Issues

number of issues are plaguing the development, implementation, and use of web measurement technology. First, there are no clear standards or language for web measurement. The available tools are still in the initial stages of development and use. Comparisons between the type of measurement accomplished by these tools is not comparable to other media. And within Internet measurement tools, there is no consistency in reporting. Because the size of the market is difficult to assess (e.g., no subscriptions), it is difficult to estimate reach, frequency, rating, and share. Finally, the audience on the Internet is still relatively small compared to television and other forms of media. Other problems include:

Caching: Caching is the storing of popular pages on a server or hard disk. The user benefits because loading time is decreased. However, the publisher cannot provide correct usage data to the advertiser because cached pages are not on the site's log file. If the publisher can access the server's log file, they may have access to information about these cached pages.
Proxy Servers: Proxy servers are a firewall between a company and the Internet. All requests go through the proxy server under one IP address. The log file will show only one user (the proxy server), although in reality, there may be multiple users.
Robots: As robots run through IP addresses multiple times a day, they inflate page visits. Fortunately, most measurement software can recognize robots.
Low Click-Through Rates: A study by I/PRO and DoubleClick found the average click-through rate on banner ads to be 2.1%. A more recent study by NetRatings, Inc. found a 1% click-through rate (Vonder Haar 1998). Advertisers, discouraged by low click-through rates, have been able to create arrangements with publishers to pay by the click-through, rather than the impression. For example, Proctor & Gamble made this type of agreement with Yahoo! in May, 1996. The problems with this arrangement are that the publisher takes on all the risk (by giving up valuable space to a banner that may or may not produce sufficient click-throughs) and that they, unlike other media vehicles, are forced to be accountable for advertising response as a result of the unique capabilities of the Internet.

In thinking about web measurement, the following questions may arise:

Can we determine whether an ad was seen by one person 1,000 times or 1,000 people only one time?
If a hit is actually the number of files requested from a site (which may be more than one if a page contains graphics as well as text), how can we accurately measure visits to a site?
How do we know whether the IP address reflects one or multiple users (in a house, company, or school)?
How can we determine whether a user looked at all the information on the web page, particularly if the ad is at the bottom of the page?
How would we know if all the images loaded correctly and the user is able to view all images?
If pages are being cached, how can we count these page views?
Can we tell whether the user interrupted the transfer or left the page through hyperlinks before the page downloads?
How can we measure ad impressions when products that strip out advertisements exist (e.g., Internet Fast Forward)?
How can we know if a user turned off graphics and did not see the ad?
How can we overcome data manipulation issues?

Do you want to know more about the issues plaguing web measurement? The following links will take you to other articles on the subject:

"Stop data spin doctors with standardized traffic counts," (Ad Age Interactive, 6/13/98).
"Net traffic ratings debated," (News.com, 6/12/98).
"Caching comes at a price," (CNET, 1/23/98).
"Getting real about usage statistics," (Tim Stehle, Knight Ridder New Media Center).
"Measuring who's on web is still difficult business," (Internet World, 2/9/98).