Why counters are (worse than) useless

David Dorn explains why the figures you see in your counters are misleading at best and dangerous most of the time

In order to understand why the figures you see on your counters are almost definitely not even close to being accurate, you need to understand how the Web works – and the concept of cacheing. When you surf to a Web page, what you’re actually doing is requesting a set of linked files from a potential sequence of servers. The first stop is usually your browser cache…

Browser Cache

99% of web browsers maintain a directory on the host PC into which the currently viewed page’s files are copied. Normally, when you return to that page, in that session, your browser takes the information from that directory – the browser cache. The result is that the Web site sees no access. Its files are not needed.

Local cache

“Local” in this sense means local to the IP block in which your PC resides. Your ISP may implement caches specific to ranges of dynamically assigned IP addresses. If that’s the case, when you request the page, the Local Cache is searched for it files first, and, if they’re there, the Web site doesn’t ge a request for them – so it doesn’t see an access.

“ISP” Cache

Many ISPs (like AOL) maintain cacheing servers set up purely to store pages from the WorldWide Web that their users have accessed. So, if someone from the same ISP as you has already been to the Web site you’re trying to get to, you’ll be served the pages from the ISP Cache – again, the Web site itself won’t be troubled to send the files – so it doesn’t know you’ve had a look. They may cascade through a local cache to get to you.

As you can see, there are numerous opportunities for your browser to find the pages you’re looking for without the Web site you’re trying to get to ever knowing you’ve read it. If the site doesn’t know you’re “been there” (that is, seen the pages) then it cannot add you to the statistics, or increment the counter.

Any count you get, then, is likely to be on the low side.

Can I apply any tricks to help?

If you’re the Web master, there are tricks that will subvert the cacheing process – but I wouldn’t ever recommend that you implement them. The main reason that cacheing exists in the first place is to speed up the Internet, or, more specifically, the Web. Consider that some Web sites are hosted on machines that really couldn’t handle the enormous numbers of visitors that it could potentially get. If local and ISP caches didn’t exist to remove some of the load, the site would slow to a crawl, or maybe crash altogether.

Then there are bandwidth considerations. It’s much less costly to provide you, as a visitor, with a set of files to make up an HTML page on the first hop of your connection than it is to serve them remotely. If, by sending a site once to a cache, you can then send it on to, say, 20,000 visitors, each of whom is dialling directly into just one server, the reduction in bandwidth necessary is massive.

In other words, cacheing helps prevent the Web becoming clogged during periods of peak usage. Using zero-length refreshes and pragma statements to force a cache bypass, then, subverts the whole process, and could lead to the Web grinding to a halt.

Can I apply a multiplier to get the right figures?

Nope! There is absolutely no way of knowing what access figures you do see represent, and no way of estimating what scaling factor to use. About the only thing you can say for certain is that you have had at least the number of accesses your counters show. That’s the  only thing – you can’t even estimate the number of people, since accesses does not equal people and individual IP addresses don’t equal people, either. When you log into AOL, you’re assigned an IP address. Next time, you log in, chances are you’ll have a different IP address. You could, then, appear twice in the stats for a given Web site on any given day – but you’re the same person.

There’s absolutely no correlation between access counts and actual views – there’s no way of determining one, and no way to be even close to accurate about a count.

So why have counters?

That’s the $64,000 dollar question. I would be very wary of having visible counters on any Web site – they’re not going to be even slightly accurate, and your visitors are not likely to understand everything that you do (now) about how they work, and why they can never be trusted. At the very best, they’re only a very rough guide as to whether or not you’re actually getting any visitors at all. To actually base a strategy for your site on a counter would be folly indeed, unless you were absolutely certain that you’d forced a no-cache policy that worked on every last file in your site.

Trust me – that’s all but impossible, unless you know the internal workings of every form of cache that’s in use by every ISP on the planet.

The bottom line is this: Counters are, at best, misleading, and, at worst, totally, completely, and utterly useless. If you absolutely must have one, treat it as a bit of fun, and don’t let your visitors see it.

This page has been viewed 135,790,876,330,749,001 times since you logged on... yeah, right!

^top

Have your say - click here

David Dorn