Exploring Algorithmically Identified Evergreen Stories


Much of digital publishing developed under the influence of traditional publishing. The home page is an manifestation of the front page.

Stories are published to the home page, but are soon pushed off the home page by a never-ending stream of new stories.

This is similar to the consumption habits by readers of newspapers. A newspaper serves its purpose for a day or two before joining the recyclables, making room for the most recent newspaper.

It seems the practical matter of real estate is lowering the lifetime value of many stories. Publishing new stories necessarily means older stories get buried. But some stories have the potential to provide value to readers well past the five-day mark. We call these stories “Evergreens”.

The lifetime of a story is determined by the value it provides to readers. For example, this story about the Giants and Dodgers game provided most of its value to readers during the first few hours. However, this story about Jean-Paul Sartre’s refusal of the Nobel Peace Prize in 1964 is still being read and discussed fifty years later.

At Contextly, we see great potential in stories that are relevant well past their publication date. That is why we developed algorithms to automatically identify and surface evergreen stories.

This moves us in the direction of maximizing the lifetime value of a story.

What Makes a Story Evergreen?

We have reviewed a number of stories that have been identified as evergreen by our algorithms. There are some notable patterns in these stories. They can be described as Seasonal, How-Tos, Reviews, and Factual.

Seasonal evergreens peak in value around the same time every year.

The leaked New York Times Innovation Report cited this story as a specific example of a successful experiment with evergreen stories. The story was about ‘love’ and ran on Valentine’s day, emphasizing the importance of showing the right story at the right time. They marveled, “…even old content can generate significant traffic without ever appearing on the home page.”

Handmade Charlotte is a client of ours. A number of their Halloween-themed stories have recently been identified as evergreen by our algorithms. This one shows kids how to make awesome Lucha Libre masks for Halloween:LucheLibreMasks

How-Tos and Reviews are often identified as evergreens by our algorithms. Good examples include How To Build a Worm Farm by Modern Farmer, A Short Guide to Tequila and Making a Great Margarita by KQED: Bay Area Bites and Adafruit‘s comparison of popular microcontrollers.

Readers’ needs for historical context or background information can result in Factual stories being identified as evergreens. Good examples include Jean-Paul Sartre’s refusal of the Nobel Peace Prize in 1964 and Nelson Mandela’s Obituary.

Our client, CFO.com, has a story from 2008 that describes the difference between corporate dissolution vs corporate liquidation, another example of a factual story having evergreen qualities.

It is interesting to note that the examples of factual stories used here have a Wikipedia-like quality to them; they probably satisfy a similar information need as Wikipedia entries.

Algorithmically identifying and surfacing evergreen stories increases the lifetime value of stories. This benefits readers, writers and publishers.

Readers gain access to more high-quality content at times when it is most relevant to them. Writers are made more productive because the lifetime value of some of their highest-quality stories increase. Publishers benefit because the total value of their stories increase.

Contextly’s mission to help publishers build high-value, loyal audiences drives the development of technology like evergreen story detection algorithms.

If you would like to talk more about evergreen stories and algorithms, I would love to hear from you!: ben@contextly.com


Comments are closed.