All Things Web: How Much Is Too Much?

	Text Index	Site Search	Feedback
	Home		Next

Contents

The Usable Web

Need for Speed

Interface Elements

Testing &
Quality Assurance

Compatibility &
Accessibility

How Much Is Too Much?

How large is the average page on the WWW?

What percentage of Web pages contain broken links?

Do you think you know the answers to these questions?

You might just be surprised.

This ATW feature reports the results of an automated Web "survey" that provides answers to these very questions. Motivated readers may wish to take a moment now, before reading further, and come up with their best estimates for average page size and the percentage of pages containing at least one broken hyperlink.

The GVU WWW Survey

One of the most authoritative and often cited sources of data about Web users comes from the Georgia Institute of Technology GVU WWW surveys. Currently in its seventh incarnation, the GVU survey reports all kinds of interesting (if not necessarily reliable) factoids about user demographics, usage patterns, and browsing behaviors.

The GVU survey has always had a "gripe" section -- where users identify the things that are most "wrong" with the current Web. In the current GVU survey, the top two reported problems are speed (the perennial favorite) and broken links. A majority/near majority of users identify these as the two single largest problems on today's Web.

Okay, but How Slow Is Too Slow?

The generalizations from the GVU surveys are valuable in their own right. After all, knowing that slow load time equates to user dissatisfaction serves as the basis for a moderately handy general Web design tip. Readers want smaller, faster-loading pages.

But the GVU results also raise (but don't answer) somewhat more specific questions, such as: "Just how big is too big?" and "Just how many broken links is too many?" In other words, it's fair to query the "State of the Web," in an effort to quantify specific thresholds for current user dissatisfaction.

The State of the Web Survey

Measuring the "State of the Web" is a technologically straightforward task. So ATW constructed its own Web "spider" and sent it out in search of data that would describe the current state of the Web. The spider was sent to discover both total page size as well as the incidence and prevalence of broken hyperlinks.

A total of 44 pages were selected at random from the AltaVista search engine. (Readers unfamiliar with sampling theory may be tempted to think that 44 pages seems awfully small. In fact, it's probably large enough (statistically speaking) to serve as an approximately representative sample of the larger Web. Arcane details regarding sample selection are available for methodologically inclined readers.)

Page Size

Total page size was actually somewhat lower than anticipated. Still, the obtained result represents a compelling if uncomfortable design guideline for Web authors. Average total page size (text and graphics) was 44 KB. In other words:

7 out of 10 Websurfers agree:
44K is too big.

Dead Links

The "linkrot" results can be approached in one of two obvious ways. One approach is to compare the number of dead links against the total number of links in the sample (incidence). The second approach is to count the number of pages containing at least one dead link (prevalence).

The incidence of linkrot was surprisingly low. The pages sampled by the SOW survey included several hundred links. By any reckoning, less than 3% of those links were "dead."

But the prevalence is linkrot is much more pervasive. Approximately 18% of the pages examined contained at least one dead link. (It seems virtually certain that it is this prevalence, rather than the incidence, of dead links that is the source of user unhappiness.)

Ironically enough, the most likely source for dead links was -- AltaVista itself! Nearly 10% of the URLs returned by the search engine pointed to either a nonexistent page or a nonexistent server.

Digging Deeper

One of the more noteworthy aspects to the survey results is that they tended to "cluster" fairly strongly into two groups. Regarding page size, there were more than twice as many "small" pages (30 K or less) as "pigs" (over 90 K).

Similar "clustering" was found in the results pertaining to dead links. While the vast majority of survey pages contained precisely zero dead links, the worst "offender" offered its readers a 50/50 chance of a new page or a "not found" error, depending on which link they tried to follow.

These results may be grounds for cautious optimism. They offer a suggestion that serious-minded, conscientious Web designers are beginning to "get" the message that smaller pages make for happier readers. Similarly, they suggest that the majority of site administrators are making a conscious effort to keep their pages free of dead links.

It's Not Fair, But...

Several sites offer suggested page size limits based on conventional wisdom in human factors research. Most of these sites exhort authors to keep total page size under 30 K. (The most obvious dissenting voice in this regard is ATW, which prefers a more modest 20 K limit for its guidelines.)

Of course, page size is not the only element in the response time equation. Fair or not, readers can't help but assess the overhead imposed by nameserver lookups and the initial server connection against your page. Increasingly slow responses from an overburdened Internet infrastructure can severely compromise the apparent "load time" of even the most spectacularly well-optimized Web pages.

But the results of this State of the Web survey make it clear: nearly one page in every five presents dead hyperlinks, and average page size is at least 150% above the upper limit implied by human factors research.

No wonder readers are complaining.