Guest Speaker: The guest speaker is Hal Barian, the Chief Economist at Google. Barian has an illustrious academic career and is known for his work in information economics.
Introduction: Sandy Boss introduces Hal Barian as the guest of honor at a flagship PRA seminar. This seminar focuses on competition issues, particularly in the online context. Barian’s expertise lies in information economics, and he founded the Berkeley School of Information. In his spare time, Barian was a regular columnist with the New York Times from 2000 to 2007.
Use and Abuse of Network Effects: Barian’s presentation covers four main topics: Positive feedback sources in markets, including network effects. Decreasing returns to data. Platform competition. Personalization.
Sources of Positive Feedback: Network effects are a type of demand-side economy of scale where the value to a user increases with the number of users. This creates a positive feedback loop, as more users attract even more users.
Supply-Side Economies of Scale: Traditional supply-side economies of scale occur when the cost per unit decreases or quality increases as output increases.
Learning by Doing: Learning by doing is when the cost per unit decreases or quality increases as experience increases. This is a significant factor in productivity improvements and is observed in many industries. The progress ratio measures the unit cost after doubling cumulative production and typically decreases by about 15%.
Online World Fixed Costs: In the online world, many fixed costs associated with production have become variable costs. Examples of such costs include data center space, custom software, productivity tools, and communication tools. This has democratized access to computing and communication technologies for even small companies.
00:09:27 Data Network Effects and User Support Services
User Support: Zendesk, a Danish company, offers user support services on a fee-based, per-usage basis, eliminating the need for businesses to create their own support centers.
Angel Funding: Online labor markets like LinkedIn and Kaggle provide access to a global talent pool, allowing startups to hire skilled workers on a project-by-project basis.
Sales: Technology has enabled businesses to reach a wider customer base through online platforms, reducing the need for physical storefronts.
Variable Costs: The rise of cloud computing and other shared services has transformed many fixed costs into variable costs, making it easier for businesses to scale their operations up or down as needed.
Venture Capital Funding: Europe has seen a surge in venture capital funding since 2010, with over 4,000 new companies founded and $27 billion raised, indicating increased ease of entry for new businesses.
Network Effects: Increased usage of product A can stimulate the usage of complementary product B, leading to positive feedback loops that benefit both products.
Indirect Network Effects: Operating systems provide an example of indirect network effects, where the popularity of the system attracts developers, leading to more software and increased user adoption.
Search Engines: The argument that search engines benefit from indirect network effects due to advertisers is questionable, as users generally do not choose search engines based on the number of ads.
Data Network Effects: The concept of data network effects suggests that search engines learn from user clicks to improve their results, but this is more accurately characterized as a supply-side effect or learning by doing, which is common in many industries.
00:14:26 Learning from Data: The Continuous Improvement Cycle in Business
Learning Systems: Learning systems involve collecting, storing, manipulating, and analyzing data to achieve continuous improvement. Managerial buy-in is crucial for their successful implementation.
Data and Learning: Unlike classic returns to scale, learning systems require conscious effort and investment in data collection, storage, manipulation, analysis, and implementation.
Diminishing Returns to Data: Data, like any other factor of production, experiences diminishing returns. The improvement process, though gradual, can lead to strong competitive interactions within an industry.
The Circulation Spiral: An old concept from information economics, the circulation spiral describes how more users lead to more data and revenue, enabling further investment and user acquisition.
Network Effects: Network effects, even the mythical ones, work in both directions. A spiral of decline can occur if the number of users decreases, leading to reduced investment, revenue, and data.
Learning in Various Industries: Learning systems are applicable to various industries, such as hamburgers and media. By investing in learning about customers and outcomes, businesses can enhance their products and services.
Cheap Data and Technological Advances: The cost of collecting data has significantly decreased due to technological advancements, making data a byproduct of ordinary business activities. Advances in sensors, cloud computing, and open-source software have further driven down costs.
Decreasing Returns to Analytics: Investing in analytics to utilize cheap data generates significant initial benefits but faces diminishing returns as investments increase.
Industries at the Beginning of the Learning Journey: Industries like entertainment and mobile phones are at the initial stages of embracing learning systems and investing in analytical talent to exploit data for improved decision-making.
00:20:19 Data, Information, Knowledge, and Practice: The Importance of Investing in Learning
Data-Driven Positive Feedback: Data analysis can lead to positive feedback loops in businesses and organizations, leading to improved performance. However, utilizing data requires managerial focus and internal expertise.
Diminishing Returns to Data: Statistical models show that the accuracy of data-driven predictions improves at a decreasing rate as more data is collected. The mean squared error decreases linearly with the number of observations, but it is always bounded by an irreducible error term.
Empirical Evidence from Machine Learning: Machine learning models initially show a strong increase in accuracy as more data is used for training. However, the accuracy eventually asymptotes to the intrinsic error rate as the sample size increases.
Improving Performance Beyond Data: Increased sample size alone cannot provide a magical improvement in performance. Better models, improved algorithms, and more features can also contribute to improved accuracy.
ImageNet Example: The error rate in image recognition tasks on ImageNet decreased from 28.2% to 3.57% in just five years. This improvement was due to algorithmic and hardware advancements rather than an increase in the training data size.
The Data Pyramid: The information economy relies on data collection, analysis, and knowledge translation into action. The data pyramid involves collecting data, analyzing it to produce information, extracting knowledge from the information, and applying the knowledge in practice.
Learning by Doing and Productivity: Organizations that invest in the components of the data pyramid become more efficient over time through learning by doing. Learning and knowledge translation are active investments that exhibit decreasing returns but are critical for success in the 21st century.
00:28:11 Network Effects and Economies of Scale in Digital Industries
Network Effects and Market Share: In simple network effects, market share plays a crucial role. The larger network offers more value, making it more appealing to users, thus increasing its advantage.
Supply-Side Returns to Scale and Size: In contrast to demand-side networks, supply-side returns to scale focus on size rather than share. Reaching a particular size may deplete economies of scale. Minimum efficient scale indicates the optimal size for efficient operations.
Experience and Learning by Doing: Learning by doing, unlike the previous cases, is not automatic but requires investment and managerial capabilities. Upsets can occur, as seen in the social network industry, where MySpace dominated and then was displaced by Facebook.
Technological Changes and Disruptions: Technological innovations disrupt conventional wisdom and challenge established knowledge. Examples include the rise of Google’s search engine algorithm and the success of the iPhone despite skepticism.
Network Effect Examples: Higher numbers of advertisers on an online search platform increase revenue, which can be reinvested in improving the service, attracting more users, and creating a positive feedback loop.
Generalizing the Concept: The same principle applies to various industries, not just network effects. Revenue increases may not always translate to profitability, as costs must also be considered.
Profitability and Scale: There is no universal rule that larger firms will be more profitable. The mobile phone industry illustrates this discrepancy, with Samsung having a higher product share but Apple having a significantly higher profit share.
00:35:55 Platform Competition and Monetization Strategies in the Digital Economy
Changing Business Practices: In the tech industry, GAFA (Google, Apple, Facebook, and Amazon) dominate the global market and compete intensely in various ways.
Omnipresent Threat of Entry: The malleability of tools and factors of production in the digital world makes it relatively easy and cost-efficient to enter new industries, fostering competition.
Data Centers and Software: Companies can repurpose their existing resources like data centers, networking, software, and coders to enter new markets, such as creating a browser while already running an operating system business.
Competition Using Different Business Models: Tech companies not only compete with each other but also explore various business models to monetize their services.
Operating System Monetization: Apple and Amazon bundle their operating system with hardware, Microsoft charges a fee for licensing, Google open-sources its operating system, and other companies compete for advertising revenue.
Office Productivity Monetization: Office productivity is either bundled with hardware, sold as a package, provided as an online service, or offered as a premium online service.
00:39:56 Combinatorial Innovation and Mobile Internet
Combinatorial Innovation: Technological changes often involve combining inexpensive components to create new inventions. Examples include standardized mechanical parts in the 19th century and electrical parts in the 20th century. Today, web tools, mobile phones, and software components enable combinatorial innovation.
The Internet of Things: The Internet of Things combines mobile phone parts with data center services. This combination leads to new devices for the home, car, factory, and office.
Personalization: Markets have different ways of using monetization, like subscription, advertising, freemium. Software is malleable, allowing for different ways of being monetized.
00:42:37 Search Advertising and Contextual Ads: A Drive for Personalized Services
Personalized Data: The Driver in the Tech Industry Hal Barian emphasizes that personalized data, not only big data, is the driving force in the tech industry. Personalization services tailored to individuals provide a significant advantage, and companies compete to deliver the best personalization.
Online Advertising Search advertising leverages the search query as a strong signal. Additional personalization is less impactful in search advertising. Location, recent searches, and context play a role, but the query remains the dominant factor.
Contextual Ads and Browsing History Contextual ads based on page content, such as ads for fishing equipment on a fishing site, are less controversial. Ads based on browsing history, not the current browsing or search, are of interest.
Newspapers and Online Advertising Newspapers traditionally made money through contextually relevant ads on sections like home and garden, travel, finance, and automobile. News stories often lack contextually relevant ads. With the split of the newspaper industry, the cross-subsidization model is no longer feasible. Newspapers now show ads relevant to users’ history or browsing history, rather than page context. About 91% of news site traffic relies on this type of targeting.
Targeting Ads: Online advertising targets users by placing cookies on their browsers to track their interests and serve relevant ads. Retargeting involves showing ads for products that a user has previously viewed on other websites. Interest-based ads are displayed based on inferred interests from users’ website visits. Targeting is time-sensitive as ads focus on users who are actively in the market for a product.
Interest-based Ads and User Control: The interests used for ad targeting are inferred from users’ recent website visits. Users can view and edit their interest list in their ad settings to opt in or out of specific categories.
Positive Feedback and Data-Driven Business in the 21st Century: Positive feedback loops, demand and supply dynamics, and learning effects are shaping online businesses. Data is used to improve products, compete with other platforms, and provide personalized services. These trends are expected to expand across industries, driving the evolution of 21st-century business practices.
Abstract
Understanding the Dynamics of Digital Markets and Innovation: A Comprehensive Analysis
In today’s digital age, understanding the intricate dynamics of market competition, data management, and technological innovation is crucial. This article delves into these aspects, drawing from a seminar by Hal Barian, Google’s chief economist, and insights from various sectors, including Zendesk’s customer service model, European venture capital trends, and the strategies of tech giants like Google, Apple, Facebook, and Amazon (GAFA). It aims to provide a comprehensive overview of how network effects, data management, platform competition, and personalization are reshaping businesses in the 21st century.
The Seminal Concepts of Network Effects and Data Management
Hal Barian, a renowned academic with expertise in information economics and a former columnist for the New York Times, presented four main topics at a recent seminar hosted by the Program on Regulation and Governance: positive feedback sources in markets, decreasing returns to data, platform competition, and personalization.
Barian introduces the concept of network effects as a primary driver in digital markets. These effects, wherein the value of a product or service increases with more users, have been pivotal in the rise of online platforms. Direct network effects are seen when the service’s value increases with the number of users, while indirect network effects arise when increased usage of one product boosts the usage of a complementary product. These effects underscore the importance of market share in the digital field, where a larger network can exponentially increase a platform’s appeal.
Complementing network effects, Barian discusses the nuanced role of data in digital markets. While traditionally seen as a purely beneficial asset, he points out the diminishing returns to data. This concept suggests that as data collection increases, the marginal improvement in outcomes (such as algorithmic accuracy) tends to plateau. This revelation is vital for understanding how companies can optimize their data usage, focusing on quality over quantity.
Data-Driven Positive Feedback:
Data analysis can lead to positive feedback loops in businesses and organizations, leading to improved performance. However, utilizing data requires managerial focus and internal expertise.
Diminishing Returns to Data:
Statistical models show that the accuracy of data-driven predictions improves at a decreasing rate as more data is collected. The mean squared error decreases linearly with the number of observations, but it is always bounded by an irreducible error term.
Empirical Evidence from Machine Learning:
Machine learning models initially show a strong increase in accuracy as more data is used for training. However, the accuracy eventually asymptotes to the intrinsic error rate as the sample size increases.
Improving Performance Beyond Data:
Increased sample size alone cannot provide a magical improvement in performance. Better models, improved algorithms, and more features can also contribute to improved accuracy.
ImageNet Example:
The error rate in image recognition tasks on ImageNet decreased from 28.2% to 3.57% in just five years. This improvement was due to algorithmic and hardware advancements rather than an increase in the training data size.
The Data Pyramid:
The information economy relies on data collection, analysis, and knowledge translation into action. The data pyramid involves collecting data, analyzing it to produce information, extracting knowledge from the information, and applying the knowledge in practice.
Learning by Doing and Productivity:
Organizations that invest in the components of the data pyramid become more efficient over time through learning by doing. Learning and knowledge translation are active investments that exhibit decreasing returns but are critical for success in the 21st century.
The Rise of Micro-Multinationals and the Impact of Venture Capital
The transformation of the cost structure in the digital economy has given birth to “micro-multinationals.” Small companies now access technologies once exclusive to large multinationals, leveling the playing field in many industries. This democratization of technology is vividly illustrated by Zendesk’s model of providing user support on an a la carte basis, eliminating the need for in-house support centers.
Zendesk, a Danish company, offers user support services on a fee-based, per-usage basis, eliminating the need for businesses to create their own support centers. Online labor markets like LinkedIn and Kaggle provide access to a global talent pool, allowing startups to hire skilled workers on a project-by-project basis. Technology has enabled businesses to reach a wider customer base through online platforms, reducing the need for physical storefronts. The rise of cloud computing and other shared services has transformed many fixed costs into variable costs, making it easier for businesses to scale their operations up or down as needed.
The European venture capital scene further exemplifies this shift. Since 2010, there’s been a surge in new companies, with significant investment flowing in post-recession. This influx indicates a robust entrepreneurial ecosystem facilitated by accessible technology and funding. Venture capital funding in Europe has witnessed a significant upswing since 2010, with over 4,000 new companies founded and $27 billion raised. This signifies an increased ease of entry for new businesses, fostering a vibrant entrepreneurial ecosystem.
Platform Competition and the GAFA Phenomenon
The article also explores the intense competition among the GAFA companies. These tech behemoths, originally focused on distinct sectors, now overlap in various services, from operating systems to streaming. Their competition is unique due to the malleability of digital products and the combinatorial innovation seen in the industry. This battle extends beyond traditional product lines, encompassing diverse monetization strategies like bundling, licensing, and advertising revenue sharing.
Changing Business Practices:
In the tech industry, GAFA (Google, Apple, Facebook, and Amazon) dominate the global market and compete intensely in various ways.
Omnipresent Threat of Entry:
The malleability of tools and factors of production in the digital world makes it relatively easy and cost-efficient to enter new industries, fostering competition.
Data Centers and Software:
Companies can repurpose their existing resources like data centers, networking, software, and coders to enter new markets, such as creating a browser while already running an operating system business.
Competition Using Different Business Models:
Tech companies not only compete with each other but also explore various business models to monetize their services.
Operating System Monetization:
Apple and Amazon bundle their operating system with hardware, Microsoft charges a fee for licensing, Google open-sources its operating system, and other companies compete for advertising revenue.
Office Productivity Monetization:
Office productivity is either bundled with hardware, sold as a package, provided as an online service, or offered as a premium online service.
Personalization: The New Frontier in Online Services
Personalization has emerged as a key trend in digital services. Companies are increasingly tailoring experiences to individual preferences, a strategy especially evident in online advertising. Personalized ads, based on user data like browsing history, have become a cornerstone of digital marketing. However, this approach raises privacy concerns and underscores the need for balance between effective targeting and user consent.
Personalized Advertising:
Personalized advertising, utilizing user data like browsing history, has become central to digital marketing. This strategy, while effective, raises privacy concerns and necessitates a delicate balance between targeted marketing and user consent.
Embracing the Digital Transformation
In conclusion, the digital economy’s landscape is shaped by complex interactions between network effects, data management, platform competition, and personalization. Companies that understand and adapt to these dynamics are better positioned to thrive. This evolution also signals a broader shift in business paradigms, where data-driven decision-making and user-centric services are becoming the norm.
This article, by weaving together insights from various experts and sectors, offers a comprehensive understanding of the key trends and challenges defining the digital marketplace today. As businesses and consumers alike navigate this rapidly evolving landscape, staying informed and adaptable is more critical than ever.
Combinatorial Innovation:
Technological changes often involve combining inexpensive components to create new inventions. This process of combining existing parts to create innovative solutions is known as combinatorial innovation. Examples include the use of standardized mechanical parts in the 19th century and electrical parts in the 20th century. Today, digital tools like web tools, mobile phones, and software components make combinatorial innovation easier and more accessible.
The Internet of Things:
The Internet of Things (IoT) is a perfect example of combinatorial innovation. It combines mobile phone parts with data center services to create new devices for various applications, including homes, cars, factories, and offices.
Personalized Data: The Driver in the Tech Industry
Hal Barian emphasizes that personalized data, not just big data, is the driving force in the tech industry. Personalization services tailored to individuals provide a significant advantage, and companies compete to deliver the best personalization.
Online Advertising
Hal Barian highlights the importance of personalized data in online advertising, particularly search advertising. Search advertising leverages the search query as a strong signal for targeting users with relevant ads. Additional personalization factors, such as location, recent searches, and context, have a lesser impact. Contextual ads, like those based on the content of a web page, are less controversial compared to ads based on browsing history. However, ads based on browsing history, rather than current browsing or search behavior, are of particular interest to advertisers.
Targeting Ads:
Online advertising relies on targeting users to deliver relevant ads. This is done by placing cookies on users’ browsers to track their interests and serve ads based on those interests. Retargeting is a technique used to show ads for products that users have viewed on other websites. Interest-based ads are displayed based on inferred interests from users’ website visits. Targeting is time-sensitive as ads are most effective for users who are actively in the market for a product.
Interest-based Ads and User Control:
The interests used for ad targeting are inferred from users’ recent website visits. Users can view and edit their interest list in their ad settings to opt in or out of specific categories.
Positive Feedback and Data-Driven Business in the 21st Century:
Positive feedback loops, demand and supply dynamics, and learning effects are shaping online businesses. Data is used to improve products, compete with other platforms, and provide personalized services. These trends are expected to continue and expand across industries, driving the evolution of business practices in the 21st century.
Search engines like Google have become lucrative through targeted advertising leveraging the internet's scale and low marginal costs. Google's auction model optimizes ad revenue by ranking ads based on bid amount multiplied by the click-through rate....
Google's auction system is a second-price auction where advertisers pay below the winning bid and not their own. Nash Equilibrium exists when no bidder can improve its profit by changing its bid unilaterally, assuming others' bids are constant....
Google's revenue is primarily driven by its online advertising model, which leverages scale and network effects to thrive on low conversion rates. The company's success stems from its focus on improving search quality, developing efficient auction systems, and employing real-time experimentation and data-driven marketing strategies....
The digital revolution has transformed advertising, enabling real-time consumer feedback, convergence of online and offline experiences, and global reach. The internet's infinite scale and interactive advertising revolutionize brand building, allowing businesses to target and interact with individual consumers....
Online education offers unprecedented access and flexibility but faces challenges like low completion rates, quality control, and the need for human interaction. It has the potential to revolutionize access, complement traditional education, and address the global skills gap....
Online learning has undergone a transformation, driven by shifts in methodologies and challenges, leading to innovative approaches like personalized learning and AI-powered tutoring. Online education offers flexibility, personalized learning, and problem-solving opportunities, but faces challenges in verification, cheating prevention, and credentialing....