Customers vs. Subscribers

Reader asks why we count 78 million online subscribers at the end of 2003 whereas many other sources put the number of people online at between 150 million and 180 million.

[Response to Top 24 U.S. ISPs by Subscriber: Q4 2003 from May 16, 2004.]

[May 24, 2004]

Dear Mr. Goldman,

This relates to your article entitled "Top 24 U.S. ISPs by Subscriber: Q4 2003" and accompanying article entitled "Top U.S. ISPs by Subscriber: How We Count".

Firstly, we very much appreciate the explanations of how you arrive at your figures and the nature of your sources. Life would be easier if others did the same—which brings me to some questions:

In connection with internal business analysis at Proxyconn regarding both the North American Internet and ISP markets as well as the global picture, we have been searching for sources of "reliable" statistics.

We came across on the web a Morgan Stanley Equity Research report entitled "Global Market Sizing of TMT Products and Services" dated September 12, 2003. It is a pretty extensive analysis covering telephone lines, installed PCs, mobile phones, cable subscriptions, Internet users and credit/debit cards (as an indicator related to e-commerce). Their estimate for North America Internet users for 2003 is 198 million (of which approximately 9 percent is Canada i.e. about 180 million US).

In one of their footnotes (on Page 3 of their report), the following explanation appears: "Internet users reflect the number of users accessing the Web (users may share/use multiple devices, users accessing the web from home and work are counted only once); source—Morgan Stanley estimates, IDC."

We also came across a Comscore Press Release dated March 10, 2004 which refers to the "U.S. online population reach 150 million people".

In the world of statistics and margins of error, the 180 million of Morgan Stanley and the 150 million of Comscore somehow seem to fit within one's subjective impression of "consistency". I am struggling with how to fit Jupiter Research's figure of 78 million US ISP subscribers into the same consistent picture.

Can you provide some insight? Am I simply missing something very obvious?

Thanks,

Alex Hill, Proxyconn

P.S. Also just wanted to check that the Q1/Q4 comparisons being shown in the table of the 2nd article refer to Q1 2004 and Q4 2003 (i.e. the article refers to Jupiter Research estimates for March 31, 2004).

Dear Mr. Hill,

Re: P.S. Also just wanted to check that the Q1/Q4 comparisons being shown in the table of the 2nd article refer to Q1 2004 and Q4 2003 (i.e. the article refers to Jupiter Research estimates for March 31, 2004).

Yes, that's correct.

The difference between the Jupitermedia and the Comscore, etc. statistics are twofold:

1) Jupitermedia reports subscribers, and there is usually more than one person per subscription (i.e., an entire family on one AOL account with several screen names pays one dialup fee). Comscore, etc. report "people" but I think they have to estimate the number of people per subscription, because the raw data that I've seen is number of subscriptions, not number of people.

2) Jupitermedia does not count at work, student populations, and the government population. The Comscore , etc. reports may do so (I do not know whether or not they do).

I think that your question is probably a very common one. With your permission, I would like to publish it, with my response, next week on ISP-Planet.

All the best, and please do not hesitate to contact us with further questions in the future,

Alex Goldman
Managing Editor, ISP-Planet

P.S. There's also a lot of useful information here.