Personalizing News Recommendations by Building Better Reader Profiles
Only a few years ago, many people used RSS feed readers and personalized news aggregators to quickly find the stories they wanted to read. The promise of many of these apps and readers was that they would use data to learn what you liked and provide you with the most interesting stories.
The technology seems almost outdated now, since Facebook’s newsfeed became the number one driver of readers to news websites, but can digital media sites still learn something from news aggregators and personalization apps? If news websites hope to keep gaining and growing loyal readers, providing those users easy ways to find what they’re looking for is step one in a good online experience.
At last month’s INMA conference, Schibsted Media Group’s Vice President Products for Data & Identity Edoardo Jacucci talked about his organization’s ambitious data strategy, which includes creating better personalized recommendations by building user profiles.
“Stereotypes Work in Your Favor”
“Stereotypes are horrible, but they work,” said Jacucci of applying common assumptions to improve recommendations to groups of readers.
Schibsted Media’s network of classified and media sites, allows it to use data to make predictions about readers’ age, sex, and other demographic data. The goal, said Jacucci, is to build profiles that allow Schibsted Media to feed personalization, and eventually monetization.
These “stereotypes” allow the organization to better react to an individual’s behavior. For example, the actions taken by readers allow Schibsted Media to reasonably predict a user’s age range based on how they are consuming content. Browsing the classifieds for a part-time job? It’s likely you are a college student. Reading about challenges for first-time home buyers? You’re probably in your early thirties.
But before it can monetize, Schibsted Media has to wrangle all of its data with its 20+ person team.
The retail industry has been using similar data to make predictions based on user behavior. Its level of sophistication allowed Target to identify a pregnant shopper before the woman was able to tell her father.
Building the Recommendation Engine for the Future of News
Building user profiles and combining them with reader data can lead the way for newsrooms to fulfill the promise of personalized news apps, but as scores of failed news recommendation engines can attest, the engineering challenge isn’t easy.
How can your newsroom take advantage of personalized news and services?
Assess What Data You Can Access
Schibsted Media used its network of sites to build comprehensive user profiles. This is also the advantage many ad networks offer. If you cannot compete on scale, what data can you use? For example, Unique User IDs (UUIDs) can be used to build basic profiles of articles users read and can serve as the basis for showing similar stories or articles.
Can you access your full suite of analytics data? Without access to its data, Schibsted would not be able to attempt building user profiles or provide recommended content.
Understand What You Can Do In-House vs. Out-of-House
Even if you do have access to your data, not all companies have a 20-person data science team to build personalized user profiles. Are there off-the-shelf solutions that still ensure readers are treated to personalized content based on their reading or behavioral habits? Content recommendation engines like Taboola or Outbrain offer this; Parse.ly clients can also use the Parse.ly API to power personalized recommendations.
Track The Success of the Solution You Choose
The most innovative use of data to build profiles still isn’t worth much if you can’t show how it’s changed user behavior. Jacucci noted at the conference that the ultimate goals of the data project would be monetization and user growth. Make sure your team knows what the goals are for the systems you’re building, and has a way to measure the results once implemented.
The Promise of News Aggregators 2.0
There is no single way to personalize users’ site experiences, but it is becoming increasingly clear that personalization through smart usage of data can lead to more options for monetization. This time, perhaps the promise of personalized news will be realized — not by a technology company, but by news outlets, themselves, which have worked so hard to produce and report the news.