Behind the News

Analysis of Industry Events

No Agreement on Internet Statistics

One indication that the Internet has truly gone commercial is that the company that became famous for measuring television viewership--ACNielsen Co.--is now responsible for one of the most controversial surveys of Internet use. Reactions to that study indicate that getting a handle on who's using the Net will prove a bit trickier than finding out who's watching "Friends."

The results of the survey, performed a year ago by Nielsen Media Research of New York City--a unit of Dun & Bradstreet-owned ACNielsen--and paid for by CommerceNet of Menlo Park, CA, had no sooner been released than the authors began catching flack from one of the survey's own instigators, Donna Hoffman, an associate professor of management at Vanderbilt University in Nashville, TN. Essentially, Hoffman said Nielsen had misinterpreted the raw survey data, failing to weight it correctly for the U.S. population as a whole, and consequently the numbers Nielsen reported were too high.

The Nielsen/CommerceNet study also conflicted with a survey done at about the same time by O'Reilly & Associates, the Sebastopol, CA-based publishing company. O'Reilly announced that there were 5.8 million U.S. users with direct Internet access, while Nielsen reported 37 million in the U.S. and Canada who had some kind of Internet access. Hoffman, applying U.S. census demographic data, said Nielsen's figure should have been 28.8 million. Since O'Reilly used a restrictive Internet definition, its numbers could be expected to be much lower. But the controversy illustrates the difficulty involved in any enumeration of Internet users or even in defining the Internet itself.

Spinning Raw Data

At Vanderbilt, Hoffman codirects Project 2000, a research program studying business uses of the Internet and other emerging media. CommerceNet is a consortium, partially funded by the federal government, of 130 companies interested in furthering Internet commerce. At a time when most estimates of Internet use were based on studies from inside the computer industry and conducted using the Internet itself, a five-person CommerceNet team including Hoffman selected Nielsen for the 1995 survey, which was to be non-proprietary and use traditional telephone polling techniques. Nielsen employed random digit dialing to poll 4,200 persons in August 1995.

Under a previous agreement, Hoffman claimed the right to see and analyze Nielsen's raw survey data. But soon after the results were announced, Hoffman began questioning the figures and saying that she couldn't get the information she needed from Nielsen. Eventually, she received the survey data and reanalyzed it. "We had a lot of questions about the methodology, and Nielsen was not very responsive. To this day some of them remain unanswered," Hoffman says. "We began to suspect that the problem was in the representativeness of the sample, and as it turned out, it was."

Essentially, Hoffman says Nielsen's sample overrepresents demographic groups more likely to use the Internet, such as those with college degrees, and underrepresents groups less likely to use the Internet, such as men and women over 55 with less than a high school education. In addition, Nielsen did not check for consistency of responses, she says. After reweighting the raw data, Hoffman came up with figures about one-third less than those Nielsen reported, namely: 28.8 million with Internet access (vs. 37 million reported by Nielsen); 16.4 million who had recently accessed the Internet (vs. Nielsen's 24 million); 11.5 million with access to the World Wide Web (vs. Nielsen's 18 million); and 1.5 million who had ever used the Web to purchase something (vs. Nielsen's 2.5 million).

A Nielsen spokesperson, speaking to the New York Times, defended the survey as containing better data than had ever been gathered before.

Since Hoffman's criticism, Nielsen has applied its own weights--based on age, sex and region--to the raw survey data and is producing revised numbers, which will be announced in a few weeks, according to Asim Abdullah, executive director of CommerceNet. However, the difference between those and the original Nielsen numbers won't be as significant as Hoffman reported, Abdullah predicts. Nielsen feels that applying weights for education and profession is too difficult, because it is hard to collect that data accurately. "The controversy is about weighting techniques, not about the health of the Internet or the potential of that marketplace," he says. "Our members are interested in getting a quick sense of where the market is today and how fast it's growing. Nobody is arguing with those facts."

Other Assessments

O'Reilly's survey, conducted by its own online research group, also used random digit dialing and conducted interviews with 29,901 individuals. Unlike Nielsen, O'Reilly rigidly defined an Internet user as a U.S. resident over 18 (Nielsen used age 16) who has direct access to the Internet and uses e-mail as well as one or more other Internet tools, such as a Web browser, FTP, Gopher or Telnet. O'Reilly excluded those connected to the Internet only through a commercial online service such as America Online or CompuServe, since those users do not have direct access to all the Internet's resources. With subscribers to online services added, O'Reilly said there would be 9.7 million Internet users. In addition, O'Reilly projected that 15.7 million would have either direct access or an online subscription by sometime in 1996.

For some observers, it's the lack of a universal definition of Internet and Internet user that overhangs all such surveys and tends to prevent meaningful comparisons. One source of Internet data who has studied such definitions is John S. Quarterman, president and editor of Matrix Information & Directory Services (MIDS)--a newsletter on Internet use and statistics--and a senior technical partner of Texas Internet Consulting in Austin, TX. Quarterman sees three concentric levels of Internet structure and use. The most inclusive is what he calls the matrix, which is all worldwide computer systems that can exchange electronic mail. The next largest category he calls the consumer Internet, which includes anyone capable of accessing Internet services beyond e-mail. The most restrictive he calls the core Internet, those capable of both sending and receiving Internet services interactively. A 1994 MIDS survey estimated 7.8 million core Internet users, 13.5 million consumer Internet users and 16.3 million e-mail or matrix users. If users of online services and other distributed networks were included, the matrix would have included 27.5 million users at that time, the study says.

Quarterman also criticizes the Nielsen survey, although he reaches a conclusion different from Hoffman's. "I thought they spent a lot of money and missed some basic points which radically diminished the value of their results," he says. As a result, he believes Nielsen probably underestimated the total number of users. Not defining Internet was a critical error, he believes. "A lot of people say Internet when they really mean electronic mail, which I find confusing and not terribly useful. The reason Netscape is worth a few billion dollars is not electronic mail, it's interactive things like the World Wide Web."

Other difficulties in surveying Internet use include the changing nature of the technology and the rapidly changing Internet user population. Quarterman said the technique of random digit dialing is "not a magic bullet to solve all the problems of Internet surveying." MIDS uses the Internet itself for conducting its surveys. Whatever technique is used, "You're not going to get an incredibly precise picture because it's changing so quickly," he says. "But you can approach estimates that make sense."

That surveys of Internet use are a hard nut to crack is probably the only thing observers agree about unanimously. "I've seen several people try to survey Internet usage by putting a survey on the Internet, which is ridiculous; only people that are Internet-involved will see it," says Steve Young, director of electronic commerce for Input, a market research firm in Mountain View, CA.

But just asking people if they have Internet access may not work either, says Jim Warren, a columnist for publications MicroTimes and Government Technology. "It's like asking teenage boys if they've had sex," he says. "They'll say, 'Yeah, sure.' It doesn't work."

At this point in tracking the demographics of the Net, it may not be literally true that your guess is as good as mine. But there's a grain of truth in that old chestnut.

--Don Dugdale

Top European Vendors Face Uncertainty

Prospects are dicey for the four leading European systems vendors, according to market research analysts. Bull, ICL Fujitsu, Olivetti and Siemens-Nixdorf (SNI) face a three-year window of opportunity to sort out problems in their domestic markets and develop a growth model to beat back international competitors. They also face a difficult time expanding internationally beyond the continent. Nor are these suppliers likely to gain much protection from the emergence of a single currency or a federal government in Europe. The only area where analysts see these companies achieving sustained success is the IT services business, but even on this front it will be hard for the four to build up major market shares over time.

The European offices of both the Gartner Group and the Meta Group see big challenges coming up for European systems suppliers. While each firm retains a strong base within its home country and locality, the Gang of Four probably has a maximum of three years of security left from their installed bases, according to Will Cappelli, European program director for services and system management strategies with the Meta Group, based in Camberley, Surrey, U.K. Cappelli sees enough life in the current user investments to sustain the vendors for the immediate future, but he insists that they must reform to meet the rapidly approaching challenges.

"Paradoxically, it is now easier for U.S. firms coming over to Europe. U.S. companies originally made mistakes here, but they now have a much more friendly approach to Europe, making it harder for traditional European vendors to compete," Cappelli says. American firms are more used to dealing with highly fragmented markets, he notes. The U.S. suppliers have also benefited from faster rationalization of their work forces, compared to Europe where reducing the head count has taken longer and cost more due to the Social Chapter legislation from the European Commission, which protects worker rights. In general, U.S. vendors more readily take a global view of markets than their European counterparts have.

U.S. companies, such as Digital, EDS and IBM, also are proving successful in grabbing substantial services business in Europe. This is not to say that firms like Bull, ICL and Olivetti cannot compete in service provision but rather that the competition will be fierce. In fact, Gartner Group rates Bull as a 50 percent-plus services operation in terms of revenue. ICL also has a strong pedigree in services, and Olivetti has an alliance in this area with Hewlett-Packard in Europe.

Player Profiles

On the Unix front, Cappelli sees HP and Sun Microsystems as presenting a serious threat to the Euro vendors, because they can offer more competitive products in terms of price and performance. IBM's AIX offerings also play a leading part, and there is the growing popularity of Windows NT Advanced Server to deal with. Even Germany, the bastion market for OS/2 in Europe, is seeing a shift to NT. Unix does have a future in Europe, but it won't be with the indigenous vendors.

Bull, headquartered in France, has the most aggressive Unix strategy among the four vendors. The company has merged its open systems and mainframe businesses, which Ed Thompson, a Gartner analyst based in Egham, Surrey, sees as a sign that it is aiming at the high-end Unix sites for mainframe replacement systems. Thompson says the company is temporarily stealing a march over IBM in providing AIX solutions on the PowerPC platform, to which Bull is dedicated. He says Bull will offer the 64-bit version of the PowerPC chip architecture in the near future as well. In fact, Bull's Unix business has grown dramatically over the last six months in Italy, Spain and the U.K., based on its growing specialization in managing projects. Overall the company forecasts a growth factor of three up to the year 2000, although Thompson reckons that Bull will do well to reach 50 to 100 percent expansion.

Due to the expanded influence of Fujitsu in ICL affairs, Thompson says, the U.K.-based company will decommit to some extent from Unix. While ICL will manage its current user base for Sparc, it is expected to adopt the UltraSparc chip as its next-generation architecture. The increasing branding role played by Fujitsu is pushing ICL toward a volume products business. The ICL identity now has most relevance to the local U.K. market, where its original mainframe power base in local government is being eroded.

Olivetti's primary backing is for UnixWare, and the Italian company will become a reseller of Digital's Alpha and Intel machines. SNI is shifting its emphasis from Mips processors to Intel while concentrating on vertical markets in the public sector, telecommunications and retail.

Neither Meta nor Gartner sees the Euro vendors making much headway in the international IT markets. SNI in particular has to strike out, having emerged from a period of losses and with cash to spend. The company recently opened an operation on the U.S. East Coast and might consider further acquisitions after buying Pyramid Technology of San Jose, CA, last year. Current Gartner figures show that SNI has a 60 percent share of the central European market, while Olivetti and Bull can claim 35 percent of the Italian and French IT markets, respectively.

Cappelli of Meta Group says that ICL is an odd case when it comes to developing an international profile, in that any expansion through acquisition would probably be done under the Fujitsu name. Bull, which recently sold the Zenith Data Systems PC and server business, is unlikely to buy out overseas vendors. It is not Olivetti's strategy to acquire abroad either, according to Cappelli, who says that more divestment is required before the Italian-owned company can even consider such a move.

One fillip that might come the European vendors' way is the change to a single currency, now dubbed the Euro. The problem is that the Euro may take two to 20 years to be fully adopted, but when it is a vast amount of conversion work will be required to retool existing systems, according to Thompson of Gartner Group. In particular, this will be the case if two currencies are run concurrently in various nations. The four European firms have an opportunity here, but the trouble is no one knows when the Euro will be introduced and how widespread its adoption will be.

Plausible Scenarios

A likely outcome for the European systems suppliers could be acquisition or merger with U.S. or Far Eastern companies. This process has already started with Fujitsu taking over ICL. Unless the remaining independent vendors in Europe can change their fortunes, they may experience similar fates. Some form of shakeout is bound to impact the European systems suppliers.

At least in Cappelli's view, an outbreak of protectionism by the European government remains improbable as a source of help for the local IT firms. He says there is no evidence that the authorities are prepared to support moribund industry sectors. There are far more successful high-technology sectors, such as software in the U.K. and biotechnology in France and the U.K., which could get protectionist relief.

Hardware is becoming a commodity, and the European vendors are faced with reselling the popular operating software platforms originating from the U.S. They can maximize opportunities in service provision through developing cross-platform portfolios, in particular offering outsourcing capabilities for distributed systems and desktops, as well as systems integration, project management and consulting. Gartner Group rates Bull's management as having the most vision and drive, as well as having extra markets to diversify into, such as smart card technology.

The bottom line for these vendors is that they must achieve leadership in their own geographic marketplace if they are to survive, let alone expand beyond the European domain. The challenge for the European IT firms is to deal with the internationalization of the IT markets in their own backyards.

--Dom Pancucci