What if all the data you have on users browsing your site and reading your free content could be analyzed to determine who the most likely to buy is? Schibsted, the large Scandinavian publishing house, has done just that.
"What we didn't really know before this project was how user behavior on the site relates to purchasing subscriptions. So that was unknown territory for us: What they are doing on the site, and what are the characteristics and patterns of those users who end up wanting to buy a subscription."
That quote comes from Eivind Fiskerud, Schibsted's head of data and analytics, speaking in an article on NiemanLab about a subscription-purchase prediction model they've come up with. Picturing it reminds me of something from an old cartoon—maybe Mr. Peabody's wayback machine? Am I dating myself?—where all this information is put in and great numbers spew out.
Schibsted can now analyze a user's actions on a site and the information they provide to know who best to pursue. The model has identified groups of readers 3-5 times more likely than average to buy a subscription. In addition, their telemarketing success rate for selling subscriptions has gone from around 1% to 6% using the model.
For telemarketing, at least, the methods are being used more and more. "Many of our clients use predictive measurements to score their names for best results," Cathy Karabetsos, president/CEO of SIPA member QCSS, wrote to me recently. "They then provide to us in callable format by score. We test, measure and watch ROI."
How does Schibsted's "machine" work? It's not quite rocket science but I'm sure it does take lots of man and woman power. Here are some attributes it uses:
- It's based on an algorithm that has been trained to identify the browsing behaviors of registered site users that go on to subscribe.
- It takes between 10 and 15 variables into account such as a reader's frequency of visits and the number of articles they clicked on, and other behaviors such as the number of devices used to access the site, past subscription history, and the proportion of weekend visits (which correlate with a higher likelihood to subscribe).
- The observational period, after some experimentation, has been set at 14 days. "The data is stored as a shared list of user IDs and individual scores based on a user's calculated propensity to take out a subscription, which the sales and marketing teams can then access and use in their own workflows."
"The importance of having control over our own data, and knowing who the user is, and building a deeper relationship with our subscribers is really key to surviving in journalism," Fiskerud said.
At some point, the technology enters into the equation as well. "The No. 1 predictor of people subscribing is how much they read," wrote Digiday, addressing The Washington Post's big digital subscription spike in 2016. "So once people have sampled, the next step is to get them reading more. To that, the paper has cut page load time 85% and studied reader behavior to determine what articles to offer them next."
The question is how long will it take for Schibsted's prediction model to become available for smaller publishers. In this advanced age, probably not too long.