Chinese search giant Baidu is using artificial intelligence and deep learning to develop an “omni-marketing” approach that has delivered significantly improved results for advertisers such as Jaguar.

Deep learning technology is enabling Baidu to recognise “very specific patterns” and learn from them, according to chief technology officer Alex Cheng.

“We’re able to highlight ‘dynamic attributes’: we can see people’s search intent, whether they’re in the market to buy something,” he said recently at the dmexco conference in Germany.

“That pattern can be picked up very precisely and very quickly, more so now than ever before.” (For more, read WARC’s report: Baidu’s omni-marketing future.)

Baidu’s ability to combine search intent, buying behaviour patterns and real-time location data means it has “a complete string of information to better serve the user the information they need, at the right time and the right place”.

The next step is ‘intelligent action’, and predicting what users need, as well as offering recommendations and greater personalisation.

“This is what we call omni-marketing: bringing all the information together with the marketing brain behind powered by deep learning and AI technology,” said Cheng. “In the last 2-3 years, we’re beginning to see some real meaningful lift and breakthrough.”

This solves three problems for marketers, he suggested, since it identifies audiences, their intent and is always on.

This approach was used for the launch of the Jaguar E-series in China, which saw customised creative developed for different audiences – ranging from a driver looking for performance at one end of the spectrum to parents for whom a luxury interior or safety was a key feature at the other.

“Based on their intent and their needs we were able to show them the right series of visuals,” Cheng explained. This took place in three phases, from pre-launch, through launch to post-launch.

The results included a 58% lift in launch date searches, indicating increased brand awareness, while a 127% lift in price search indicated increased purchase consideration; there was also an 83% lift in test-drive video search.

“These are all very solid numbers, based on deep learning and better prediction of user behaviour,” Cheng stated.

Sourced from WARC