Big data cases 1-4: How Amazon, Google, eBay, Target deploy big data technology

For companies, 100 theories are not as practical as a successful benchmark. The main idea of this paper is to find 49 samples of “doing” big data.

This paper attempts to sort out the general law of the value of big data from the perspective of enterprise operation and management: one is data driven decision-making, mainly to improve the prediction probability to improve the success rate of decision-making; the second is the data driven process, mainly Form a marketing closed-loop strategy to improve the conversion rate of the sales funnel; third, data driven products, in the product design phase, emphasize individualization; in the product operation phase, it emphasizes iterative innovation.

Top Natural Data Company's various packages

From Google, Amazon, Facebook, LinkedIn, to Ali, Baidu, and Tencent, they have become natural big data companies because of their large user registration and operational information. Large technology companies such as IBM, Oracle, EMC, and Hewlett-Packard have invested in big data to provide other companies with "hardware software data" solutions by integrating big data information and applications. Our focus is on the value of big data, and the first category of companies bear the brunt.

Below are typical examples of the data mining value of these natural big data companies.

01 Amazon's "Information Company"

If a company in the world has found the most value from big data, so far, the answer may not be Amazon. Amazon also has to deal with massive amounts of data, and the direct value of these transaction data is even greater. As an "information company", Amazon not only gets information from each user's purchase behavior, but also records every user's behavior on their website: page time, whether the user views the comment, the key to each search Words, products viewed, and more. This high sensitivity and attention to the value of data, as well as its powerful data mining capabilities, has made Amazon far beyond its traditional way of operation.

Amazon CTO Werner Vogels gave a presentation on big data at CeBIT and described the business blueprint of Amazon in the era of big data. For a long time, Amazon has been trying to target customers and get customer feedback through big data analytics. “In the process, you will find that the larger the data, the better the results. Why do some companies make mistakes in business? That's because they don't have enough data to support operations and decisions,” Vogels said. “Once entering the big In the world of data, there is an infinite possibility in the hands of enterprises. From the infrastructure that supports emerging technology companies to the mobile devices that consume content, Amazon's tentacles have reached a wider field.

Amazon recommends: Amazon's business links are inseparable from the "data driven" figure. A friend who bought something on Amazon may be familiar with its recommended function. The recommended function of "People who bought X products and also bought Y products at the same time" looks very simple, but very effective, and these accurate recommendation results The process of drawing is also very complicated.

Amazon Forecast: User demand forecasting is the use of historical data to predict the future needs of users. For books, mobile phones, home appliances, etc. - Amazon's internal products called hard demand, you can think of it as a "standard" - the forecast is relatively accurate, and can even predict the demand for related product attributes. However, for the soft demand products such as clothing, Amazon has not been able to predict very well for more than ten years, because such things are subject to too many interference factors, such as: the user's preference for color styles, put on and fit, Lovers and friends like it or not... This kind of thing is too volatile, but it will sell badly, so it needs a more complicated forecasting model.

Amazon Test: Do you think that a certain page of text on Amazon's website just happened to happen? In fact, Amazon will continuously test new designs on the website to find the solution with the highest conversion rate. The layout, font size, color, buttons, and all other designs of the entire site are in fact the best results after multiple prudent tests.

Amazon Record: Amazon's mobile app gives users a smooth, ubiquitous experience while also collecting insights into each user's preferences by data on their phones; more notable is the Kindle Fire, embedded The Silk browser can record the user's behavior data one by one.

The data-oriented approach is not limited to the above areas, Amazon's corporate culture is a cold, data-oriented culture. For Amazon, big data means big sales. The data shows what is valid and what is not, and new business investment projects must be supported by data. The long-term focus on data allows Amazon to offer better service at a lower price.

02 Google’s intentions

If there is a technology company that accurately defines the concept of "big data", it must be Google. According to comScore, a search research company, Google’s search terms were as high as 12.2 billion in March alone in March 2012. Google's size and scale make it have more ways to apply big data than most other companies.

The Google search engine itself is designed to allow it to seamlessly link thousands of servers. If there is more processing or storage needs, or if a server crashes, Google engineers can easily get more servers by adding more servers. The result of bringing all of this data together is that companies not only benefit from the best technology, they also benefit from the best information. Here are three of the highlights of Google Inc.

Google's intent: Google not only stores the network connections that appear in the search results, but also stores the user's search keyword behavior. It can accurately record the time, content and method of people's search behavior, and people search on Google's website. And a large amount of machine data generated when passing through its network. This data allows Google to optimize ad sorting and turn search traffic into a profit model. Google can not only track people's search behavior, but also predict what the searcher will do next. Every search request entered by the user will let Google know what he is looking for. All human behavior will leave a trail on the Internet. Google has taken a good position to capture and analyze the path. In other words, Google can predict your intentions before you realize what you are looking for. This ability to capture, store, and analyze massive amounts of human-machine data and then make predictions based on it is a data driven product.

Google Analytics: Google has more ways to get data outside of search. Companies install products like Google Analytics to track the footprint of visitors on their sites, and Google also gets the data. The site also uses Google Ad Network to show ads from Google Advertisers on its site, so Google can not only gain insight into the performance of ads on its site, but also showcase the performance of other ad sites.

Google Trends: Since the search itself is a "intention database" for netizens, it is of course possible to predict the next trend based on the ups and downs of a particular search volume. Google Trends can predict sales of travel, real estate, and automobiles. The most famous of these predictions is the trend of Google flu, tracking the spread of influenza and other diseases worldwide, and analyzing the spread of influenza and other diseases worldwide according to the search of netizens.

03 eBay's analysis platform

As early as 2006, eBay set up a big data analytics platform. In order to accurately analyze the user's shopping behavior, eBay defines more than 500 types of data to track and analyze customer behavior. Oliver Ratzesberger, senior director of eBay Analytics, said: "On this platform, structured data and unstructured data can be combined to promote eBay's business innovation and profit growth through analytics."

eBay Behavior Analysis: In the early days, every feature change on eBay's web pages was usually determined by a product manager who knew the feature very well, based on the personal experience of the product manager. By analyzing the user behavior data, any changes to the functionality on the web page are left to the user. “Every time there is a good idea or idea, we will select a range of users to test on the website. By analyzing the behavior of these users, we can see if this idea has the expected effect.”

eBay Advertising Analytics: More significant changes are reflected in advertising costs. eBay's investment in Internet advertising has been great, introducing potential customers to eBay's website by buying keywords for web search. In order to measure the input and output of these keyword ads, eBay has established a completely closed

04 Target's “Related Data Mining”

Using advanced statistical methods, merchants can build models through user purchase history analysis, predict future purchases, and then design promotions and personalized services to avoid user churn to other competitors. Target, the third-largest retailer in the United States, can “guess” which pregnant women by analyzing all female customer purchase records. It found that female customers would buy a large amount of fragrance-free lotion in about four months of pregnancy. This led to the discovery of 25 items that were highly correlated with pregnancy and produced a “pregnancy prediction” index. After deriving the expected date of delivery, you can take the first step and send the discount coupons such as maternity clothes and baby cots to the customer. Target also created a model for buying changes in women's behavior during pregnancy. Not only that, if users buy baby products from their stores, they will regularly give them according to the baby's growth cycle in the next few years. These customers push related products to create long-term loyalty for these customers.