The Seattle Times Co. last month rolled out a predictive analytics tool to get a better handle on its readers as it attempts to find ways to staunch falling circulation.

The publisher, which manages all advertising, production, marketing and circulation operations for The Seattle Times and its Hearst Co.-owned JOA partner the Seattle Post-Intelligencer, slices and dices its subscribers by news and information consumption as well as other criteria, said Janet Farness, strategic research manager.

“At this first stage we’re looking at various segments of our audience and many of the segments are oriented to the print product only,” she said. “We are not meshing our online registration into this database at this phase.”

Understanding readers

Like every newspaper publisher, The Seattle Times Co. (combined, daily 378,104; Sunday, 457,010) is looking for ways to retain readers.

Being armed with tools to better understand its readers would yield countless benefits.

The publisher realized the need for predictive analytics to decrease costs and improve the targeting ability of its existing marketing and acquisition retention efforts, Farness said.

The tool is currently subscriber-based, although Farness and research manager Nadine Selden see the potential for targeting particular advertising products to particular readers as well. Right now, though, the plan is to drive subscriptions.  

“[We’re interested in] understanding a little bit more about what different segments are interested in with our product offerings and perhaps doing a little bit more in the CRM area of connecting with them on content or events they might be interested in - so both acquisition and retention,” Farness said.

Algorithm predicts

The tool, essentially predictive algorithms, was developed by Chicago-based Apollo Data Technologies, which in addition to The Seattle Times Co., inked a deal with Knight Ridder Inc. and is in the beginning stages of implementing a similar project at The Kansas City (Mo.) Star (see sidebar).

In a nutshell, Apollo provided The Times Co. with algorithm software that works on top of the publisher’s marketing database (Maax from Astech Intermedia).

The Maax database includes data from the subscription database - Equifax data, Claritas data, some classified private-party data and some online subscriber data.

“This database is of the entire market, not just those who are doing business with us,” Farness said.

“We built them a predictive model that identified who their subscribers and non-subscribers are and what segments they fall into,” explained Jeff Kaplan, co-founder and principal of Apollo. “Based on their subscriber database, we had a survey they did that asked geodemographic and attitudinal questions - questions like, ‘How important is being well informed in the news to you?’ or ‘Do you prefer to buy the newspaper only on days you want it?’”

(Editor's note: Minnesota Opinion and Research Inc. conducted The Seattle Times Co.'s original survey of newspaper and online audiences upon which the predictive analytics algorithms from Apollo Data Technologies were based.)

Kaplan said that by using that information the publisher can predict what types of readership segments people fall into.

“Apollo created two algorithms that can predict the larger marketing database of 2.1 million individuals and what segment they would be in based on that (Market & Opinion Research) survey,” The Seattle Times Co.’s Selden said.

Kaplan said the main reason firms like The Seattle Times Co. are turning to organizations like Apollo is that their margins are getting crushed because of free online advertising.

“They have to think of ways to increase customer retention and increase subscriber base,” he added.

The Seattle Times Co. plans to do plenty of tests to monitor how the tool is working.

“We’re kind of setting up various Petri dishes full of experiments and then we’ll be measuring and correcting,” Farness said. “I think incremental improvement in your market always hits your bottom line pretty nicely - so if we can improve acquisition efforts by a half of a percent or a percent, or improve retention by one or two percent, that’s very significant.”