A GRP is valuable in a model where you can purchase media on a cost per point (CPP), where your ad dollars are valued against the share of the audience being exposed to media. The GRP model works in television (or has worked to date) because Nielsen provides ratings estimates on the size of the audience and the type of audience who is viewing television programming. Nielsen's reporting is based on a sample size and estimates total U.S. audience for this type of programming. If the system is accurate, then a GRP is indeed valuable and can be used to estimate the cost and impact of an advertising unit in this form of media.
The problem is that Nielsen data is becoming less and less valuable at an alarming speed, as consumers' media consumption behavior changes.
As the model shifts from a linear format of content delivery, where the network dictates when a show will be available (i.e., "Lost" on Wednesday at 9 p.m.) to a non-linear format (i.e., "Lost" initially airs on Wednesday but is also available on-demand, online and in iTunes), the system for estimating total viewership will need to change. Nielsen is beginning to revise its methodology to account for DVR and TiVo usage, beginning to estimate the size of the audience that time-shifts the viewing of the show within the current television format, but it is missing the audience who views these programs on other formats.
To use as my example, the show is currently the 10th ranked show on television, according to Nielsen, generating approximately 10.6 million viewers each week. However, the audience for "Lost" is certainly tech-savvy and probably not all tuning in at that time. This hypothesis is justified when you examine the number of people watching "Lost" on-demand at ABC.com (numbers not available currently) or the number of people downloading "Lost" on iTunes each week (as of last week when I checked, episodes of "Lost" accounted for 6 of the top 20 most downloaded shows on iTunes). It's very possible that if aggregated, the entire audience together for "Lost" across all available media formats, the show may be watched by a large enough audience to warrant it being considered the 6th or 5th ranked show. Of course there will be duplication among the audiences, but right now we can't prove it one way or the other!
One more example, though slightly different, can be seen in the U.K. pop charts over the last few weeks. A song called "Crazy" by Gnarls Barkley (a combination of Danger Mouse and Cee-Lo from Goodie Mob) reached No. 1 on the U.K. pop charts--before the album was even out--solely on the strength of the sales from Internet downloads. The methodology for how the charts were scoring sales changed recently to include digital downloads, and we now have a better idea of what U.K. kids are listening to.
My point with all of this is that a GRP only works if the model for buying is on a cost per point. If we revise the counting and measurement methodologies to reflect actual delivery or actual exposure in the way that we currently buy online, then an estimated share is no longer applicable, and we can revise the model based on impressions or OTS (opportunity to see). A few years ago, I tried to revise the buying model to reflect cost per visitor rather than cost per impression, and for a little while that worked, too, but it because onerous and didn't have the strength to stay in place, nor does the GRP--not until the methodology for measurement from Nielsen includes all opportunities for exposure and actual viewership.
The bottom line is, that while I believe we will be able to develop the methodology for an iGRP in the next year or so, I see its use as short-term and not applicable in the next five years. The speed of change in the last year has been astonishing, as a number of organizations are looking into easy-to-measure engagement and other elements, so how much longer can we keep trying to evaluate a GRP?