a critique of Net Worth: Shaping Markets When Customers Make the Rules, by John Hagel and Mark Singer
Esther Dyson popularized the idea that the internet will reverse the power balance between customers and marketers. Customers will control their own data, and marketers will bid for the attention of the customer. The insight of Net Worth, a visionary and practical new book by John Hagel and Mark Singer of McKinsey & Company, is that there is a powerful new opportunity for companies to take advantage of this shift toward customer control, by managing customer data on behalf of customers. The strength of the book lies in its comprehensive scenario of the evolution of "infomediaries," and the plan it puts forth for growing an intermediary.
But it would be unwise to use Net Worth unmodified as the core of a business plan, since there is a crucial flaw in the book's theory of infomediary evolution. Hagel and Singer argue that infomediaries will begin by offering a wide but shallow set of services, serving all buyers, covering all categories of consumer purchases. There are cultural and technical reasons that this scenario is implausible. Instead, I believe, infomediaries are more likely to evolve from narrower, deeper networks -- affiliate, community, syndication, and business trading networks -- which will offer a richer set of services to start, and will spread like ivy plants, outward and towards each other over time.
An infomediary -- in the Hagel and Singer scenario -- is a company that manages customer data for the benefit of customers. The customer lets the infomediary collect information about his or her activities and preferences -- and in return, it scours the world for helpful information -- and protects the customer from irrelevant and intrusive advertising. The infomediary is a butterfly business: it collects money from consumers -- for agent and filtering services --- and from vendors -- for targeted marketing and market research.
The difference between infomediaries and traditional direct marketing companies, such as Cendant and American Express, is that infomediaries are relentless protectors of customer privacy. They will command huge fees for advertising because they can deliver leads to interested and imminent buyers. Infomediary services need to act in customers' interest -- that's the only way that customers will trust them with all of their information.
This infomediary business has powerful network-effect economies of scope. The more customer data the infomediary has, the more it can serve the needs of the customer base, and the more valuable its marketing services are to vendors. An infomediary service will command great customer loyalty, because switching costs are high. The more data it has about a customer, the longer it would take that person to "train" a new service about all of his or her buying preferences.
However, in order to succeed, this business model requires a very large customer base, and a very large information processing infrastructure. According to Hagel and Singer's model, a new infomediary will require billions of dollars in marketing and IT, and seven years of patience before profits start to roll in. Hagel and Singer believe it will be imperative for entrepreneurial infomediaries to get a head start by partnering with traditional businesses that already have deep pockets and a base of customer relationships.
Hagel and Singer argue that in order to be successful, an infomediary needs to be built and managed as a monolith. Consumer marketing companies -- even credit card companies and grocery stores with card systems -- capture only a small fraction of a household's transactions. A successful intermediary should cover no less than absolutely every aspect of household commerce: soap and travel, housing and entertainment. Because of the network-effect, increasing-returns characteristic of the business model, the successful infomediary will become the one dominant player in the consumer economy.
The monolith scenario is implausible, for cultural and technical reasons. There are cultural barriers on both the marketer side and the consumer side of the business model. A broad-based infomediary requires a breathtaking amount of consumer trust, and marketers have deeply engrained habits that make it difficult to earn that trust. In addition, the price and feature comparison-shopping services offered by broad-based infomediaries address consumers as rational individuals, whereas consumer buying has a strong emotional and communal component. The emotional side of buying is changing in the internet era, but Hagel and Singer are wrong in their argument that emotion is on the way out.
The technical problems with monolithic intermediaries have to do with the volume of data, and with the limits of artificial intelligence. The monolith scenario implies gathering and analyzing the entire consumer economy. The scale is too big to be practical. Also, the monolith scenario depends on general-purpose agent technologies that will help consumers in every aspect of their economic lives. General-purpose agent technologies will be not be good enough for this scenario any time soon.
It seems more likely that infomediaries will grow out of networks and communities based on market segments and audience segments. If an infomediary starts with a particular domain -- a community of customers, or a set of services -- it seems much more likely that the infomediary will be able to earn trust, offer value to start, and grow from there.
The key to success for an infomediary is building customers' trust. Hagel and Singer argue that infomediaries will gain economies of scope and customer lock-in because of the richer, and more valuable set of data that they gather because customers trust them. Gaining that trust will be difficult, and losing it is easy.
The biggest barrier to building that trust is the mindset of traditional consumer marketing as it has developed over the last 100 years. The language of marketing is military, and the enemy is the customer: you target customers, you plan campaigns. You bombard consumers with messages until some of them relent and buy your product. It is hard to imagine marketing companies reversing these deep assumptions, embracing an attitude of "shielding" customers, rather than targeting them.
New interactive marketing firms are colonizing the internet by following the strategy manual of offline marketers. Axciom and Experian profit by harvesting customer data from free or cheap sources, processing it, and reselling it for use in direct mail and advertising. Online data miners such as Engage and Aptex, and ad services such as DoubleClick, AdForce, and NetGravity are exploiting online data resources in a similar way. Permission marketers, such as Yoyodyne (now AOL) and Netcentives offer a marginal improvement over the data harvesters. But consumers still get a poor deal. They have no idea how much their demographic information is worth to the company that purchased it for a $.50 coupon. Consumers are essentially selling Manhattan for a handful of buttons and blankets.
The online leaders are doing a terrible job of earning customer trust. AOL and GeoCities have faced backlashes against the disclosure of customer data. Amazon seemed to be doing a good job of providing personalized and customer-friendly information. But it was recently revealed that Amazon sold its "objective" book reviews. Yahoo is doing an increasingly bad job at its core directory service, as it attempts to grow its marketing and commerce services. The New York Times recently revealed that Microsoft has been secretly gathering a stash of data about Windows 98 customers.
It is hard to imagine a company coming from any position in today's consumer economy -- online or offline -- earning the level of trust it would take for individuals to hand over all of their personal data. I would even be suspicious of an infomediary start-up founded by people with good intentions and high integrity. Given the capital requirements of the infomediary business, an infomediary startup would probably be acquired by a traditional business that has demonstrated disrespect for customer privacy.
While it will be difficult for any business to build enough trust to manage customer data, it will be easier to build customers' trust in a particular domain. Once that trust is established, an infomediary may be able to gradually extend that trust through a partner network, by maintaining a rigorous set of privacy policies across the partnership.
The second cultural roadblock to Hagel and Singer's scenario is the emotional nature of consumer buying. In the early days of mass production and mass marketing -- through the teens in the U.S. -- advertisers stressed their products' qualities -- cheap, strong, nutritious, tasty. The birth of modern advertising came in the teens and twenties, after the translation of Freud into English -- advertisers discovered how to tap and stimulate non-rational desires. Advertising messages started to sell security, status, and sex appeal, instead of insurance, automobiles, and soap.
Net Worth assumes that consumers are rational. Hagel and Singer argue that advertisers sell image, and people buy images, because they don't have information. Give people enough information about pricing and products and that's how they will buy. The rise of infomediaries will deal a death blow to the agents who deal in sizzle. It does seem likely that the trend away from mass media, and toward target marketing and measurability, will reduce the dominance of branding as a marketing strategy. By assuming a rational buyer, however, I think that Hagel and Singer are missing a key point. Oddly enough, the element that's missing from the book's argument is the subject of Hagel's last book -- community and context.
The infomediary scenario in Net Worth -- a single application handling household purchases like a cross between a butler and a corporate procurement application -- misses the mark. That's only a part of how we'll buy. There are two faces to electronic commerce. One side is rational -- price, selection, convenience. The other side is emotional and social -- community, process and context.
People make household purchases in the context of a process, relationship, or community. People buy food in the context of family and friends; they buy financial services in the context of life events and household budgeting, they buy kids' sporting equipment in the context of soccer league. In online communities, affiliate networks and category destinations, businesses are assembling the context surrounding a purchase, and helping customers buy in context.
Within these networks and communities, business will be able to help customers store and manage information, and facilitate purchase. For example, a service could store contact information about a soccer team, send out reminders of games, and send requested information about soccer gear. Infomediaries of this type will be able to add value based on context, not just on price and features.
The success of the monolith scenario depends on a single company being able to gather and make sense of consumer data on a scale unprecedented by orders of magnitude.
Hagel and Singer anticipate that infomediaries will use a data-gathering approach similar to the monitors used by MediaMetrix to estimate total web usage. A small piece of software lives on the user's PC, and tracks all of the customer's clicks and purchases on the web, and sends the data to a central server, where it is stored and analyzed. Unlike MediaMetrix, which extrapolates from a sample of 40,000 users to the entire internet population, an infomediary wouldn't use sampling -- it would trace every user's actual movements.
This approach would cause similar problems to those that large web sites are grappling with today. Most web sites, including leading commercial sites, are only in the first stages of getting basic value from site usage data. Today's commercially available tools don't scale to the highest volumes. Tools to segment customers and measure marketing objectives are in early stages of development. Companies are just starting to develop methodologies to use the data effectively. Infomediaries will face all of these challenges -- but on a scale many times larger than the largest web sites.
Infomediaries will also face challenges integrating PC-based data gathering software with web sites. In order to track user activity on a web site whose content is stored in a database, it will be necessary to hook into the database content hierarchy. It's even more critical to hook into the site in order to get information about customer transactions. Hagel and Singer raise this issue, and suggest that infomediaries will need to build partnerships with web sites. But it won't be practical to perform this level of integration for the entire World Wide Web at once. There is a handful of very popular sites that most users visit in common -- and then a nearly infinite list of sites that people use separately. In order to make a service meaningful for an individual, it's important to provide hooks into the minority interests that make that person's life special -- collections of hymn books or fishing gear.
A monolithic intermediary will also run into trouble analyzing the data that it gathers. The authors observe that marketers waste a lot of money on customer acquisition, because customer data is so thinly spread. Even the companies with biggest databases -- banks, consumer marketing companies -- know only a tiny fraction of an individual consumer's data. However, companies today are struggling painfully to build coherent data architectures and make sense of the customer information they have already. It is hard to envision solving this problem any time soon with a single data warehouse of the entire consumer economy.
The analytics scale problems are hard enough with traditional demographic and transactional data. Throw in web usage data and the problem gets even messier. To make matters more difficult, decision time frames are getting shorter on the web. It used to take three to six months to analyze and plan a direct mail campaign. Companies expect to make web marketing decisions in one to four weeks -- or faster. Crunching the entire consumer economy, including browsing and shopping as well as buying, down to the level of individual user, at close- to-real-time speed -- is orders of magnitude greater than what people are doing successfully today. My estimate is that it will take several decades to be able to approach the analytical requirements of the infomediary monolith envisioned in Net Worth.
These scale issues are good reasons to envision infomediaries arising from a defined network, rather than a universal one. An infomediary service could get started by serving a defined but useful collection of sites -- say, travel and vacation reviews and restaurants. As these networks get a better sense of their customers, they can add products and services -- and add the integration into new sites as they go. An infomediary based on a subject or user community would be able to gather and analyze customer data at the state of the art of today's analytical technology -- rather than a decade ahead.
Another reason the monolith approach seems implausible is its dependence on the use of general-purpose agent technology. In Hagel and Singer's scenario, infomediaries start by offering agent services, which look for deals and information based on an individual's personal profile. The computer will search the web and find deals on toothpaste and stereos and vacations and mutual funds.
There's a fundamental flaw to this scenario -- despite the hype, general-purpose agents don't work very well. Hagel and Singer overestimate the potential of both main types of agent technology -- filters, which search for text, and bots, which search for data.
While text searching and filtering services are useful assistants, it still takes a lot of human effort to make sense of the data they gather. I've subscribed to Inquisit's news filtering services, which retrieve articles on topics of interest based on keywords. Inquisit works pretty well. But it takes a lot of effort to train it to weed out irrelevant stuff. And once you train it, the service still delivers a lot of junk -- it's not smart enough to weed out pointless press releases and poorly written news articles. After a while, I turned the service off, and returned to human editors to do most of my first-pass filtering -- the Wall Street Journal, News.com, SJMercury, MarketWatch, Meckler, CMP.
I found Inquisit difficult even in a work context, where I'm highly motivated to find relevant information. In a personal context, I would have even less time and patience. I doubt that general-purpose filtering services will be able to find what customers want. If I'm looking for a car or an investment, I would much rather have high-quality human editorial, and peer commentary with a reasonable signal to noise ratio.
Hagel and Singer's scenarios assume even more sophisticated artificial intelligence techniques which infer unstated desires from a user's online behavior. For example, they suggest, an infomediary would notice that you've browsed web sites relating to wine, cycling, and Europe, and would suggest that you might be interested in a cycling vacation through the wine country of France. In order for collaborative filtering and behavioral profiling systems to come up with that sort of brilliant suggestion, the system needs a lot of information to start with -- the results won't be very helpful to start. Even with a lot of data to chew on, will the technology be good enough to be a primary source of recommendations, rather than a useful aide to a human search?
I believe that artificial intelligence will continue to improve over time -- but not enough to create the scenarios the authors envision. Human editors, peer recommendations, and customer service agents will continue to play a key role. Human agents, assisted by software agents, will use technologies such as chat and telephony (Webline and Acuity) in web sites, and wireless handheld computers in stores to deliver smarter and more cost-effective personal service.
Hagel and Singer's software agent scenario also runs into difficulty with structured information like prices and product features. According to the scenario, infomediaries will launch shopbots that will skitter across the web, gleaning data about the prices and features of televisions and compact disks, and return information about the best prices. The effectiveness of shopbots will be enhanced by use of XML. Merchants will be able to tag the data on their web sites as a "price" and "product," so the shopbots will be able to compare prices and features on different cites.
Hagel and Singer acknowledge that merchants will resist being compared -- merchants already modify their sites to avoid pesky comparebots -- and merchants will resist meta-data standards. However, Hagel and Singer believe that this resistance will be overcome by the increasing power of infomediaries. Once the infomediary represents enough buyers, merchants will acquiesce, in order to gain access to the customers.
The book makes the strongest case that I've seen yet about how XML will be adopted for consumer e-commerce. Data standards trace the existing lines of power in an industry. Just as large grocery chains can force farmers to barcode their apples, powerful infomediaries will force online merchants to code their products. However, until infomediaries have that power, the results of shop-bot searches will be mediocre. So I don't agree with the book's premise that general-purpose agents will be the way for infomediaries to get started.
Another problem with shop-bots and other data agents is that is that there is no such thing as a general-purpose solution. XML enables businesses to define and standardize on vocabularies to define products -- but these vocabularies don't immediately pop into existence -- they need to be built laboriously, domain by domain -- food and audio products, furniture and financial services. The task of standardizing every consumer product in existence is too big and diverse for a single company to take on. Another reason, I think, that successful intermediaries will begin with an area of domain expertise.
I think that infomediaries are more likely to evolve from networks -- affiliate networks, communities, and syndication networks. Customers will be able to gradually extend trust from a reliable company to partners that also pledge respect for privacy. Networks will be able to meet customers' needs for context and community, as well as their rational preferences for price and selection. These networks will be able to provide buying-support services based on a close understanding of the customer base or the subject matter. Infomediary networks will be able to extend their site integration, agent services, and analytics by domain, at a manageable scale.
Traces of a network-based approach to the management of customer data are emerging among online communities and networks.
Networks such as these, I believe, will have the best chance to develop an infomediary business model, by providing buying services that protect customer' privacy. In order to become infomediaries, however, companies will need to reverse their orientation -- from managing customer data on behalf of marketers, to managing customer data on behalf of consumers.
What will be a wrenching shift in consumer commerce will much smoother in business to business commerce networks. In fact almost everything about the infomediary scenario makes more sense for business to business.
Consequently, I expect to see the first examples of spectacular success with the infomediary model in business to business commerce.
While I believe that infomediaries will evolve differently from the trajectory that Hagel and Singer describe, I believe the core premise of the book is correct, and the basic scenario is strong. The book draws a compelling picture of how the internet will empower buyers, by providing easier, cheaper access to information, personal service, and the ability combine as groups. The book is full of valuable and practical ideas about how businesses can profit from the power shift between sellers and buyers.
The richness of the scenario makes the book useful even where it is wrong. By assembling such a fully articulated plan for of the business and technical evolution of infomediaries, the authors have developed a model that rewards critique and adaptation. Despite the flaws in its argument, Net Worth is essential reading for executives whose business is the internet, or whose business will be transformed by the internet.
Copyright © 1999 Levin Consulting.