ONLINE advertisers are not lacking in choices: They can display their ads in any color, on any site, with any message, to any audience, with any image.
Now, a new breed of companies is trying to tackle all of those options and determine what ad works for a specific audience. They are creating hundreds of versions of clients’ online ads, changing elements like color, type font, message, and image to see what combination draws clicks on a particular site or from a specific audience.
It is technology that could cause a shift in the advertising world. The creators and designers of ads have long believed that a clever idea or emotional resonance drives an ad’s success. But that argument may be difficult to make when analysis suggests that it is not an ad’s brilliant tagline but its pale-yellow background and sans serif font that attracts customers.
The question is, “how do we combine creative energy, which is a manual and sort of qualitative exercise, with the raw processing power of computing, which is all about quantitative data?” said Tim Hanlon, executive vice president of VivaKi Ventures, the investment unit of Publicis Groupe.
“I think it’s clear that the traditional process of agencies is clearly not going to survive the digital era without significant changes to our approaches,” Mr. Hanlon said.
The push to automate the creative elements of ad units is coming from two companies in California, not Madison Avenue.
Adisn, based in Long Beach, and Tumri, based in Mountain View, are working both sides of the ad equation. On one, they are trying to figure out who is looking at a page by using a mix of behavioral targeting and analysis of the page’s content. On the other side, they are assembling an ad on the fly that is meant to appeal to that person.
Both companies assume there is no perfect version of an ad, and instead assemble hundreds of different versions that are displayed on Web sites where their clients have bought ad space, showing versions of an ad to actual consumers as they browse the Web.
That might lead to finding that an ad for a baby supply store is more popular with young mothers when it features a bottle instead of diapers.
(Adisn and Tumri both measure the ad’s effectiveness based on parameters the advertiser sets, like how many people clicked on the ad or how many people actually bought something after clicking on it. They compare those with standard ads they run as part of a control group.)
Adisn’s approach has been to build a database of related words so it can assess the content of a Web site or blog based on the words on its pages.
Adisn then buys space on Web sites, and uses its information to find an appropriate ad to show visitors to those sites. If a visitor views pages about beaches, weather and Hawaii, it might suggest that the visitor is interested in Hawaiian travel.
Based on that analysis, Adisn’s system pulls different components — actors, fonts, background images — to make an ad. For example, it might show an ad with a blue background, an image of a beach, and a text about tickets to Hawaii. “Once we’ve built this huge database of hundreds of millions of relationships” between words, said Andy Moeck, the chief executive of Adisn, the system can “make a very good real-time decision as to what is the most relevant or appropriate campaign we could show.”
Simple Green, the cleaning brand, began working with Adisn this year to advertise a new line of products called Simple Green Naturals.
“If it’s a woman looking at a kitchen with a stainless steel refrigerator, they can show a stainless steel product,” said Jessica Frandson, the vice president for marketing for Simple Green. While Ms. Frandson gave Adisn a general idea of what she wanted, she also let the agency do almost random combinations with about 10 percent of her ads to see which of those combinations had the highest click-through rates.
“If it wants to be purple and orange, if that’s going to be appealing to my customer, then so be it,” she said.
Even Mr. Moeck said he was often surprised by the success of certain ads. “Some of it, I just scratch my head and say, ‘I have no idea,’ ” he said.
Tumri’s approach is slightly different. It creates a template for ads, including slots for the message, the color, the image and other elements.
Unlike Adisn, it does not buy ad space, but lets clients — like Sears and Best Buy — choose and buy space on sites themselves. And rather than building a contextual database like Adisn, Tumri uses whatever targeting approach advertisers are already using, whether it is behavioral or contextual or demographic, and assembles an ad on the fly based on that information.
“It’s reporting back to the advertiser and agency saying, ‘Guess what? The soccer mom in Indiana likes background three, which was pink, likes image four, which was the S.U.V., and likes marketing message 12, about room, safety and comfort,” said Calvin Lui, chief of Tumri.
Some advertisers are using that information just to see which version of the ad works best, but Mr. Lui emphasized that the appropriate ad is not static, and changes all the time as content on the page changes.
While the planners and buyers in advertising agencies are intrigued by the idea of measuring each part of an ad, the creative staff that designs ads is less focused on measurement and more focused on the overall effect.
“I think the creative community has to get very comfortable with results-based outcomes in marketing,” said Mr. Hanlon, whose company has an interest in Tumri. “There are a lot of creative people who didn’t sign up for that kind of world.”
Bant Breen, the president of worldwide digital communications at Initiative, the Interpublic Group media buying and planning firm, had a similar view. “The traditional creative process right now is not structured to essentially deliver hundreds of permutations, or hundreds of ideas for messaging,” said Mr. Breen, whose firm is using Tumri to determine which ads are working.
“There’s no doubt that there will be a lot of data that can be collected that could be applied to the creative process.”But, he said, “that’s not necessarily an easy discussion to have with great art directors.”