Non-subscribers visiting WSJ.com now get a score, based on dozens of signals, that indicates how likely they’ll be to subscribe. The paywall tightens or loosens accordingly: “The content you see is the output of the paywall, rather than an input.”
The Wall Street Journal thinks it might know your reading habits — and your potential spending habits — better than you know them yourself.
For the past couple of years, the Journal — home to one of journalism’s oldest paywalls — has been testing different ways to allow non-subscribers to sample its stories — refining a subscription prediction model that allows it to show different visitors, who have different likelihoods of subscribing, different levels of access to its site.
Non-subscribed visitors to WSJ.com now each receive a propensity score based on more than 60 signals, such as whether the reader is visiting for the first time, the operating system they’re using, the device they’re reading on, what they chose to click on, and their location (plus a whole host of other demographic info it infers from that location). Using machine learning to inform a more flexible paywall takes away guesswork around how many stories, or what kinds of stories, to let readers read for free, and whether readers will respond to hitting paywall by paying for access or simply leaving. (The Journal didn’t share additional details about the score, such as the exact range of numbers it could be. I asked what my personal score was; no luck there, since the scores are anonymized.)
“I think back to maybe eight months ago, when we were looking at all these charts with a lot of different data points. Now we’ve got a model that’s learned to a point where, if I get a person’s score, I pretty much know how likely they will be to subscribe,” Karl Wells, the Journal’s general manager for membership, told me when we spoke last week, with a Journal spokesperson on the call. “What we’ve found is that if we open up the paywall — we call it sampling — to those who have a low propensity to subscribe, then their likelihood to subscribe goes up.” (The Journal’s model looks at a window of two to three weeks.)
The Journal has found that these non-subscribed visitors fall into groups that can be roughly defined as hot, warm, or cold, according to Wells. Those with high scores above a certain threshold — indicating a high likelihood of subscribing — will hit a hard paywall. Those who score lower might get to browse stories for free in one session — and then hit the paywall. Or they may be offered guest passes to the site, in various time increments, in exchange for providing an email address (thus giving the Journal more signals to analyze). The passes are also offered based on a visitor’s score, aimed at people whose scores indicate they could be nudged into subscribing if tantalized with just a little bit more Journal content.
The cost of a subscription doesn’t vary based on a reader’s score. As of the publication of this story, the Journal has been offering $222 for a full year of digital all-access; there’s a student deal at $49 for a year.
“These passes play a role in making our subscription model more predictive,” Wells said, since now the Journal can collect additional data on the person who’s been hopping around the site for a while. “They also still put a value on our content, telling you we’re still a paid-for site, and that you’re being welcomed in as a guest to enjoy 24 hours of access.” A person willing to hand over an email is also more willing to eventually pay for a subscription. Other targeting possibilities open up as more people hand over their emails: the Journal could email a subscriber with a slightly higher propensity score a little more often, or recommend specific newsletters on topics they’ve already shown an interested in. (Passes are the least common of the site experiences a non-subscribed reader might see. They’ve been undergoing some technical upgrades, so if you’re hunting around for one right now you might not see one.)
In addition to guest passes, WSJ.com has publicly tested other ways for non-subscribers to try out its stories. In a feature introduced in August 2016, a Journal reporter who shared a link to a Journal story on social media could unlock that story for readers. The sharing option also used to extend to subscribers who shared a Journal story on social media, though that channel is no more as the Journal moves completely towards its reader score-driven paywall system.
Wells and the spokesperson didn’t share any specific numbers around conversion rates or how many guest passes had been offered, pointing broadly to the organization’s latest subscriber numbers. The Journal now has 1,389,000 digital subscribers, according to News Corp’s latest earnings report, up from 1.08 million a year ago.
The Journal is hardly the first to use propensity modeling techniques — common in the app world for trying to convert users into paying users — to increase subscriptions. The Financial Times has for years been using reader data to more efficiently target readers with the offers that they are more likely to respond to (Ken Doctor wrote about these efforts for us as far back as 2010, here, here, and here.)
Scandinavian media giant Schibsted developed a prediction model that identifies, based on many signals, readers who are 3× to 5× more likely than average to buy a subscription, and then advertises offers to them differently. (Wells actually presented the Journal’s developments on this front at a paid-content summit hosted this month by Axel Springer in Berlin, at which news organizations like the Journal and Schibsted brands discussed how to improve digital subscription strategies.)
Wall metaphors were once sufficient for news organizations’ paywall strategies. A paywall could be deliberately “leaky” or “porous.” The gates could be fully opened for public emergencies like severe weather events or terrorist attacks. It could be a hard wall you could peek over, but with no gaps in between. That metaphor doesn’t quite work for the Journal anymore.
“If you think about paywalls broadly, there have been metered, freemium, and hard paywalls. Metered considers people who will want to read more than, say, five stories. Freemium assumes, this and not that is the type of content people will pay for. This is what we’ve tried to move on from,” Wells said when I tried to ask about whether certain types of articles were always more likely to get a reader to pay, and what other specific holes there were for WSJ.com. “Our model now is to flip that and start with the reader. The content you see is the output of the paywall, rather than an input.”