs and discovers artists? A social network has also been established to enable users to establish contact with other users and recommend their favorite songs to their friends. Last.fm | Play music, find songs and discover artists? We don't use expert annotation, but mainly use user behavior to calculate the similarity of songs.

4) Social networking

Personalized recommendation technology in social networks is mainly applied in three aspects: personalized project recommendation to users by using their social network information, conversation recommendation of information flow and recommendation of friends to users.

Facebook keeps two kinds of most valuable data: one is the social network relationship between users, and the other is the user's preference information.

Facebook has launched a recommendation API called InstantPersonalization, which can recommend friends' favorite items to users according to their favorite information. Many websites use Facebook's recommendation API to personalize their websites.

Clicker, a famous TV drama recommendation website, uses InstantPersonalization to recommend personalized videos to users. Clicker can now use Facebook's user behavior data to provide personalized "content streams" that users may be interested in. More importantly, users don't need to input too much data on Clicker's website (by rating, commenting or watching? Clicker.com？ Wait on the video. ), Clicker can provide such a service.

In addition to using users' social network information on social networking sites, social networking sites themselves will also use social networks to recommend conversations of other users on social networking sites. Every user can see and comment on all kinds of sharing of friends on Facebook's personal homepage. Every share and all its comments are called a conversation, and Facebook has developed an EdgeRank algorithm to sort these conversations so that users can try to see the latest conversations of familiar friends.

Social networking sites not only recommend content to users according to their social networks and user behaviors, but also recommend friends to users through personalized recommendation services.

5) Other applications

Because e-commerce enterprises have basically realized the dataization of all aspects of business processes, they can make full use of big data technology to mine and analyze these data in order to optimize their business processes and improve business profits. In addition to the applications mentioned above, big data can be applied in many other aspects in the e-commerce industry.

① Dynamic pricing and special price

E-commerce enterprises can use data to establish customer data, understand how much money users like to spend and what products they like to buy, so as to make flexible pricing and discount policies by tracking customer consumption behavior and using big data analysis. For example, if the analysis shows that users' interest in a particular category of goods has greatly increased, e-commerce companies can offer discounts or buy one for one.

② Customized discount

E-commerce companies can use data to determine customers' buying habits and send targeted special prices and discount codes to customers according to previous buying methods. When customers stop buying or just take a look, data can also be used to attract customers again, for example, by sending an email to remind customers of the products they have seen or inviting them to complete the purchase.

③ Supply chain management

E-commerce companies can use big data to manage the supply chain more effectively. Data analysis can reveal any delay or potential inventory problems in the supply chain. If an item has a problem, it can be deleted from the sales immediately to avoid damaging the customer service problem.

④ Predictive analysis

Predictive analysis refers to the use of big data technology to analyze various channels of e-commerce business and help enterprises formulate business plans for future operations. Data analysis may show that there are new purchasing trends or unsalable goods in the online shop department of e-commerce enterprises.

Using this information is helpful to plan the next stage of inventory and set new market targets. It is challenging to keep abreast of the latest trends of e-commerce, but using big data technology can greatly improve the profits of enterprises and help enterprises establish successful forward-looking business. If you don't tap the power of big data, you may miss the opportunity of market success.