Measuring Time Spent On A Web Page

The Monday Note covers the intersection between media and technology and the shift of business models. It is jointly edited by Frédéric Filloux, a Paris-based journalist and Jean-Louis Gassée, a Silicon Valley veteran currently general partner for the venture capital firm Allegis Capital in Palo Alto. Their column appears on CBS News.com each Monday.

How much time is actually spent on websites? New technologies are emerging, starting with time spent on individual pages and drilling down to page segments. Such technologies will lead to improved monetization; they could even spell good news for paid sites. Here is why.

First, display ads. Banners and other modules still represent 30% to 45% of the sector (depending on the market). For a brand, display ads remain the best way to actually be seen on a web page, as it is seen in a magazine or in a newspaper. At least in theory. In fact, there are several catches. The first one is the discrepancy between the size of the average computer screen and the length of the average web page. It takes about 5-6 scrolls to get to the bottom of a page. (Some sites require as many as 25-30 scrolls - the gateway to carpal tunnel syndrome.)

Evidently, not all modules get the same amount of viewer attention. As a result, all modules do not hold the same value for the media, nor do they create the same ROI for the advertiser. However, today, ad spaces are sold roughly at the same price, the main variable being the type of page and the editorial context (home page or article page, in a sport, business or politics section of the site).

The second catch is the advertising module's actual goal: is it supposed to be just seen (Chanel brand awareness, for instance) or clicked-on (win a trip to the Bahamas)? The latter sort, clicked-on, conflicts with the editorial environment. On the one hand, the media (its editors, at least) work hard at making the content compelling, relevant and interesting; therefore, the last thing they want is visitors clicking away and going to the Bahamas vacation site. On the other hand, the advertiser wants editorial context without too much ability to retain attention. In short, so-so content will make visitors more inclined to click on the Bahamas banner. In these conflicting goals lies one of the main problems of Internet advertising: a growing number of advertisers want to pay for performance, i.e., when people actually click on the module.

This makes time-measurement relevant. This week, I saw technology developed by the French company Alenty, created in 2007 by two engineers who once worked for NetRatings. Their previous employer, in itself, makes the story interesting: NetRatings (later acquired by Nielsen) enjoys a dominant position on the Internet measurement market with a system using panels of individuals. In Europe as well as in the US, this type of measurement system is seen as highly questionable. But it is nonetheless a great marketing success as all ad agencies embraced it. Nielsen is the reason for such state of affairs: being the reference in TV measurement, nobody dares question their ability to measure Internet advertising - big mistake.

Alenty spent two years developing technology measuring the actual user viewing time per module. It uses variables such as the visibility of the module in the page (is it on my screen when I read a story for instance?); or was the ad fully loaded in the page, etc. For details, go to this self-explanatory demo.

Last December, Alenty conducted a multi-site analysis on a sample of 120m impressions. Here are some of the more salient findings:

• 27% of banners are not seen at all; either because the user bounced too quickly to another page or site (below a certain threshold the ad is considered not seen), or because it was never on the part of the screen seen by the user. With such metrics, performance varies widely between types of websites: the best are media (yesss!) and services; the worst are games and e-commerce sites where people come for precise and intense search-and-click mode. Weirdly enough, even banners located on the top of pages do not get 100% viewing rates but between 85% and 95% simply because a measurable proportion of users leave the page before the ad is fully deployed (this, in itself, gives an idea of the frenzied zapping behavior seen on the Internet.)

• On average, for each module, ads are viewed for 15 seconds. Again, it varies from a mere 8 seconds for e-commerce sites, to 21 seconds for medias sites (yessss again! It's our day, folks). For the entire page, the average view time is 33 seconds.

• Modules on the footer of a page rate poorly, of course. But it is even worse than expected since Alenty detected rate of viewership as low as 10 to 20%.

• Visibility of a page and click-through rates are not correlated. For instance, you might have a poorly viewed page bottom, but with a decent click-rate. Why? When reaching the bottom of the page, the viewer is through with content, and therefore more prone to click on the banner (better put the Bahamas ad here).

Based on this analysis, Alenty derivates several new concepts. The first one is the real CPM; for example, a module sold at €3 for 1000 impressions (that's the CPM) but viewed only 66% of the time, will in fact bear a ROI cost of €5 for the advertiser. Alenty computes the real cost of a campaign by aggregating variables in a 24-column spreadsheet.Second concept : the result isn't computed in Cost Per thousand of ads impressions but in Cost per thousands of minutes. Results vary widely.

The time-spent metric is used in television. Therefore, it's no surprise to see several French TV networks teaming up in order to evaluate Alenty's technology. (In France, TV is a fiercely competitive business, but networks cooperate when necessary. By contrast, newspapers have a much lower competitive metabolism, they cooperate less on commercial and industrial matters - no comment).

The first one to broadly embrace Alenty's technology was Canal+, the global pay-TV network. When, in France, Canal+ decided to rebroadcast shows on its player it didn't want to get caught in the low CPM quagmire. "We are serving premium content, targeted to upmarket viewers", says Roger Coste, general manager of Canal+Régie, the network's sales house. "We needed something else that could value our content. Measuring Internet advertising by duration and visibility allowed us to guarantee the exposure of campaigns in a rather precise manner and, also, to increase our revenue for each campaign. This had a positive effect on our bottom line: we are able to charge 30% to 40% more this way than by relying on the classical CPM system". This also yielded a better campaign management process: once an advertiser has been served with the amount of time he purchased, its campaign is replaced by another one. No more low yield ads. Hulu, the Internet-only TV network backed by NBC, News Corp and Disney does it, and its revenue is already higher than Google's You Tube.

For the digital news business we are in, time measurement offers a promising future. At first, it will increase the revenue par square centimeter of screen, so to speak. No more dumping large chunks of inventories to bottom-feeding "ad networks" specialized in bulk selling at fire-sale rates - which is the commercial equivalent of slow motion suicide. No more saturating our news sites with ads in a downward yield spiral either. Instead, for our future, we're looking at a smaller number of ads, but paid-as-viewed ads. Everyone will enjoy the change: brands will have better (and guaranteed) exposure in lieu of today's shots-in-the-dark; sites publishers will have more leeway in creating innovative designs; and viewers/readers will be less bombarded with irrelevant and overwhelming ads.

Campaigns purchased on a "view-time" basis will also generate new financing models for rich multimedia contents, such as simply selling "pre-roll" 10-30 seconds screens before a great Flash-based narrative production. (See Monday Note #84 for examples.) It works for TV series on Hulu, it should do fine for The New York Times, Le Monde, or the Nordic dailies Verdens Gang or Aftonbladet.

One more thing. The pay-for-view-time system could solve the paid-for-content equation. Today, we see dictatorial decisions: "This content is worth paying for and this one is not". It doesn't work. Tomorrow, measuring the actual time, in seconds or minutes, spent on a given page (or part of a site) will yield smarter and more credible monetization models. It will open the door to charging for actually viewed content, i.e. the article you've actually read, the multimedia production you spent six minutes on - all this without discouraging exploration, serendipity, which must remain a the core of news consumption.

By Jean-Louis Gassée and Frederic Filloux
Special to CBSNews.com