To be precise, trying to find a single set of metrics to compare the audience of different types of websites it insane. That is why Nielson/Netrating’s decision to measure time spent on websites, rather than page views, does not really matter much. Those looking for a single measure across all websites are reducing a mistake of the dotcom boom.

The mistake is to think of all business websites as internet businesses. This is as ludicrous as thinking of all companies that accept customer orders by phone as telecoms businesses.

It is not rocket science, but since people seem to keep making this mistake, I think it is worth repeating the lesson.

A retailer is a retailer regardless of whether they operate on the high street, over the phone, or on the net. They are not in the same business as a bank because they have similar locations, or operate through websites.

While I am correcting wrong impressions, it is worth pointing out that the web is not the internet. The internet delivers various services: email, file transfers, chat an instant messaging, remote logins, and countless others, of which the web is just the most visible.

It has always been true that the internet as been a means of delivering lots of different things. As the web has become more sophisticated, it has also splintered into multiple different media:

• Traditional websites: primarily written content. Advertising is similar to magazine ads, spatially separated from content.
• Video and audio, which can also have ads embedded before, after, or during the main content, temporally separated.
• Web apps: computer software delivered on line, like Google Apps, or even webmail.

Now consider this. If some had suggested, before the internet came along, that they could devise a single metric for measuring the popularity of magazines, TV, shops and computer software, how sensible would you have thought it?

The relevant measures for a content website are similar those for a magazine. We can measure visitor numbers and circulation. They are subject to different errors, but they are trying to measure the same thing. We can also measure page impressions for a website, which we do not do for magazines — but we would if we could!

Similarly viewing or listening hours are relevant to a video website like You Tube, as they are for a broadcaster. Admittedly reach is also relevant here.

Furthermore the same measures used on different sites are not comparable. It is not difficult for a visitor to a video site to spend five minutes there but only seeeing a single brief ad at the beggining. Someone spending five minutes reading text, could easilly have seen several ads. On the other hand, the ad in embedded in a video forces you to pay attention to it.

As I said all this is very basic stuff, but it is remarkably easy to forget when faced with lists of apparently authoritative numbers.