The Essential Public Radio Frontpage

The Problem

I was the web producer for WESA 90.5 FM, the local NPR affiliate in Pittsburgh, PA, and as much as I loved the job, I couldn’t help but notice how much of my time was taken up shuffling stories to different spots on the homepage. I knew that no matter how dedicated or responsive I tried to be, I would never be able to beat an automated system built into the site itself. But could I design a system that could automate editorial decisions like which story belongs in the top spot, which in second, and so on?

My Approach

In the original system, the news editor decided where stories went, he told me, and I made the updates. So I started paying close attention to his decisions to see if there was an underlying pattern.

It turned out there was a distinct flow that stories normally followed in their life on the frontpage. Top stories moved to the second spot when a new top story came along; the story in the secondary spot moved to the tertiary, and so on. Of course, not every story merited top billing from the start, so sometimes they might start in the secondary or tertiary spot. But wherever the story started, it almost always followed the same path. Sometimes more than one story would hit the frontpage at the same time, but even this didn’t really change the overall pattern, but provided multiple simultaneous inputs.

Our Solution

I created a module that would allow the news editor to assign a newsworthiness score from 0 to 10 to any story. I developed a mathematical formula that would derive a present newsworthiness score by taking this number and causing it to decay to zero over a set period of time. Views on the story would slow that decay, so that popularity would keep a story in a prominent position for longer. This would allow a second-spot story that performed well to potentially stay in the second-slot for quite some time.

We ran some tests with this formula, and it seemed to work out well. As a safeguard, we added options for the news editor to “bump” the newsworthiness score up or down.

With this in place, the news editor was able to manage the layout of the front page directly by focusing on the importance of each story from a journalistic point of view. The formula was designed to reflect the station’s journalistic commitments and priorities, but it became a primary means by which those commitments and priorities were implemented. The layout was automated, meaning even when I was asleep or on vacation, the front page would highlight the most important, recent, and popular stories up to the moment, and could even weigh those factors more honestly and consistently than we might have as human beings.

What We Learned

Design is about solving problems. Often we think of those solutions in terms of interfaces, but sometimes the most important design decisions are hidden in algorithms or mathematical formulae. In this case, I was able to recognize the algorithm hiding behind editorial decisions and derive a single mathematical formula that reflected a news station’s journalistic standards and priorities. The key to the success of this approach was that I recognized the subjective, human core of it in the judgment of a qualified news editor. The editor would tell me how important a news story was, and my algorithm would focus on the questions that follow: How important was it two days later? How important was it two days later if it remained the most popular story on our site? How important was it two days later if no one was reading it? More broadly, this case study reinforced my commitment to stay on the lookout for all the little design problems we assume are unsolvable.