By Daniel Pimienta,
FUNREDES Head
[email protected]

In INET93, wonderful works have been presented on the evaluation of the Internet, in term of existing and projected node figures and therefore user figures (by Tony Rutkowsky and John Quarterman). The offered figure, as for 10/93 (John Quarterman: [email protected], see December issue of Matrix News for details), is of around 2 millions nodes and an estimated 15 million end-users for pure Internet, and an estimated 40 millions for all Internet gatewayed network together (this did not include videotex users, such as minitel users). The pursuit of the growth pattern shows that each human beings will be part of Cyberspace by the end of the century... Nobody believes it: so what are the tricks which torn the forecast?

We are very good in evaluating the node figures, but we are so far incapable of any credible estimation of the user base. This situation, a persistent heritage of the technology driven emergence of the networks, is less and less acceptable when the Internet is entering a market driven pattern. Hereafter, we analyze why this have to change rapidly and suggest some possible actions.

SOME GOOD REASONS TO START COUNTING SERIOUSLY THE USERS

The computer market, has been the first to be concerned by the networking phenomena. This is, between mid 70's and mid 80's, when Bitnet was in geometrical growth (measured in a constant figure of new nodes each day). The computer industry was having both a close follow-up and a good support for the networking growth.

The second market taking over, with the strength of the Internet (mid 80's to mid 90's), is clearly now the telecommunication market (bandwidth, protocol converters, routers and other telecom products). This is why the key measurement unit should be now the traffic pattern, rather than the number of node.

The next coming will be the information providing market: in the foreseeable future, the global revenues in the emerging information market may go much beyond the computer and telecommunication one's. Assuming that almost every Internet user is a potential customer, the forecast for the information consumed from the Internet are directly derived from the number of users, and much less concerned by the technical infrastructure. Information industries need to know TODAY the growth of the Internet in term of users in order to forecast their products for tomorrow.

Beside this basic and strong reason, there is a question of credibility and seriousness in the way the Internet growth is measured in new geographical area. Larry Landweber is doing a fine work filling flags over the world map. The color of the flag tells us the type of connectivity (Internet, OSI, BITNET, UUCP, Fidonet) but in noway the real impact in the flagged country. John Quarterman's Matrix Map Quarterly is providing a much more complex (and though closer to reality) perception, but is not yet providing user figures.

What real effect have a full Internet node accesed by 0.1% of the potential user base? Is not a Fido or UUCP solution reaching 20% of the potential users far more important? What makes us believe that a few users node benchmark will easily transform into a nation-wide network? What comparative value has a framework with identical flags measuring few users in some case and a strong penetration of the potential market in another one? What if the user growth pattern is lower than the demographic rate in some developing countries (ref. Syd Goodman's Inet93 presentation: [email protected])? What is the point in flagging a country where the network access is reserved to non nationals users (as in Africa, ask Michel Perdreau: [email protected])?

THE LIMITATIONS OF COUNTING THE USERS BY AVERAGING METHODS.

1) The actual figure of 15 millions full Internet users is computed by multiplying the number of nodes by an estimated average number of user per node of 7.5. I do not discuss the figure, but I do strongly question the stability of it. If one look at the statistical distribution function of the number of node per users he or she have to admit first that the variance factor is high (from hundreds of user in Bitnet systems to one user node ... and even a growing number of client Internet nodes with 0 users)! The obvious guesstimate is that the variance factor is higher than 1 making this function hyper exponential. The law of statistics calls for carefulness in using only the average for such type of function.

2) The average number of users per nodes, as a function of time, is obviously decreasing with the PC prices and the decline of centralized architecture. As was suggested recently by John Quarterman in the LAyC Network Meeting in Caracas, with the growth of resource servers, nothing forbid to consider it will go, at some point of time, below 1.

THE HIDDEN PROBLEM BEHIND THE LACK OF USER DEFINITION

What is a "real user"? Is somebody having an Internet address without using it to be accounted? Or a person connecting just once a week only to read some listservs messages? How about someone practicing daily e-mail but not yet capable to navigate within the information ressources? Obviously, the information market forecasters do appreciate the difference. So are doing the people who try to evaluate the real impact of networking in the developing countries and the effectiveness of the budget invested to create the user base...

A field study conducted in Venezuela (by Fabio Chacon: [email protected]) shows than only 30% of the user base makes efficient use of the services. Our field experience makes us convinced this is not a local curiosity. Our guesstimate says (totally empirical, in developing countries): -60% are infrequent users (they have an e-mail address but do not use it),
-from the remaining 40%,
--60% are few times a week e-mail only users, --20% are frequent e-mail users, but are not yet skilled in information access,
--20% are fully skilled users,
from the last categories how many strongly participate to the open circulation of the information (the one's who make listservs live)?
---Probably less than 10%...

Can we be so sure this is very different in the industrial world? The lack of user count strategy may be hiding a serious concern in the Internet and it's capability to sustain the predicted growth.

RECOMMENDATIONS:

Time has come to face the challenge of Internet users counting.

Counting methods are hardly	applicable,	some	statistical
oriented methodology have to	be applied. There		is enough
creativity and resources around	the Internet	to react	against the
actual paralyzed state.	Hereafter	are some doable
recommendations:

1. Enhance the country flagging system, providing a second more user oriented charts (showing an estimated percent of user penetration by segment like research, education, commercial). If this is well managed and some prestige value is associated to it, then the incentive for network administrators to be charted in the second one may be enough to trigger the process.

2. If the professional of opininon surveys are capable to predict the result of political elections with a .1 % confidence interval with an around 1000 units sample, why should not they in situation to do a survey of Internet users with a 20% confidence interval with a 10000 sample (which is by the way much easier and economical to contact)? If it happens difficult for the ISOC to find the budget for this sane exercise, I am sure a good negotiator could convince one of the companies to offer it on the basis of how good it will do for his image and business prospects to be the first positioning in the Internet :-).

3) Beyond the global user count, what is required is a dynamic model of categories of users (some efforts had already be done, see, for example, Pablo Liendo's onion layers model: [email protected]) and figures to help understand the changes of user habits, and what are the triggers which make them change from categories. This would produce benefits not only for the information provider industry, but also for the Developing Countrie's Science and Technology Authorities and also for the International Agencies in order to plan and regulate their policies.

4) We are probably reaching the point where the key parameter for the Internet growth figure is no more financial nor technical: this is the capability and propensity of people to become users, and to climb the different layers of services potentialities offered to them. People around the Internet should consider that good pieces of investments should be dedicated now in USER TRAINING. There are sporadic and creative efforts underway to provide on line training to the Internet Resources and more and more good didactic books appear in the market. A more systematic approach to user training will be required soon, and will have, as a prerequisite, a good understanding of actual user figures and a credible way to measure the impact of policies on the user capabilities changes.