The rise in reader revenue in the news industry has been a welcome panacea for our woes - not least the declining print market, our collective reliance on high-traffic advertising and changing reader habits. Over the past twenty years, we have seen a variety of paywall models, including the somewhat elusive “dynamic” paywall which has seemed to date to be the preserve of only the most sophisticated and technically advanced of news organisations.
But as technology improves, and publishers become more proficient in data handling, dynamic paywalls are becoming more commonplace.
The Financial Times launched a dynamic paywall about a year ago, using Zuora as a tech partner. The AI paywall has become a key driver of subscription growth, with a 92% increase in conversion rate, a 118% increase in progression through the subscription funnel and a 78% uplift in subscriber lifetime value. Besides these excellent numbers, the new paywall has also reduced manual overhead, freed up engineering capacity, and empowered the FT to pursue bold new growth ambitions with confidence.
(If you would like to learn more about the FT’s paywall, please register here for a webinar with FT experts and Zuora to discuss what it takes to build and launch a dynamic wall. And if you can’t make it live, sign up anyway and we’ll send you the recording to watch in your own time.)
During the INMA Subscriptions Summit this year, it was highlighted that 22% of 264 news brands declared using a hybrid / dynamic / smart paywall, the proportion of which has increased 4 times since 2020.
In our work with publishers around the world, FT Strategies has seen how paywalls tailored to content preferences, user behaviour, and demographic data, can more effectively convert engaged readers while retaining those with lower willingness to pay through ad revenues, compared to a one-size-fits-all paywall.
As INMA highlights in their article on dynamic paywalls: “For years, dynamic paywalls were a bit like driverless cars – everybody talked about them, but only few had seen them.” Similarly, everybody seems to have different ideas of what dynamic paywalls actually do. So, this blog series will walk you through what to consider and how to build your dynamic paywalls. First of all, let’s start with some simple questions facing publishers looking to introduce or optimise a dynamic paywall.
Understanding what we mean by dynamic
A dynamic paywall is a system that adjusts both when to show a paywall and what to show on it. Dynamic paywalls exist on a spectrum: at one end are rules-based systems that rely on pre-set rules and reader segmentation, while at the other are what Jonathan Harris from Zuora calls "self-learning" AI paywalls that leverage multi-agent reinforcement learning.
The key difference between a classic paywall and a dynamic one is that the former is triggered based on a fixed rule applied to all readers (e.g. a standard paywall appears after three articles within a 30-day period), while the latter is triggered at different times or with different subscription offerings based on user behaviour and characteristics.
As Harris explains: "The key difference with true AI paywalls is that they learn and adapt automatically. Rules-based systems, even those that use propensity scoring, require constant manual adjustments – you're always chasing audience behaviour. With reinforcement learning, the system explores what works and doubles down on successful patterns, optimising for long-term subscriber value rather than just immediate conversions."
What elements become dynamic can vary by case, depending on data availability, user needs, and historical trends. Some of the elements that make a paywall dynamic include the following:
|
Access model |
Call-to-action or paywall messaging |
Packages and prices shown to readers |
|
Adjust when to show the paywall and what content should be gated |
Adjust marketing messages to convince readers to subscribe |
What subscription packages to promote and how prices are shown to readers |
|
Example: Readers who exhibit behaviours similar to those with a higher likeliness to convert may encounter the paywall after fewer free articles. Meanwhile, those with a seemingly lower willingness to pay may be allowed to read more, sometimes through free registration, before hitting the paywall. Example: Lock political content for readers who frequently consume politics, and business content for those who engage more with business coverage. |
Example: A reader who prefers business content might see the message, “Grow your business with best-in-class market news” or “Must-read for business leaders.” In contrast, a political news reader might be shown, “Stay informed with your local politics” or “Know what’s happening behind the headlines. Subscribe for deep political insight.” |
Example: Readers who resemble high-engagement subscribers might see a premium digital package, while those who engage lightly may be offered a lower-priced subscription with a great emphasis on the level of discounts they can enjoy. |
Note: Publishers sometimes conflate dynamic paywalls with dynamic pricing, but we have intentionally excluded dynamic pricing from this discussion, as it requires a completely different set of steps to implement. However, if you are interested in exploring dynamic pricing, this post is a good place to start.
Rules-based approaches are often a practical starting point, particularly when data is limited or internal capabilities are nascent. More advanced systems using AI, enable real-time personalisation and optimisation at large-scale.
To dynamically adjust the elements mentioned above, publishers often build or refine a machine learning model, typically a classification model, to predict how likely a user is to subscribe based on available behavioural and contextual data. This model then triggers tailored responses (e.g. locking content, adjusting call-to-actions) in real time or near real time. Clean and structured data is essential for ensuring the model generates accurate and effective predictions. The specific dataset used to train the model will depend on data availability, but a typical minimum set of data includes:
- Real-time user behavioural data: eg., time on site, content consumed, number of articles read, traffic sources, and device used
- Engagement data: eg., registration, newsletter sign-up, and usage of product features
- Historical subscription data: eg. acquisition and churn timings
In the process of training a model and creating pre-set rules for a dynamic paywall, other types of data, especially first-party data such as age, location, and work role, can offer significant advantages to news publishers. As many in the industry recognise, first-party data holds enormous potential, and with the right strategies, publishers can unlock substantial value from it. The Wall Street Journal, a pioneer in dynamic paywalls, uses more than 60 user signals, while the FT’s latest experiment orchestrates across about 250 user states to inform its AI-powered paywall, highlighting the scale and diversity of data required for effective personalisation.
A clear understanding of dynamic paywalls helps you properly scope your project while clean datasets are a key enabler of long-term success. With this foundational understanding in place, the next blog post will walk you through how to build and optimise a dynamic paywall with a step-by-step guide.
At FT Strategies, we offer a range of support services that news publishers can leverage to optimise their paywall strategies. These include building data infrastructure suited to dynamic paywalls, advising on Build or Buy decisions, supporting vendor selection, developing predictive financial models and co-creating / implementing your paywall strategy. To find out more about how we can support you, please get in touch with our expert team today.
About the author
