“Having a machine-learning framework to say who’s likely to churn, register, and subscribe has been a critical step in us making those experiences more tailored.”
Paywall company Piano introduced its LTx propensity wall to lure users into paying, similar to the ones used by The Wall Street Journal, Financial Times and other media outlets to probe the “likelihood to action”.
Paywalls, dynamic walls, intelligent walls, and propensity walls all use several signals to determine each visitor’s propensity to subscribe and analyze what action has to be taken to improve the scores. Actions can include, for example, periodically sending extra free articles, newsletter subscription forms sent before leaving the page, and targeted social media ads with offers.
It takes Piano two or three weeks to set up an LTx paywall tailored to the signals and data acquired, which is exactly the timeframe needed to gain a new subscriber – according to Trevor Kaufman, Piano’s CEO, to Neiman Lab.
“People’s focus on a site tends to be very intense at given periods of time…90 days from now, your loyal audience will largely be a different group of individuals with maybe 30 to 40 percent overlap,” says Kaufman. “We wanted to make our system more adaptable to accommodate that. Having a machine learning framework to say who’s likely to churn, register, and subscribe has been a critical step in us making those experiences more tailored.”
Companies nowadays tend to aim at involving customers beyond simple subscriptions, and propensity is about making the most of a client’s lifetime value. According to Piano, metrics are not that reliable in determining a customer’s long-term propensity.
“Even with subscription websites, the page metrics [tend to] have driven short-term value, as opposed to long-term value. That starts to transform the way you think about operating a media business, from pageview to customer lifetime value,” explains Michael Silberman, Piano’s SVP of strategy.
However, access to metrics appears to have helped Piano’s clients in generating traffic and conversions. Since the paywall’s launch in June, a client had a 20 percent increase in its paid conversion rate and another one saw a 75 percent increase in visitor’s subscriptions.
The propensity wall is in place for five clients over eight sites. Piano serves 1,300 publishers like Business Insider, TechCrunch, and The Economist, though they have to opt into the paywall setup. The company has not uncovered the names of the clients using it.
The machine-learning algorithm that powers the LTx propensity wall was built using the random forest technique. The developers beta-tested the site for two months refining its reaction to real-time prediction data and what actions are appropriate in various situations.
“Do you show the subscription offer plus the offer to register for temporary access? Do you show them just the offer and not register? Another use case might be that different meter height for users depending on their subscription propensity score,” Silberman said.
Each client of Piano’s propensity wall can fine-tune its propensity factors if needed: “For a local newspaper website, one of the things we’ve discovered — no surprise — designated market area [DMA] is important. For at least one of them it’s not where the user is in terms of DMA but if the content is from that local DMA,” he explained. “Another client knew going in that a lot of their users were converting on the content of one particular author. We added ‘count of articles read by x-author’ as a metric as opposed to generic ‘number of authors read’ in the algorithm.”
When the propensity wall is live, Piano divides users into sets and assesses the existing propensity distribution. The LTx is then able to kick in, giving visitors more options tailored on the 76 metrics acquired on them, and generate a larger number of subscriptions and pay ups.
Personalizing online user experience is becoming a common practice among many companies. Media outlets, retail stores, airlines – all hope to increase revenues and stand above the competition, and a very good way to do this is to “get closer” to the customer, by speaking to him one on one.
However, this is not done by employing more workforce dedicated to customer support, as one might logically believe. Companies increase profit by boosting sales just as much as by saving money, and automation is a great way to achieve this.
This is when complex, data-hungry algorithms step-in. And this is why paywalls are appearing all over the web – gathering data, profiling users, and using this edge to better manipulate people’s habits as consumers.
“Commercial experiences in general are becoming more 1:1. It’s not just the messaging, but the pricing, the purchase experience you have, the products you’re offered — should all be relevant to you,” said Kaufman.