When media buyers purchase inventory from television networks, they typically rely heavily on Gross Rating Point (GRP) metrics which are designed to reflect reach.

And for good reason: if you’re buying media to get in front of consumers, you need to know how many consumers you can reach through various media channels and properties.

One company, Nielsen, dominates the television market, and its GRPs heavily influence how billions upon billions of dollars are spent each year.

Needless to say, the stakes are high and any mistakes could have a notable impact.

Some media buyers are dealing with that reality today, as Nielsen discovered that some of its GRPs have been inaccurate since the end of January.

Joe Mandese of MediaPost’s MediaDailyNews explains:

In the notice it sent to clients Wednesday, Nielsen said the problem began when it implemented a new method for calculating average audience estimates in its main TV ratings processing system beginning in February, and that it appears to have been delivering “incorrect data for time-shifted data streams” since Jan. 31.

Nielsen’s so-called C3 ratings, which include time-shifted viewing for three days after a program and its advertising have aired, is the official currency for the national TV advertising marketplace, and the basis for most TV advertising buys and audience guarantees.

Nielsen said it became aware of the problem when clients “inquired about elevated average frequency levels” for the time-shifted viewing data, indicating that the glitch may have over-inflated TV audience estimates.

As Mandese observes, the timing of this mistake wasn’t the best, as Nielsen is currently pushing a new GRP-equivalent for online properties.

Nielsen’s television GRPs is calculated based on Nielsen’s television panel. While there’s nothing inherently wrong with panels in theory, they aren’t perfect.

One of the biggest criticisms of panels, for instance, is that they may not be representative of the audiences they’re being used to measure and analyze. More generally, statistical sampling is always subject to error of various kinds.

So are panels harmful, or useless?

No. But when it comes to digital media buying, Nielsen’s flub reminds us that panel-based data and analyses shouldn’t be trusted blindly. It may not be possible to measure audience directly on television, but online, panels are definitely not the end all and be all.