That famous John Wanamaker quote about half of advertising spending being wasted may not apply to the Super Bowl : A study by research firm Communicus suggests 80% of the ads don't sell stuff.

Indeed, the $4 million advertisers spend for a 30-second Big Game ad actually buys a much bigger chance that their ads won't work, according to the Tucson, Ariz.-based firm. In general, Communicus has found about 60% of ads it tests don't increase purchase or purchase intent.

The firm reached its Super Bowl conclusions through interviews with more than 1,000 consumers before and after they were exposed to the ads in the 2012 and 2013 games. Before the game, Communicus asks people what they've bought recently and what they intend to buy in the categories spanning the Super Bowl advertiser base. Then it follows up with the same group within the next couple of weeks and asks similar questions as well as others in order to gauge whether consumers actually saw Super Bowl ads for various brands. For categories with longer purchase cycles, such as automobiles, Communicus uses other measures such as changes in intention to investigate buying a brand.

CEO Jeri Smith said it takes its first read on a typical ad four weeks after it starts running to allow for multiple exposures. She said one thing that she suspects hurts Super Bowl ads is that many of them don't air regularly after the game and "we find that one ad exposure often isn't enough to make anything happen."

She believes there are other factors at play, too.

"The advertisers really dial up the entertainment quotient to pop to the top of the USA Today rankings and such," she said. "But we find the brand association with Super Bowl commercials is much lower than you'd get with a typical buy, just because of the way the creative is structured."

Super Bowl ads actually do better than average in ad awareness, with 44% of people remembering they've seen an average Super Bowl ad vs. 32% for other ads that get similar gross-rating-point exposure, Ms. Smith said. But because the creative often focuses less on the brands, she said, people not surprisingly remember the brands less often in Super Bowl ads. People who remember seeing a Super Bowl ad recall the brand 35% of the time vs. about half the time for other ads, she said.

The Super Bowl works better for new products, Ms. Smith said, because the ad message tends to be clearer. Yet, some brands without any real news have fared well too.

Budweiser's "Brotherhood" ad from the 2013 game, showing the relationship between a Clydesdale and its trainer, fared well in terms of entertainment value and increasing purchase intent, she said, because the iconic horse is so well linked to the brand and the creative was compelling. "Beer is an affinity product," she said. "I want a beer that makes me feel good about myself."

But Tide's 2013 "Miracle Stain" ad, which was well liked and did well on branding, still didn't increase purchases or intent, according to Ms. Smith. "It didn't tell people anything they didn't already know," she said. "And unlike Budweiser, I don't buy Tide because of my personal connection."

Long-form commercials such as Tide's 60-second spot "run the risk of people being so caught up in the story that they forget about the brand," Ms. Smith said. And in "Miracle Stain," the brand appeared at the end as "kind of the anti-hero" by eliminating a stain that looked like former San Francisco 49ers Quarterback Joe Montana.

Meanwhile, last year's Mercedes spot did well because it drove home the message that the brand had a more affordable model coming and helped boost interest in the brand more broadly. But most of the automotive commercials "go to the bottom of the list in terms of effectiveness because they all run together in people's minds," she said.

Communicus uses some of the same purchase-comparison metrics copy pre-testing firms use, but gauges the ads' effect on consumers who've seen them (or not seen them) in the real world rather than through forced viewing. Asking about purchase intent is not precise as sales data but is superior in isolating the impact of the advertising with the individuals involved, since they're going back pre- and post with the same individuals and the only real variable is whether they saw the ad (given that both groups are just as likely to have been influenced by promotions or other factors).