• by Steve Smith , Staff Writer @popeyesm, July 25, 2008

Nielsen's many panels, research and measurement products now span almost all consumer touch points, so increasingly the company is exploring ways of integrating and overlaying the data to generate new insights. A series of "fusion" studies recently from Nielsen Connections, for instance, tries to map actual product buying patterns against TV and Internet usage. As Howard Shimmel, senior vice president, client insights, explains, this is the kind of data that gives media planners much more precise methods of targeting their TV buys. His colleague, Pete Doe, managing director, Nielsen Connections, offers a crash course in fusing databases.

Behavioral Insider: What holes does this fusion data fill for your clients? What can they see through these studies they can't see elsewhere?

Howard Shimmel: A direct link between marketing targets and media usage. A brand manager at Procter on one of their Gillette men's razor brands has a lot of great data that comes from syndicated services like our HomeScan or their own proprietary research. They have a very narrow marketing target.

What ends up happening is that gets handed to their media agency, and that gets converted from this very narrow marketing target. Some of it may be behaviorally based, like someone who uses a competitor's razor or is  a heavy razor buyer. Some of it may be attitudes.

But that target gets handed over to their media agency, and the agency says, for the purposes of executing television, we think that looks like men 18 to 49 in households with income over $75,000. It gets to this big broad target that has been used for the past 30 years to transact television. And the agency picks out programs and negotiates buys and tracks rating delivery against that [broad target].

Through the fusion of HomeScan as an example, we are allowing Gillette and their agency to go in and say, let me figure out what programs actually do best against heavy razor buyers who don't buy the Gillette brand. Let me go right to the marketing target and look at real TV ratings data against that marketing target. The key thing is, we tend to see they can get anywhere up to 10% to 20% increase in delivery against their marketing target if they are making media decisions based on what programs, networks or day parts do well with that marketing target.

BI: How long has this feature been in the market -- and for how many clients thus far?

Shimmel: On the consumer packaged goods side, we have been in the market for a couple of years for more than ten clients. Pete and I are bringing this application to media companies now because we are hearing that media are looking for more relevant metrics to show to advertisers. These are more relevant metrics than just submitting plans on traditional demographics. We're also broadening this out to additional categories. So we have a product in the marketplace now that is for the movie studios. Think about a category that needs to make sure they get their advertising right. They have one swing at bat in a really tough, competitive environment. We're going to bring a product to market over the summer for the pharmaceutical companies and financial services. 

BI: What is the methodology behind the Fusion products? You essentially put together multiple databases?

Pete Doe: Correct. The foundation of all of this is the National People Meter. Around 15,000 households are reporting TV [usage] on a daily basis, around 24,000 persons... And that is the basis for all of these products. On top of this, we are overlaying this Internet database that comes from Nielsen's NetView panel, which is similarly sized measuring home and work Internet use. Then we have the movie benchmark studies conducted by Nielsen Entertainment for about 3.600 adults, 18+, [which] gives insights into people's movie-going habits and preferences. Then there is the HomeScan panel run by 100,000 households that tracks their consumer purchases via a scanner. They scan each article they bring back from the store. Nielsen Mobile has a monthly survey of around 3,000 individuals aged 13+ where we track their mobile Internet usage.

BI: I gather from these various databases you are not necessarily talking about the same consumer monitored across all those platforms. This is in place of a single source study.

Doe: They are different respondents. We have separate research studies and panels. It is our job to statistically match the various surveys in as accurate a way as possible based on the common characteristics. The obvious ones are demographics and household characteristics. But we also want to go beyond that and include TV usage information where we have it on the other surveys or the People Meter. A key point we are always making is that for cross media surveys to be reliable, you need to go beyond demographics. You can have two people with identical demographics and household characteristics but one may be into sports and one may not and they will have quite different media usage patterns. So picking up media usage information is very important in integrating cross media.

BI: Explain what data fusion is.

Doe: It is a technique that has been around for decades now. It was invented in the United States but seemed to be adopted more readily and earlier in terms of media usage in Europe in the late ‘80s and early ‘90s. There is a wide body of work and literature around its best practices. I would argue that the results obtained from a good data fusion are actually going to be better and more reliable than from a single source survey that may suffer from very poor cooperation rates or measurement techniques that are not quite as reliable. You can get single source by asking people what they do in terms of TV and product use, but it is better to meter TV and product use and then fuse them together.

BI: I gather it would be impractical to have consumers who are being monitored at every touch point.

Doe: Yeah that's it. One of the points I would make as well is that there does seem to be in some cases a schism between data fusion and single source, and we are trying to bridge that gap. Our philosophy is that data fusion and single source can work well together. One thing we have created within Nielsen Connections is our convergence  panel. It has some single source measurement. It allows us to assess the TV and Internet usage of people in the same home.

BI: Whenever we start using actual behavioral data to find affinities, it always produces unanticipated results. Is any data popping out at you as surprising nuggets?

Doe: One thing I noticed when we started doing Internet TV fusion, it was clear that within homes that had access to the Internet, heavier TV viewers tended to be heavier Internet users. And it turns out that lighter TV viewers tend to be lighter Internet users.  Looking at the TV viewing of women 25 to 54 who were weight-loss-product users, there was a lot of kids' programming. Thinking about it a bit more deeply, it seemed related to having kids. It made sense when you thought about it -- but [was] not something you would immediately come up with

Shimmel: Looking at the movie database pretty closely, one of the things we have seen is that you wouldn't necessarily expect early-morning network television (‘Today,' ‘Good Morning America') to do well with the movie audience. It's not a place that attracts a lot of dollars. [Movie] dollars tend to go to prime time and sports and late night. And we have seen that those morning day parts work pretty well for a bunch of different movie genres. Insights like that are going to sustain the use of this data in the marketplace.