Nielsen changes will push down web ratings for all publishers

Nielsen is preparing its most sweeping methodology change since it started rating website traffic in New Zealand.

At the moment, the company uses software cookies to measure every visit to a website in its survey (which covers practically every major site in New Zealand), and other metrics en masse.

But in December, it completed “initial recruitment” of 3000 people who will form a panel used to extrapolate national ratings – not unlike the panel used by Nielsen to measure TV ratings.

Nielsen is now fine tuning the people panel.

From mid-way through this year, it plans to launch a new “Hybrid” Net Ratings service, which will combine results collected from the humans with results collected from the software cookies.

Overall, I like the idea of a people panel (which already operates in the US). The cookies used in Nielsen’s mass survey can be deleted by a user (or, unknown to them, by their network administrator or security software), complicating matters, and inflating results (more of which shortly). A people panel is a much more contained environment, and will mean much less voodoo is required to determine key measures of audience engagement, such as time spent on a site, and the average duration of each visit.

Yet the new system gives me the heebie-jeebies on two levels.

One is that the people panel approach can only measure web activity from one location – home – while some sites are mostly accessed from work (yes, I’m thinking of NBR, but the same logic applies to others).

I put this concern to Nielsen’s commercial director, Amanda Wisniewski, who responded that, in such a situation, reader panel results could be normalised with data pulled from the traditional survey.

Inflated results
Nielsen plans to release monthly traffic figures based on its people panel.

With the cookie-based survey, publishers – and advertisers – tend to focus on domestic unique browsers numbers – but my understanding is that, in point of fact, the current system measures unique browser sessions rather than unique people. So if I access a website from my home PC then, later in the day, my home PC, I register as two unique browsers. If, at either location I, say, close Safari then open the same website through IE, that’s a new unique browser session too. 

Nielsen doesn’t usual make public comment on the blow-by-blow details of its methodology, when I put the above paragraph to Wisniewski, she replied. “Yes, basically you are correct.”

It follows – I would say - that monthly figures under the current, cookie-based system are exaggerated. As cookies get deleted, or people log-on from different PCs, they can be double or triple-counted. It's a problem that gets worse the longer the survey period.

I’m proud of NBR’s weekly domestic unique browser number (which has averaged 40,575 over the four weeks to March 6, fact fans, compared to 25,286 for the same period a year ago).

But I’m dubious over NBR’s sky-high monthly figure (and other sites).

I don’t think this is a particularly controversial take on events.

And indeed Wisniewski conceded that, “Yes we do see some browser inflation across longer periods, particularly in a month (which is why we recommend using average daily unique browsers).  So the fact that we will now be issuing a monthly unique audience (people) number yes it is likely that generally it will be lower than the monthly unique browser number.”