(FORTUNE Magazine) – SOPHISTICATED new techniques of market research are yielding fascinating and sometimes startling facts about how consumers behave and how advertising works -- or doesn't. The new methods, introduced a few years ago, have begun to prove their value in an era when familiar brands are dying, new brands are flopping, and all brands are having a tougher time getting on cramped supermarket shelves. On this market research frontier, some of Madison Avenue's most venerable verities are being disproved. Most surprising finding so far: Trying to goose sales by a big increase in a product's TV ad budget often doesn't pay. The buzzword for what this research produces is single-source data. The ungainly name is meant to tell you that the research tracks the behavior of individual households from the TV set to the checkout counter. Chicago-based Information Resources Inc. pioneered the technology (FORTUNE, July 25, 1983) and monitors more than 3,000 households in each of eight small-town markets, from Pittsfield, Massachusetts, to Visalia, California. In the sample households, microcomputers record when the television set is on and which station it is tuned to, sending this information by telephone line to a central data-collection office. Sometimes IRI sends out special test commercials over cable channels. At the supermarket checkout, a member of a sample household presents an identification card. The checker records the purchases by scanner, and the data are sent to IRI. ''You put all that information together,'' says Gerald Eskin, vice chairman of IRI, ''and you have more insights into people's buying patterns than ever before.'' IRI's revenues have nearly tripled to $75 million over the past four years. A.C. Nielsen Co., a unit of Dun & Bradstreet, grabbed a piece of this fast- growing market last July when it began testing a similar operation involving 3,500 households in Sioux Falls, South Dakota, and the same number in Springfield, Missouri. Nielsen can send test commercials over the air, so members of the sample need not be cable subscribers. More competition comes from ScanAmerica, a 200-home pilot project launched last November in Denver. ScanAmerica is a joint venture of Control Data's Arbitron Ratings Co., which measures TV viewing, and Selling Areas-Marketing Inc., or SAMI, a subsidiary of Time Inc., publisher of FORTUNE. Unlike the other two services, ScanAmerica does not send test commercials to homes of sample viewers. Instead it simply monitors sample households' viewing and consumption patterns and plans to sell the data. ScanAmerica participants record purchases at home by passing a penlike wand over goods bearing bar codes. The information is transmitted daily through a device attached to participants' televisions. The single-source services are reshaping advertising. Graham Phillips, chairman of Ogilvy & Mather U.S., frankly acknowledges that ''we've changed major campaigns and spending levels for several of our clients'' after single- source tests. A while back, for example, such tests suggested the agency shift ads for Campbell Soup's Swanson frozen dinners from a serious to a light approach; Campbell marketing executives subsequently clocked a 14% rise in sales of Swanson dinners. MAJOR ADVERTISERS like Procter & Gamble, General Foods, General Mills, Campbell, and Quaker Oats say they have gathered a valuable crop of previously inaccessible market data through the services. By correlating test panelists' TV habits and grocery purchases, for instance, Campbell discovered an extraordinary appetite gap between viewers of two daytime serials: Search for Tomorrow fans buy 27% more spaghetti sauce than average, but 22% less V-8 vegetable juice; devotees of All My Children feel so-so about spaghetti sauce -- but purchase 46% more V-8 than the viewer norm. Why the divergence in tastes? ''We've been looking for an explanation, trying to unlock this thing,'' says George Mahrlig, Campbell's director of media services, ''but it's still a big mystery.'' Mahrlig says Campbell has pitched the virtues of Prego spaghetti sauce on Search and those of V-8 on Children and will probably do so again, a low-risk, high-gain strategy aimed at maximizing sales among the likeliest customers. Marketers also say scanning lets them take chances on long-shot new products that might not rate an expensive trial. A regional market test could cost $2 million, for example, vs. a tenth of that for a test through a scanning service. Quaker Oats recently tested an advertising campaign ''for a business where we weren't really very hopeful,'' says John Blair, vice president for marketing research. ''Maybe in one case in ten do you move the needle -- and that's what happened here, in a big way.'' For competitive reasons Blair declines to identify the product, but he says the scanning test's ''strong sales response'' has persuaded Quaker Oats to take the product national soon. The scanning services' speed and cheapness pay off in other ways. After only an eight-week trial last winter, Quaker Oats decided on a national launch of Oh!s, a new dry cereal. ''Electronic scanners helped us to decide to move ahead much sooner than usual,'' says Blair, ''and in a highly aggressive manner.'' The services are just as valuable when they point out duds. A negative but accurate verdict that kills a product quickly is preferable to a long, expensive, and perhaps inconclusive regional trial of a product that will fail. ''Scanner data give you an earlier read on new products,'' says D. J. Jefferson, market research manager at Kitchens of Sara Lee. ''You get the data store by store and week by week.'' In addition to other tasks, clients use scanning services to test so-called advertising weight, meaning spending levels. About half the time the results have been shocking. According to the tests, even doubling TV spending from, say, $5 million to $10 million a year may not pay. Says a somewhat rueful Campbell Soup Co. market manager: ''It doesn't move the needle.'' Advertisers may test a $10-million ad campaign with one panel and a $5- million campaign with another. Such tests typically take a year. ''If it turns out that the people who received the $5-million campaign bought the same amount as the $10-million group, that raises the possibility that a reduction in spending might be in order,'' says Randall Smith, president of IRI's database division. Richardson-Vicks decided against a stepped-up media schedule for Olay Beauty Cleanser, says Jurgen Fey, director of marketing research in the personal care products division, after IRI tests in Pittsfield and Marion, Indiana, ''made clear this would not increase consumer offtake leading to a payout.'' Translation: The extra ads wouldn't pay for themselves. Most marketers find the scanning data persuasive. Unless an ad campaign is changed or a product repositioned, notes a market research executive, ''increased advertising is usually not sales effective, and that runs counter to the intuitive judgment of an awful lot of people.'' After supervising more than 450 ad tests, IRI executives say that usually an ad campaign's copy and focus matter more than the number of times a spot runs on TV. THIS NOTION is understandably spooky to people who work in advertising agencies, which generally get paid a percentage of what their clients spend. ''It's frightening,'' admits Ronald Kaatz, director of media concepts at J. Walter Thompson U.S.A., ''but brands double their ad budgets in most of these tests, and nothing happens to sales. That says there have to be a lot of other variables out there.'' Agency executives, protecting their own interests, point out some valid considerations. Slashing an ad budget solely on the basis of a relatively brief panel test may be a risky strategy. Brand loyalty built over the years might not be eroded by a one-year drop in advertising. But the brand's long- term franchise with customers could suffer from a sustained advertising decline, they say. In addition, even carefully constructed market tests can be flawed. The scanning services don't know if viewers are actually watching a television set when it is on, or whether a commercial break will send them to the bathroom. Test participants, who are paid volunteers, might also skew the numbers: They may react in atypical ways simply because they know they are sending a message to advertisers. ''The most serious problem with this kind of data is it presumes you can track a commercial on a cash register,'' says Dave Vadehra, president of Video Storyboard Tests, an advertising research firm. ''It takes years to change buying pattterns.'' Only for new products is the single-source process ''a pretty viable concept,'' says Vadehra. Since the product had not existed, initial purchases can be directly linked to advertising. Other advertising executives note that single-source testing has not been perfected. Tests in small-town markets cannot predict nationwide results, they say. And single-source tests, for all they tell, do not answer the all- important question of why people are motivated to buy a product. ''Certain TV programs appear to be the best vehicles for selling your products,'' says Laurence R. Stoddard, an advertising consultant. One of those programs may work fine, but ''then you try it again and find it doesn't sell anymore.'' Other commercials, coupons, and price cuts are constantly roiling the market scene, he explains. Against that background, even ad experts often find it hard to tell why any ad succeeds or fails. SOME advertising executives on the creative side say scanner testing could lead to more distinctive commercials. ''One way to break through the commercial clutter today is to have a commercial that's better than the next guy's,'' says Malcolm MacDougall, president of the Hill Holliday Connors Cosmopoulos ad agency's New York office. Scanner tests should ferret out the winners from the losers faster than more subjective methods, such as interviewing a ''focus group'' of consumers. ''Sales numbers are the only way to judge advertising,'' says MacDougall. Where high-tech testing is headed still is not clear. The three main single-source contenders all have drawbacks. While the IRI system monitors network viewing, it can only test commercials over cable systems. This gives IRI's sample a slightly upscale tilt, according to Nielsen, whose system is the only one that can cut into the signals of regular over-the-air stations with test commercials. Nielsen's critics note the system is operating in but two markets and registers only viewing habits of entire households, not individual members. In contrast, ScanAmerica uses ''people meters'' that keep tabs on each viewer -- but only if the viewers remember to push buttons on hand-held devices that register their presence before the tube. ScanAmerica's wand lets participants keep track of all bar-coded products they buy, not just supermarket purchases, but panelists might forget to use it. Market researchers cheerfully concede their brave new world is still in transition. Laurence N. Gold, a marketing vice president at Nielsen, says national television viewership data in the future may not be measured through the current method of written diaries or people meters. Years from now, says Gold, sensors in homes will automatically detect and report who is watching TV -- and perhaps even transmit physiological reactions of viewers to programs they're watching. Such fanciful projections are far removed from the limited but valuable lessons being learned from the current technology. ''It has helped us run our marketing operation more efficiently,'' says Rudolph Struse, General Mills' director of corporate marketing research. He cautions, however, that he has not yet gleaned any general principles of what makes effective advertising. Other marketers agree. Case by case, they can find out whether campaigns have a reasonable chance to succeed. But none claims to have uncovered the key to what makes people buy. ''The one thing we have learned is, there is no magic formula that works on everything,'' says Ogilvy & Mather's Graham Phillips. Adds John Blair of Quaker Oats: ''All we can do is bring as much science as possible to what really is an art.''