Behavioral tools drive custom newspaper concept closer to reality

The online “custom newspaper” is an idea that has circulated in the media industry nearly as long as the word “online” has existed.  It was thought that, someday, a newspaper would be able to custom deliver news that had relevance to you and your interests. Not into fashion? No more style section for you.

Until now, the concept has remained separate from reality. But a few companies believe the age of the automatically personalized publications is just around the corner.

HOW IT WORKS

Thanks to improvement in data collection technology, using behavioral targeting to deliver customized content might soon become an industry standard. Much like Amazon is able to review your past purchases and items viewed to give you “recommended items” publishers can use technology to measure their reader’s viewing and buying habits to deliver custom editorial content.

“Everything you do on a website is logged,” said Mike Franklin, professor of computer science at UC Berkeley. “Anything you do generates a bunch of information that provides a bunch of clues about what you are trying to do and what your interests are.”

Last week, Franklin presented at the UC Berkley’s media summit about changes in technology that would be relevant to media companies. Chief among his observations were that websites should track user actions.

“Collecting is easy, the hard part is making sense of it,” Franklin said.

THE TECHNOLOGY

Once the data is compiled, the technology can designate users by interest, demographics and even location. Publishers can categorize articles by topic and then match the two data sets – although such a feat is much harder than it sounds.

“Technology matters,” said Web analytics expert Avinash Kaushik. “It's just that at the moment, the horsepower provided by current tools vastly outpaces our ability to actually use them.”

Engineers have been able to collect large amount of data for years, and nearly every content management system has a way of categorizing articles. The trick is getting multiple systems to work together.

“Newspapers don’t have a single view of their reader,” said Ed Hubbard, director of product marketing at DTI International. DTI, whose clients include The Telegraph in London, has just put the finishing touches on the latest version of? its CMS, ContentPublisher, which incorporates behavioral targeting features.

DTI hopes to correct the problem of several datasets by combing circulation, classifieds, and editorial databases to give newspapers a one-stop shop for information on their readers.

"The more intelligence a company has on their specific audience, the more they'll be able to do new things. It wont just be CPM," said Hubbard.

Franklin, the Berkeley professor, along with Tom Kuhr, is trying to deliver content targeting technology though their company, Truviso. The company makes software that can interpret the large amounts of data collected by publishers and deliver real-time analytics that can be used to deliver customized content.

“But if the organization doesn’t have the maturity to put these rules into place, we’re not a good fit,” Kuhr said.

CATEGORIZING CONTENT

Collecting consumer information is one piece of the behavioral targeting puzzle, but publishers still need a way of categorizing content.

“We have a pretty smart taxonomy engine ‘reading’ the stories,” said Hubbard.

Individual publications can also set up more detailed taxonomies, as a legal publication will need more nuanced tags about law than a general interest newspaper.

Automatically scanning articles and determining the context is nothing new to the web. Google’s has been making a fortune by serving relevant advertising for years.

“Most companies have some sort of classification system, and linking into that system is fairly straightforward,” said Kuhr.

While Google’s algorithm scans a webpage’s HTML to assign keywords, DTI’s system only scans the article’s text and attempts to accurately determine what the story is about.

"It reads every word using not word matching, but word meaning," said Hubbard. For example, a stabbing that took place downtown would have something in common with a stabbing in another area of the city, but not a food article that mentions the word "knife."

Hubbard admits that while he is confident in his system’s ability to automatically tag articles, no computer can yet match the thought process of a human being.

"One of the best journalist’s tools yet to be refined is an 'intelligent reader' that can pour through yours and other people’s content,” he said.

AROUND THE CORNER?

Because there are so many independent computer databases that have to work together, it will take some time before behavioral targeting becomes the industry standard, though progress is being made.

To help streamline the implementation of custom content, Kuhr envisions the equivalent of an ad network for content taxonomy.  That is, multiple publishers collaborating on an industry-standard tagging system to make the implementation of behavioral targeting systems easier.

According to Hubbard, a taxonomy sharing network is already underway among European publications, “so they can build better business models to deliver syndicated material."

Kuhr said that the largest problem for small publishers is that behavioral targeting systems need to be customized to each publication’s needs and many small to mid-size companies do not have the technical staff to build custom software. Truviso said that it can roll out its system in as little as two weeks, though typical installations cost $350,000 and can take months.

But like Truviso and DTI, Kaushik believes the technical barriers will eventually fall. Now it’s a question of publications setting up their workflow and taxonomy to easily integrate behavioral targeting information.

“It is not a tools or technology problem,” said Kaushik. "It is a people and listening problem.”