Were there 9 million or 30 million people online in 1995? Why is there a 345% difference between "Hours Used Per Month" among different surveys? Does 30% of all online subscriptions and traffic come from new media industry Noise?

The major purpose of collecting accurate online user statistics (including commercial services, the Internet and the World Wide Web) is for business planning. And tracking the size of the market, segmenting the user groups and exploring potential growth trends and inhibitors are the most important tools a company has for strategic business initiatives using the Net.

Unfortunately, NNI has found that most online statistics presented, whether from respected firms offering random survey data or schools delivering self-selecting studies, have minor and sometimes serious flaws ranging from over-counting users or badly skewing important usage statistics, to poor analysis or lack of market segmentation.

While many studies should be unusable, they are instead being applied by companies for major corporate WWW rollouts and investments. This has caused what NNI calls Vanity Telecom—Corporate vanity as the primary reason for putting up a Web site, spurred on by hyped statistics instead of serious market analysis.

And the Net gets a whole lot smaller when proper market segmentation is applied. The total number of people online isn’t the crucial statistic—it's the number of users in your company's market segment (what NNI calls a Layer). For example, it's ridiculous to assume that everyone with a TV set is going to watch PBS programming. In fact, NNI found that most layers may be too small to support the number of companies vying for the online customer.

Inter-NOT: Online statistics Reality Check, 2nd edition, is the first report to cross-reference all the major studies, including Yankelovich Partners, Intelliquest, Nielsen Media Research, FIND/SVP, Jupiter/Yahoo, Georgia Tech, Project 2000, the NPD Group, and others, starting in 1994 and proceeding to September 1996.

Inter-NOT also features NNI's proprietary in-depth interviews with 135 users and non-users in our "Man on the Street" analysis, as well as a new market segmentation method called Layer Analysis, based on 15 years of historical telecom new media rollout data. Inter-NOT is the first in a five-report series titled Inter-NOT: Reality Check of the Information Age.

Total Online Population—Not a Mass Market

There never were 25 million US online users and there still aren't 25 million today. NNI believes there are only 15 million people online (encompassing the Internet, World Wide Web and commercial services), accounting for only 6% of the population and 10% of households. In short, the overwhelming majority of the US population is NOT online today.

The oft-cited figure of 25 million online users was a 'seat-of-the-pants' analysis, made up when a reporter put a computer programmer on the spot. As Daniel Dern, author of "Internet Guide for New Users", stated in 1995:

For the last three years there have been 25 million online users. If the industry is supposed to doubling every year, how can this be?

NNI's conclusions, laid out in detail in the body of the report, are based on cross-referencing four studies and research from O'Reilly, Inteco, the NPD Group, and Network Wizards, as well as weighted averages from over 25 published research studies.

Other Trends

• Slow Market Shifts in User's Gender, Education, and Income

Since 1995, there has been some shift in the online community— from Phase 1 users, (users who are online primarily for work and have been using online services for years) to more casual, non-work users. Therefore, there has been a gradual demographic shift towards more females, the less educated and the less affluent getting online. For example, NNI found that 33% of online users are woman, compared to 31% a year ago. However, these shifts in no way reflect the average American family. For example, when the US Census data is compared to Nielsen's findings, there is a 116% difference in those having a college degree, and an 89% difference in households incomes over $80,000.

• NNI Believes the Average User is Online for 16 Hours a Month

Tracking hours of use shows a major variance among companies. For example, an average of random surveys for 1995 found the average user on for 18 hours per month, while Jupiter/Yahoo showed 80 hours per month, a 344% difference.

• Massive Churn in the Commercial Markets

As predicted in the first edition, up to 50% of users who try an online service do not continue, sometimes immediately, sometimes after a few months. However, no study has properly included churn in their models.

• Shifts In Subscription Accounts

Subscription data and recent findings by The NPD Group and Odessy Research confirm a shift from commercial service providers such as Compuserve toward Internet ISPs. Yankelovich believes there will be a 20% drop in commercial online service subscriptions by year end 1996. NNI's assessment is that WWW users are more oriented toward work and e-mail, and therefore more stable than those interested in entertainment applications, such as chat.

Total Online Statistics

As the exhibit below exemplifies, online statistics today are all over the map. From a high of 43 million adults online according to Yankelovich, or the 50 million adults who have access according to Nielsen, down to the more rational versions from O'Reilly, Dr. Hoffman, or NPD Group, NNI found serious problems with most statistics.

(In millions)

43.0 (21.5%) Adults use online services, Yankelovich Partners, 8/96

50.0 (24%) Adults (16+) have Access to the Net, Nielsen Media, 8/96

35.0 Adults (Over 16) use Internet & Online services, Intelliquest, 7/96

36.7 Adults have accessed the Internet, Fairfield Research, 7/96

10.4 Use the Internet, Boardwatch, 6/96

7.6. Computers hooked up to the Internet, Network Wizards, 5/96

23.5 World Wide Web users, IDC, 5/96

9.6 Online households in the US, 1995, Jupiter Communications, 4/96

15.0 Subscriptions to online services, Simba, 3/96

7.0 Home Internet use The NPD Group, 3/96

9.0 Internet and WWW Users in 1995, Morgan Stanley, 2/96

9.5 Use online Services, Find/SVP, 1/96

10.0 Or less use the Internet Dr. Donna Hoffman, 1/96.

37.0 Have access to the Internet Nielsen Media, 11/95

5.8 Adults connected to the Internet, O'Reilly & Assoc., 9/95

8.6 Internet users, Forrester Research, 9/95

9.1 Subs. for AOL, Compuserve, Prodigy, Microsoft, 9/95

6.7 Households accessing online services, INTECO, 8/95

40.0 Use the Internet "Launching a Business on the Web", 8/95

28.5 Adults (14%) are online Yankelovich Partners, 5/95

16.0 US Internet users, Management Forum International, 9/94

25.0 People use the Internet. Common Knowledge, 1993-1996

Sources: The Companies, NNI, 1996

How can there be a 150%+ difference between studies from the same timeframe? An examination of the major flaws reveals many of the problems.

Major Flaw 1: Terms, Definitions and What's Examined (Almost Anything)

Inteco measured households, Intelliquest studied "Adults over 16", FIND/SVP included the entire family; the list goes on and on, with each study counting different things. For example, Nielsen counts "access", (someone who does not necessarily use the Internet or a commercial online service but could through someone else), while O'Reilly measured actual users.

And the networks and services people use, whether commercial services, WWW, or the Internet, are sometimes broken out, such as in the NPD Group study. Sometimes only the Internet is counted, as in the case of Morgan Stanley. In the Yankelovich study, anyone online by any means is counted.

Survey Methodologies and Statistics

All of the traditional methodologies used for collecting online statistics have problems and numerous caveats attached. There are three primary methodologies used by online statistics researchers:

1) Seat-of-the-pants statistics

2) Random telephone surveys

3) Self-selecting online surveys

And then there is NNI's methodology:

4) Statistical comparisons, and the subsequent application of historical new media modeling.

Major Flaw 2: Over-Counting Users

Yankelovich found 21% of US adults online, mostly by measuring anything that moves. This is not a study of Internet users alone, or commercial online service users, but all types of online users, including non-Internet e-mailers, users on free trials, or even people who have tried but no longer use online services. However, Yankelovich is not alone in this methodology, which predictably yields misleading results.

Major Flaw 3: No Accounting for Households Without Phones

NNI estimates that 8% or more of the population lives in households without telephones (the FCC believes it was 6% in 1995)—a figure greater than those who live in online households. However, almost all random surveys do not account for this properly.

Major Flaw 4: Consumers Don't Know Terms

In interviews with consumers, NNI found that almost all non-technical users think that America Online is the Internet, Information Superhighway, World Wide Web, or even a CD-ROM. A flaw in Nielsen's research, focused on by Dr. Hoffman, was that 13% of the users who stated they had online access at the beginning of the survey recanted by the end.

Major Flaw 5: Self-Selecting Surveys Are Skewed

NNI found a pattern with self-selecting surveys always skewed in the same directions:

• More male users

• More educated users

• More computer/new media industry users

• Many more hours spent online

• Higher-end technology used

• Greater technological optimism

For example, GVU had a 38 point spread between males and females, while Nielsen found only a 20 point spread. Meanwhile, Survey.Net found that 40% of survey respondents worked in computer-related fields, compared to Nielsen finding that only 11% of online users were in these fields.

Major Flaw 6: Politics (Know Your Salesman)

Whenever a statistic is presented the reader should ask why the study was done: Was it done by someone who is selling a book, using the findings to make some piece of legislation, or puffing up their own figures for investors? For example, the Nielsen study was conducted by Commercenet, a industry group set up to promote commerce. (And did you know that Commercenet received a $2 million grant from the government?) Intelliquest on the other hand, has merged with Pipeline Communications, an Internet access/content provider.

Major Flaw 7: Online Subscriptions and Puffery

Online subscription statistics are based more on companies creating numbers to please investors and advertisers than accurate accounting. It is a well-known fact in the new media industry that puffery occurs regularly in online subscription numbers provided by the services themselves. For example, Compuserve adds its Japanese service, Prodigy counts more than one user per household, and AOL counts many of its free trials.

NNI has filed a complaint with the FTC to investigate the billing practices of AOL and other service providers. In a rather seamy side of the business, many online firms bill customers, even when they signed up only for a free trial. NNI believes that this could account for as much as 5% of all accounts.

Also, NNI found that new media "noise" accounts for at least 30% of some ISP accounts (especially for those from Microsoft and AT&T). From new media freebies, to users with over 10 accounts (including business accounts), one thing becomes clear:

There is little relationship between online subscriptions and households or the number of online users.

Major Flaw 8: Combining Business and Consumer Services

Almost all statistics presented have not clearly segmented the market for business and consumer customers.

Major Flaw 9: Application Breakouts—Cuisinart Results

Without proper consumer and business segmentation first, the application breakouts are just a Cuisinart of results, not serious statistics for planning. For example, Intelliquest stated that only 17% of users give entertainment services high grades, and other scores these sites "abysmally low". NNI believes this score must be the result of combining users who never visit entertainment sites, such as business users, with those who view them all the time, therefore compromising application breakout’s usefulness for planning.

Solution 1: NNI Market Segmentation—Layer Analysis

It is ridiculous to assume that everyone using the Internet will be a potential user of a business or entertainment Web site. The exhibit below highlights NNI's major Layers (updated from the first edition release in June 1996). NOTE: This is a work in progress and more reliable data is required.

The significance of Layer Analysis is straightforward. It isn’t the total population that is important, but the layer your customers are in.

The key tenets of Layer Analysis are:

• There are distinct layers of users, and each layer is online for its own purposes, focusing on specific content, applications and services, with little, if any, overlap by different user groups.

• Besides the stratification of users and their applications, NNI believes that the average number of sites visited regularly by an individual will be only 7-15.

• Each group must be thought of and analyzed as a separate entity for marketing and in terms of current and potential revenues and growth projections.

	% of Users	Population Totals
Business Applications
Corporate Business	28%	4,200,000
High Tech & Telecom *	7%	1,050,000
New Media Industry	6%	900,000
Schools	22%	3,300,000
Govt., Non-profit	10%	1,500,000
Entertainment & Personal Use (Hobbies and Discussions)
Entertainment	20%	3,000,000
Kids**	4%	600,000
Adult	3%	450,000
Total	100%	15,000,000

*"High Tech and Telecom " are users employed by high tech or telecom, cable, or wireless companies, who are not directly responsible for online product.

** "Kids" are 1-18 years old and go online, sometimes supervised, but not through educational channels.

Furthermore, each layer has very different usage patterns for commercial online, Internet, and WWW traffic. For example, while academics primarily use the Internet, consumers generally use commercial services such as AOL.

There also seems to be a common belief that the online market was born in 1995. While user numbers have dramatically increased over the last three years, there is a long history of online use since the 1980s. The truth is that there have been seven major groups which comprise the foundation of all online services. NNI refers to these as Phase 1 users. These include:

• Bulletin Boards (over 100,000 according to Boardwatch)

• Corporate Databases, including Lexis

• Financial Services, including Telerate

• Schools, Universities, Government Agencies

• The New Media Industry

• Journalists

• Computer Industry, Programmers and MIS Managers

Applying Historical Models to Layer Analysis — NNI Places a Bet

In the original version of this report, NNI predicted that the Entertainment sectors of the industry would be hit with massive turnover and that 35%-50% of commercial (including AOL) traffic would be chat services. NNI's prediction was formed by first applying Layer Analysis segmentation, then referencing the historical models of other telecom products, including pay-per-call services, Minitel and videotex user models. Based on recently released information from AOL — NNI was 100% accurate.

The companion report, Inter-NOT: The Terrible Twos: Online Services Learning Curve, details Layer Analysis and looks ahead to the next 12 months, using historical models as the predicting algorithms.