Amazon sells 4, products every minute, of which about 5% are presented to users by personalized recommendation engines. When browsing the Amazon website, the algorithm will predict what you want at this moment, and select a group from about 353 million products to push to you.
what drives personality recommendation is Amazon's evolving purchasing atlas, that is, the "entity elements" in reality-all store information such as customers, products, purchases, activities and store locations-and the digital presentation of the relationship between these elements. Amazon's purchasing atlas links the purchasing history with website browsing, Prime Video viewing, Amazon music listening and data from Alexa devices. The algorithm uses collaborative filtering, combining diversity (the degree of dissimilarity of recommended products), surprise (the amazing degree of recommended products) and novelty (the degree of freshness) to generate the most complex recommendations in the world. With rich data and industry-leading personalized recommendations, Amazon now accounts for 4% of the US e-commerce market, and its closest rival Wal-Mart has a market share of only 7%.
in order to compete with Amazon, Google announced the launch of Shopping Graph in April 221, an AI model for recommending products when users search. There are more than 1 billion people who use Google to search for goods every day, and shopping images connect them with more than 24 billion commodity lists of millions of merchants all over the network. This model is based on Google's unique Knowledge Graph, which captures information about entities and their relationships in the vast network, including structured and unstructured data from Android system, sound and image search, Chrome extension of Google browser, Google Assistant, Google email, Google photos, Google maps, YouTube, Google cloud services and Google payment. Google Shopping Map allows 1.7 million merchants to display related products on Google with simple but interlinked tools, and Google can meet the challenge of Amazon.
Data maps like Amazon and Google rely on product usage data (that is, behavioral data generated when users use platforms or products) to grasp the connections and relationships between enterprises and their customers. The concept of data map comes from social network and graph theory, which defines social map as the presentation of contacts and relationships between people, such as friends, colleagues, bosses, etc. Everyone is presented as a node, and relationships are the connections between points. This concept comes from the work of social psychologist Stanley Milgram. In the past two decades, this concept has provided a practical lens for analyzing the structure and dynamics of organizations, industries, markets and society. In 27, Facebook launched a social platform of the same name, which allowed developers to create applications and integrate them into the information flow and interpersonal connections of websites, making digital social maps popular.
leading technology companies use data maps to provide personalized recommendations, upgrade products, optimize advertisements and so on. The most successful examples, such as Amazon's purchasing map, Google's search map, Facebook's social map, Netflix's movie map, Spotify's music map, Airbnb's travel map, Uber's travel map and LinkedIn's career map, use the constantly collected user usage data and unique algorithms to get rid of competitors from product development to user experience.
this paper discusses how enterprises can learn from the methods of leading enterprises in data mapping to create new competitive advantages.
data network effect
to understand the data map, we must first understand the data network effect, that is, the effect that the data generated when users use a product or service makes this product or service more valuable to other users. Different from the direct network effect (such as Facebook and LinkedIn), the data network effect does not need to increase the number of users to enhance the network value, but the existing users continue to use and generate more extensive and in-depth usage data, so that the algorithm can produce continuous improvement results. For example, Google's 2 trillion searches each year help Google enrich its knowledge map, improve its search engine and provide users with better search results. And if users no longer use the platform, the improvement of platform service quality will stagnate and become less helpful.
The data map is not static, and it reflects not the data at a certain point in time, but what data scientists call dynamic data. This is part of the reason why it is impossible to draw the data map manually. It is necessary to use technology to collect and interpret millions of data generated by a company's products used by consumers all over the world in real time.
success factors of data map
Data map leads enterprises to collect user behavior data and quickly use it to improve all aspects of products and services. These companies constantly modify the method of classifying and labeling product data, looking for the relationship between entities, so that the algorithm can better classify and provide personalized recommendations. The company also constantly updates its algorithm to generate personalized recommendations based on the latest and most relevant data to help attract customers. Let's take a look at the key behaviors of companies that successfully use data maps.
learn quickly and extensively. Data map captures personal life, work, entertainment, study, listening, socializing, watching, trading, traveling, consumption and all other activities that can be linked with business. Digitalization enables the company to observe and sort out these customer data extensively, thoroughly and quickly. For example, Facebook's social graph analyzes the data of 2.8 billion people and their social activities all the time: what they are doing, who they are friends with and unfriended, where they have been, what brands they are discussing, what movies they are watching, what music they are listening to, and so on. LinkedIn's career map captures in real time how 774 million professionals who work for 5 million companies and participate in more than 9, educational institutions respond to recruitment information, update their status and use live videos. In addition, the career map also provides users with targeted advertisements, study suggestions, news push and more information according to other factors such as user skills. Now LinkedIn is a subsidiary of Microsoft, and it has been incorporated into Microsoft's data ecosystem, so as to create a more dynamic data map.
the user data of traditional enterprises are stored independently in the databases of different functional departments. In order to gain digital advantage, enterprises must organize data into interactive maps, which can be analyzed by algorithms to generate insights and provide personalized value for each customer.
enrich product lines with data maps. Leading enterprises in data mapping use a series of cross-domain concepts such as shopping, travel or search to organize professional knowledge into a map format that can be recognized by machines. For example, Airbnb's travel map gives a list of more than 7 million houses, which are labeled with attributes (city, landmarks, activities, etc.), characteristics (customer evaluation and business hours, etc.) and their relationships to generate more advanced recommendations, not only recommending rental houses, but also recommending the best dinner places and the best time to visit scenic spots. This ability to expand the product range allows Airbnb to provide customers with better services than traditional hotels, whose data are stored in isolated departments (the reservation department is responsible for booking rooms, the concierge department is responsible for recommending tours, the convalescent department is responsible for booking massages, and so on). Similarly, Netflix is constantly improving the presentation and classification of film and television works in 75, sub-categories, as are Spotify's music and radio programs.
in order to win at a critical moment, Facebook conducted a near-real-time personalized social network content comparison test for 3 billion users. Before pushing the content, Facebook will filter it in the list to be pushed, and according to the user's past behavior, it will narrow the scope to about 5 articles that the user may care about. Facebook will then use a proprietary neural network to rate and sort these contents, and then sort them by media types, such as text, photos, audio and videos with advertisements.
although many companies claim to be customer-centric, few can make good use of data maps and algorithms as leading enterprises. Think about it: does your company use AI algorithm to provide customers with continuously improved products so that they will not turn to other companies?
get started
if you want to compete with the leading enterprises in data mapping, you must understand one thing: the success of the strategy depends not only on whether you have a lot of information, but also on collecting relevant product usage data in real time to realize the data network effect and create advantages. If we can observe the interaction between more users and products, enterprises can get richer data; By selling more products to more diverse user groups, we can accumulate more diverse data and help achieve product differentiation. Companies that are not good at using data maps can refer to the following suggestions for improvement:
1. Develop a data map strategy. First of all, we should let the executives who know the industry cooperate with data scientists to build a data map conceptually, examine the future trend and think about the possible business impact. Many companies with less resources than Amazon or Netflix have already done this. For example, Stitch Fix, a personalized fashion service company founded by a business school student in 21, now has a market value of more than $1.6 billion, largely because of its fashion map.
think about whether the data owned by our company can provide unique advantages. You may have a proprietary data collection method that can obtain detailed information that other enterprises cannot obtain. Maybe you have an advantage in the depth and breadth of data, and you can get complementary data from your partners. Your mobile data (relative to the scattered data used by competitors for batch processing) may be faster. Think about whether we can improve the data range, depth and speed of our company through acquisitions (such as Microsoft's acquisition of LinkedIn and Activision) and alliances (such as cooperation between Google and Shopify).
2. Establish a proprietary algorithm. It is no longer enough to conduct different types of analysis independently. Leading enterprises in data mapping use proprietary algorithms to conduct descriptive analysis under the overall framework ("What happened?" ), diagnostic analysis ("Why?" ), predictive analysis ("What will happen?" ) and normative analysis ("What should happen?" ）。 Your data mapping infrastructure can change from the traditional structure for analyzing static data (batch processing and independent analysis) to analyzing changing real-time data. Reference should be made to other enterprises in the industry and other similar algorithms. For example, if your success indicator is the degree to which customers accept recommendations, how does your recommendation engine compare with leading companies such as Netflix, Spotify and Amazon?
3. Build trust. Managing customer data is a great responsibility. Most customers regard computers, algorithms and machine learning as complex black boxes, and many people think that digital companies use and even abuse their personal data to make a fortune. Enterprises must use algorithms in a trustworthy way, and must obtain permission to collect and analyze data and provide value. Explain what your company wants to do with data in a language that consumers can understand.
if consumers feel that their personal data has been abused, they will lose trust in the company. Enterprises should not only invest resources in technology, but also explain it in a way that consumers can understand and accept. Customers are increasingly looking forward to improving their understanding of digital products and how services supported by AI can be realized. Countries require enterprises to use data within local legal restrictions.
4. Organize the upgrade. Business leaders must deploy necessary resources, upgrade technical infrastructure and meet the requirements of data mapping. Talents with extensive and in-depth knowledge in both data science and business must be hired. Data organization must be regarded as a connective organization connecting all parts of the enterprise, and it is recognized that modern organizations must properly deal with two conflicting powerful factions: one believes that data and algorithms have strong problem-solving ability, and the other does not. The contradiction between the two sides is a major feature of modern organizational operation culture: for example, Reed Hastings, CEO of Netflix, balances Silicon Valley's emphasis on analysis with Hollywood's emphasis on creativity.
5. Make profits through data maps. The construction of data atlas is used to support and formulate strategies, indicating that the value lies not only in product design and manufacturing, but also in how to solve specific problems for customers. The insight provided by the data map will help you choose the most suitable profit mechanism and plan a clear path from data to business results. You can use personalized recommendation based on data network effect to keep current income and profit. For example, Netflix uses real-time data to improve user retention rate; You can also use data maps to develop more perfect ways, strive for new sources of value, and broaden income and profit streams, such as Apple's entry into the credit card, television and medical industries; It can also counter the competitors who have mastered the data map in the market, such as Disney+ successfully entering the streaming media industry.
reshaping advantages
Data maps will reshape competition in every field faster than most people expected. Every enterprise should go beyond the demand of using data to improve operational efficiency and realize the competitive advantage of data map. Senior leaders must invest in upgrading the data infrastructure to get a real-time and comprehensive understanding of the interaction between consumers and their products and services. With this structure, we can work out a unique solution to solve the customer's problems.
for the leading digital enterprises, the continuous exploration in data mapping and other fields is creating new competitive advantages for them, and they are getting rid of their competitors in product development, user experience and other aspects. Therefore, their experience is worth learning widely.
Click to declare the annual list of new growth in 222
Vijay Vijay Govindarajan and N. Venkat Venkatraman)| Wen
vijay govindarajan is the Cox Distinguished Professor at Tucker Business School of Dartmouth University and an executive researcher at Harvard Business School. Venkate Carterman is the David mcgrath Chair Professor of Management at Boston University's Questrom School of Business.