The Valuation of Online Ads - up in the air again.......

Neilson is going to de-emphasize the Web 1.0 staple, the page view and look at other - woollier in our view - measures of Ad effectiveness going foward.

"Based on everything that's going on with the influx of Ajax and streaming, we feel total minutes is the best gauge for site traffic," said Scott Ross, director of product marketing at Nielsen. "We're changing our stance on how the data should be" used.

Nielsen will still provide page view figures but won't formally rank them. Ross said page view remains a valid gauge of a site's ad inventory, but time spent is better for capturing the level of engagement users have with a site.

Ranking top sites by total minutes instead of page views gives Time Warner Inc.'s AOL a boost, largely because time spent on its popular instant-messaging software now gets counted. AOL ranks first in the United States with 25 billion minutes based on May data, ahead of Yahoo's 20 billion. By page views, AOL would have been sixth.

"It is not that page views are irrelevant now, but they are a less accurate gauge of total site traffic and engagement," Ross said. "Total minutes is the most accurate gauge to compare between two sites. If [Web] 1.0 is full page refreshes for content, Web 2.0 is, 'How do I minimize page views and deliver content more seamlessly?'"

For example, he said, MySpace may have 10 to 11 times more page views than YouTube, but myspace.com users spend only three times more minutes on the site, Ross added. Therefore, measuring total time spent on a site will make it easier for advertisers to mold their ads to how users are actually accessing content, he said.

I'm not sure this is the right way to go.......as Read/Write notes:

On balance I think it will be a step forward if Nielsen does indeed drop page views for 'time spent on site' in its rankings.

It's not yet a totally satisfying change, because with the likes of Google you want to somehow measure relevancy and with blogs you want to measure engagement. But it's at least a step away from page views, which have become too easily exploited - not just by some blogs, but also by the likes of Facebook and MySpace (which both make the user go through extra clicks to get to what they want). What do you think of this change by Nielsen?

The question we have is - how long do you need to "clock" an Ad. For eg, mean user time on Broadstuff is about 5.5 minutes - assuming we had Advertising, any extra time spent on site is probably irrelevant once a particular Ad has been "clocked", so volume must still be a part of the measure. Also, looking at Broadstuff for example, a huge amount of the "pageviews" are hoovered off via RSS and email - do we attach an estimated "average time", or do we just assume its been read and clocked.

The way we would think about it is to:

(i) Work out the minimum "human clocking time" for the type of Ads served
(ii) Calculate all impressions of all types above this level, discard all below - and do research on real usage % of RSS and email feeds
(iii) Ignore time above that cycle per Ad....just because I look at a page for 5 minutes doesn't mean the Ad got any more attention than 1 minute.

This should kill the MySpace et al 6 clicks to freedom play...each of them will be over in a second (or will we now see huge arguments for the subliminal value of Ads?)

Just using time is as open to abuse (or over-favours rich form media) as just measuring impressions.

(This applies to display Ads only, clearly click through Ads are measurable via activity levels)

(Postscript - A further thought - the problem with metrics is people optimise towards them, so the question is which set of metrics will give the least unpleasant user experience? We suspect a combination has to be used or else either one is sent on to interminable click-throughs, or else inteminably long "stickitricks")

Post - Postscript...this post got picked up on Techmeme so I've been following the discussion with growing frustartion - loads of people criticisizing Neilsen, but no one putting up an alternative - this post is typical - and until that happens we will be stuck with the imperfect metrics we have. If one cannot suggest a better system (as we have tried above), how about not posting?

Unless of course the point is not to solve, but to pontificate for points