Fusing samples by MultiBasing

Fusing samples by MultiBasing

The half-life of change

The organizing principle of life has always been change. Science sometimes expresses change in terms of half-life, the amount of time it takes for half a substance to be eliminated by another organism and replaced. For example, the half-life of molecules that make up tissue in the human body is less than two weeks. That's barely enough time to get to the next paycheck. The 95% life of brain molecules is a little less than three months, while over three million red blood cells in our body are replaced every second. So if you are just now wrapping your mind around a new idea like fusing samples together to get better information for planning and buying media, don't settle in just yet. Things are about to change, just as they get started.

The Holy Grail and other cups

Let's start at the beginning. The Holy Grail of information in media planning and buying is single source research. This would measure people's demography, psychographics, all of their media habits (including an accurate picture of TV viewing) and their buying habits in just about every product category. If we could all drink from that cup, we would confidently select television programs to advertise on that reach current and potential customers. Of course, we haven't really figured out how to do that yet. Basically, television audiences are so fragmented today that we need people meters, while measures of all other media and buying habits are voluminously handled by other techniques. No one has found a way to palatably stuff this down the throats of several thousand respondents. So, we have resorted to other methods to integrate data from different samples, in the quest for the grail.

For a long time we matched demos of product buyers with media audiences and planned and bought according to profiles. We still do that, sometimes with pretty sophisticated cross-tabs. Sometimes we ascribe samples, which is a minor league form of fusion, whereby all respondents are questioned in broad areas like demos and some media habits, while only some are queried in detail on product usage and detailed media habits. Then the demos of the some are matched with the demos of all and their special responses are seeded into the whole sample (replicated).

The most advanced state of data integration today is fusion. It has been around in Europe for a while now, where people meter data from BARB has been fused with product data from TGI. Fusion works like this. Several key demographic characteristics like sex, age, income, etc are selected as the best indicators for matching people up with others who act just like them in many other ways. So, respondents in each of two samples are matched up by linking demos and all of their other habits are imposed on each other. The samples are fused.

Don't abuse when you fuse

For a short time there we thought we had almost caught up with Europe by openly exploring fusion. Several major agencies are looking into their own systems, industry groups are reporting their findings and researchers who have done this work in other countries are being imported into the U.S. All this excitement overlooks a flaw in fusion technique, which needs to be addressed (read changed).

Current fusion practice holds that once you have identified the key demographic descriptors that will blend two samples, what is true of people in one sample is true of their counterparts in the other. So, for example, if age, sex and income are the keys, a wealthy young man who plays golf is no more likely to watch a telecast of the Masters Tournament than one who is not a golfer. Being a golfer is not one of the keys to the kingdom. So if you're selling golf clubs, beyond the obvious choice of the Master Tournament, fused data as it currently exists has limited use. This is true of every product category, because fusion is just a more sophisticated step in a long line of a rich heritage of information dances called "it's the best we can do."

MultiBasing common fusion

Remember half-life? We have already reached the half-life of traditional fusion, even before it has begun in the U.S. Fusion is changing here already. There is a new technique developed by Telmar called Fusion by the MultiBasing Technique, which has all of the elements of current fusion and includes other key descriptors, like golfing in the example cited above. What happens is by simply inspecting how much more golfers view the Masters than the average person, that factor is imposed on the respondent data from the fused samples. In concrete terms, MRI, which does not do as accurate a job at measuring absolute viewing levels (since it doesn't use people meters), does give us good measures of relative viewing). So if golfers view the Masters at a rate 50% more than the average person in MRI, that 50% increase is applied to the demographically fused rating from the two samples and the people meter rating from Nielsen. Hyperfusion. MultiBasing.

But there is one more benefit to MultiBasing. The sample donating the rating (NSI sweeps) is linked (kids would say hooked up) with a surrogate sample (MRI) which has product data. MultiBase linking maintains the integrity of each sample, because it does not permanently fuse the two together. This prevents a distortion of the information inherent in each separate sample, allowing the subscriber to dig around for more nuggets. Basically, the samples are fused and rubbed together,

So everyone wins. Advertisers can directly match up their schedules with people who have bought their products or might buy. They can also see if In the past, people who viewed their schedules bought more product, proving their campaigns were successful.. Agencies can plan and buy directly against their clients' customers and they can build a store-house of case histories that track past successes with schedules they planned and bought, showing how viewers contributed to sales. The media can use hyper-fused data to sell prospects on the power of their property by demonstrating how they can directly reach out to customers and competitors' customers. And researchers can refine the tools of fusion, safe in the knowledge that for now at least they have the best possible expression of what's going on out there and can better answer the question "Is my advertising working."