NEW YORK: The lack of precise minute ratings for each ad occurrence in US TV audience data can lead to a 20% under-estimate of television sales lift, research has shown.

An analysis conducted by Sequent Partners and Nielsen for the Council for Research Excellence (CRE), and presented at the Advertising Research Foundation's Audience Measurement Conference earlier this week, found that the problem was an infrequent one.

But in the 10% of the cases it did occur the effect could be significant, reducing television ROI by an average 20% - enough for an advertiser looking at a marketing mix model to rethink the medium's role in a campaign.

"The media data inputs for marketing mix models are frequently drawn from planning or buying systems that often are lacking in the models' requirements for granularity, precision and accuracy," observed David Poltrack, Chief Research Officer, CBS Corp.

"Modelers need more disaggregated data, variability and exposures properly aligned by week in order to tease out the impact of different media on sales."

For TV data, modelers currently rely on Nielsen's AdIntel service, which is based on average quarter hour audience estimates rather than precise minute ratings, according to MediaPost.

Sequent partner Jim Spaeth compared the situation to the "last click attribution" issue that has bedevilled digital media, where too much weight is given to the final point of contact with a user before they click through to take an action and not enough to all the other marketing to which they may have been exposed.

Over recent years, it has become evident that overly aggregated, or estimated, media delivery data doesn't always line up well with weekly brand sales data –leading to under-estimates of ROI and, as Sequent demonstrated, television being sometimes undervalued.

Spaeth indicated that Nielsen, which in the past has been criticised for a failure to measure online audiences properly, was already working to address the problem, but added that more work had to be done to devise better ways to align audience data with actual advertising exposure.

Data sourced from Council for Research Excellence, MediaPost; additional content by WARC staff