The paywall models of metered, freemium, and hard all have one thing in common: Their access rules are determined by content, rather than customer data. This has led to a one-size-fits-all approach.

The Wall Street Journal, however, has spent the last year moving away from that model. Jon Buckley, director of digital subscriber acquisition and retention, shared this journey in an INMA Webinar presented on Wednesday.

Buckley introduced The Wall Street Journal’s strategy by sharing the main goals of its digital subscription strategy: to maintain traffic and grow the subscription base. After its initial foray with a freemium model, the WSJ has since innovated from a content-led to a customer-led approach.

The first iteration of this was implementing the capability to open or close the paywall, depending on the visitor. The second iteration, which the company has been working on the past 12 months, was to create a personalised model that can qualify between subscriber and non-subscriber when they hit the paywall.

Most current paywall models have one thing in common, Buckley said: their access rules are led by content, not the consumer. The experiences are the same. “Our vision was very much reader-first,” he said. “We wanted to put in place something that was personal and dynamic.”

WSJ started off by first understanding when people buy, using a predictive model that analysed more than a year’s worth of subscription data, drilled down not just to the day, but to the hour. “We were able to identify the key points in the day, month, and year when we needed to lock down our site to drive higher subscription volume,” Buckley said.

The second item put into the model after predictive data was advertising demand, which resulted in a real-time demand model.

“That was 12 months ago; we set out to create a dynamic paywall and we did that,” Buckley said. “Then we knew we needed to add an additional layer, with intelligent personalisation that’s powered by knowing more about our potential customers. We use 65 key signals that influence the propensity to buy, from education and location, to articles accessed and time on the site. Over time, this model became more intelligent.”

Every non-subscriber who visits the WSJ Web site is given a “likelihood to subscribe” score. This score is then piped into the onsite DMP, to determine the user experience — which is determined based on their likelihood to subscribe.

Buckley explained, “When we recognise someone with a low propensity to buy from us, instead of pushing the hard sell on them, we show them a subscription message to let them know we are a paid model, but we offer them a complimentary 24-hour guest pass.”

From an onsite perspective, WSJ was able to leverage the paywall, which also allowed the company to focus more on the media buying opportunities side.

“Based on that, we’ve seen terrific results,” Buckley said. “We’ve seen a five-times increase on our on-site conversion rates. We’re now over 1.3 million subscribers.”

The six fundamental steps of their process were:

• Understand demand.
• Understand the signals.
• Create the model (accept that will take time).
• Operationalise the scores.
• Deploy the relevant experiences.
• Test, test, and test again.

When this process started two years ago, the team was a core group who spanned different divisions: editorial, advertising, and leadership. “It used to be that editorial decided what articles were paid; now, it’s the data that tells us the reader’s likelihood to subscribe that determines the paywall,” Buckley said. “It’s really been a collaborative effort.”

Q&A

INMA: How long did it take to implement this new model?

Buckley: It was two years in total, though much of the first year was spent looking at the other models out there and what approach we should take. Then we started honing in on the subscriber data. And then once we had that understanding, we needed to quickly pull in the ads to determine what a dynamic paywall would look like (that part took the better part of a year).

INMA: How does the WSJ digital edition come into it?

Buckley: It doesn’t; this model is just the WSJ main Web site and content.

INMA: When you are looking at how you optimise for conversion, does that happen in real-time?

Buckley: Yes. For example if a new user comes to the site today and reads a few articles, and then they come back tomorrow — at that point, the system recognises that they’ve been to the site before, and will begin adapting the content and experience to them in real time. When we’re looking at someone who has a high propensity to buy, we ask what they look like compared to the subscribers we already have. We give them an experience based on the learning we’ve had from previous customers.

INMA: Where are you getting those 65 variables?

Buckley: It’s a fusion of first-party and third-party data. Some of it comes from previous readers who have come through and either have or haven’t signed up, while other variables are layered in from third-party data.

INMA: Did some of these variables rise to the top as far as being most important?

Buckley: The most important one for us was frequency, how many times the visitor had been to the site. For us, five is the magic number. If we can get you to the site five times in a 30-day period, you have a 120% chance to subscribe. We also look at what sections and sub-sections that people are signing up on, which are generally our core areas of content.

INMA: Were there other key metrics you evaluated?

Buckley: The two main goals were to maintain traffic and grow the subscription base. We’re now building out the churn propensity to understand why people stay or why people leave. The two key metrics have definitely been traffic and to drive new subscriptions.

INMA: Where do you go from here? Are there specific hurdles you’re looking to tackle?

Buckley: It’s still in its infancy, really. We need to be serving up different messaging to users, determining new experiences and the right experience for the right user — not just the initial access point, but going all the way through the process.

INMA: Have you had any significant cases of consumers having a work-around or cheat the model?

Buckley: When we had the freemium model, we had about a million paywall avoiders in any given month. Going from that to the model we have now, with far more variables involved, we haven’t had nearly that problem.

INMA: What is the background of your team directly involved?

Buckley: Data scientists, for the most part. They built the propensity model and how that operates. The product and engineering team then built out the solution.

INMA: What about newsletters? How do they fit into this strategy?

Buckley: Yes, from both a willingness to subscribe and a stick rate for anyone who’s signed up for a newsletter, makes their propensity very high. So that is key.

INMA: Which peer publishers do you generally benchmark best practices against?

Buckley: From a data standpoint, Financial Times. Outside of the publishing sector, it’s those like Netflix and Spotify.

INMA: Has a similar level of sophistication been applied to retention?

Buckley: Yes, that’s being built out now. We’ve got a good understanding of acquisition strategy, and now we’re building the scoring for propensity to churn. The learnings that we’ve gotten on the acquisitions side, we’re taking a lot of them onto the members side, to see what customers look like who are willing to stay or willing to go.

INMA: When you see different users on different devices, do you see them having different purchases?

Buckley: Similar to other publishers, we see a lot of browsing on mobile and higher conversion coming on desktop. Once they’re subscribed, we heavily push them to read us on more devices, as this means they are more likely to remain with us.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .