New generation of Web analytics applies ‘big data’ to newsroom decisions
New data tools, perhaps best described as Web analytics on steroids and with psychic powers, are making their way into newsrooms and changing the way that editors decide what stories to promote, where, and when.
It’s part of an emerging technology trend called “big data” -- a process of gathering large, comprehensive, complex datasets and using advanced computer algorithms to visualize them, extract patterns, and use them to make decisions.
These tools, which originated in the labs of the titans of technology and finance, are becoming more mature and affordable, and they're spreading to other industries: health care, science, the military, e-commerce, and now news organizations.
They're showing up in the form of new analytics services that can replace an editor’s hunch with a scientific prediction of what story will perform best.
For years, tools like Google Analytics and Omniture have told site owners which content performed well in the last day, or maybe the last hour. More recently Chartbeat and now Newsbeat accelerate that process to show real-time activity across a website. But even those were just quicker ways to see a record of the past or present.
Replacing the human hunch with an algorithmic crystal ball
The next wave of tools claim to use a crystal ball of website data and patterns to see the future. And they promise to help news publishers squeeze more money out of the content they already produce.
“We created this model where I can take any piece of content created over the last day or two days ... and model how well that’s going to perform in any given position … about 15 minutes into the future.” Mortensen said. “And since we know how well the future is going to play out, we can come up with a set of very specific recommendations about what to put where, for how long.”
How does it work? Essentially, it tracks story views and headline clickthroughs in real time and compares those data to historical benchmarks to identify stories that are overperforming or underperforming. Based on that, it projects the best stories to put in each page position for optimal traffic in the near future.
By putting the right content in the right place at the right time, clients increase page views by 30 percent on average, and Web editors “can spend less time doing the same tasks better,” Mortensen said.
Another new product is called JumpTime Traffic Valuator, founded by people with backgrounds at major media companies such as Yahoo and MTV. It focuses on the revenue potential of each page on a site, showing a publisher how much money each article and each piece of page real estate is generating.
JumpTime analyzes the complete paths (sequences of page views) for each visitor and calculates the total advertising value of each page. This helps a publisher identify articles that are popular editorially, serve valuable ads, and drive the visitor to other pages that also have high-value ads.
So in a simple example, JumpTime would recommend a page with a $10 CPM that is driving traffic to three other $10 CPM pages (a $40 total CPM per visit), rather than a $20 CPM page from which most users exit the site. Without looking at all the pages visited by the user, a publisher might think the $20 CPM page is more valuable to promote.
“As an editor, you know what’s going to be popular, but you can’t know what happens next. That’s where the computer comes in,” said Michele DiLorenzo, co-founder and CEO of JumpTime. “To use a chess metaphor, if I told you to play chess and you could only think one move at a time, you would never win a game.”
The process sounds complex, but in the end the editor sees just one simple number, the total dollar value or the traffic-generating power for each story and home page position. Msnbc.com and ESPN are among the organizations that have tested this system.
Both JumpTime and Visual Revenue display their data by creating an overlay of current pages that shows the performance or recommended content for each position. Other features enable an editor to search and sort stories by topic or category.
The challenges of data-driven editorial decisions
These tools provide powerful insight for the person sitting in the newsroom deciding what to place on a home page or what to promote on social media. But how far should editors go in letting data drive editorial decisions?
Both of these tools are careful to leave the final decisions in human hands. Visual Revenue presents a series of recommended changes that an editor can accept or reject. JumpTime also lets publishers decide whether it’s appropriate for a newsroom editor to look at raw dollar values, or instead show them measures of engagement.
Visual Revenue also makes its recommendations “within a framework of editorial instructions,” Mortensen said. So each news organization can give it detailed guidelines, such as specifying a blend of local, regional and national stories, no more than 20 percent from wire sources or all less than 18 hours old.
Implementing a system like this forces newsrooms to be strategic about what exactly their home page should be.
“If I walk into any newsroom today and ask them, ‘Could you please be so kind and get me that five-page document which describes the strategy for what content we promote on our homepage and section-front pages?' the response I’m very likely to hear is, ‘That doesn’t exist,’” Mortensen said.
Can editors go too far in pushing content based on financial impact instead of journalistic values? Perhaps. But many news organizations today face strong reminders that they have to succeed as a business to succeed at all.
Speaking of money, these predictive analytics systems aren’t cheap. Visual Revenue starts at about $1,000 a month, and the company usually works with properties that have at least 5 to 10 million page views a month. JumpTime’s cost depends on the size of a publisher and the complexity of custom installation. It also is most attractive for big companies, but that includes big chains made up of small newspapers.
Even if these services aren’t right for your newsroom, others will surely come along. The era of big data is arriving for newsrooms, and it could eventually touch more than just analytics.
It seems only a matter of time until someone trains computers to comb through past news content and audience data to suggest the most effective variation on a headline for a story, or even suggest which types of stories to write.
On Thursday, IBM debuted computer chips that actually work like a human brain, instead of just a really fast calculator. Someday the computers might do the writing themselves --wait, they already do.