Amid the wealth of information streaming out of websites, banner ads get little of users' attention. For website operators that depend on advertising revenues and online retailers, that's not good news. So could they make those banners more appealing by tailoring the content more closely to users' personality type? Glen Urban, an internet marketing researcher at the MIT Sloan School of Management in Cambridge, Massachusetts, thinks so, and has developed an "ad morphing" system along with colleague John Hauser to do just that.

Urban says people who are looking to buy things online tend to behave in distinctive "cognitive styles" when presented with information: deliberative or impulsive, holistic or analytical, and visual or verbal. By changing the appearance of banner ads to fit in with personality type, it is possible to make the ads more appealing to the user, he says.

Offline, good salesmen know what customers' cognitive style is, and whether they really want to buy, by observing their body language and by asking pertinent questions. On the web, no such face-to-face diagnosis is possible, and web shoppers don't usually respond to questionnaires, says Urban. This leaves online merchants with few ways of working out what sort of customer they are dealing with.

Now websites can be primed to "read" potential customers from the way they interact with web pages. It uses a program called the Bayesian Inference Engine running unobtrusively on a user's computer to monitor the person's click patterns and so to determine how they respond to different textual and visual cues. This is then used to categorise the user's cognitive style.

Buy? Ignore?

This information is stored in the user's web browser as a chunk of data – a cookie – that website operators can use to serve up banner ads that fit the user's personality type. Another algorithm monitors what happens when such a banner ad appears – "click through", "buy" or "ignore" – and this is fed back to the inference engine, so that its predictions get better over time.

In a controlled panel study for a large automotive manufacturer, the percentage of participants who clicked on a banner ad went up by 57 per cent when "morphed" ads were served up instead of one-size-fits-all ads.

The system will enhance existing behavioural targeting techniques used with banner ads, says Urban. These techniques exploit cookies that contain browsing history and search data.

Because trust is key in commerce, consumers should know when they receive morphed ads, he adds. If online shoppers opt out if the system, the morphing cookie that has them pegged will be deleted.

Personal experiences

The web promises to customise itself for the user, says George Pappachen, chief privacy officer of WPP-owned market research firm Kantar Group, and ad morphing is in keeping with this promise. Companies try to cater to online shoppers by noting their personal preferences, but shoppers sometimes find that creepy: transparency can offset such problems, he points out. "Consumers should have control in deciding what content is relevant to them and how it is presented to them," he says.

But even when consumers are happy to have targeted adverts, serving personalised adverts is tricky, he warns. An appeal to the impulsive gadget buyer could backfire if they themselves as a more deliberate shopper. "For customisation to succeed, the result has to align with the consumers' opinions of themselves."

Live trials of the system will begin in January, when a large telecoms company, which Urban prefers not to name, will deploy the technology.