|
Getting your Trinity Audio player ready...
|
For the last four years, The Wall Street Journal has been building an AI paywall that adapts to reader behaviour and decides how many free (sample) articles they should get access to.
The Wall Street Journal’s paywall houses a machine-learning algorithm that measures reader activity across 60 variables including visit frequency, recency, depth, favoured devices and preferred content types. This forms a propensity score, a unique subscription probability, that then helps inform how many sample stories users can access.
In short, reader activity shapes how much Wall Street Journal content they can sample.
Over the last few years, the WSJ has operated a deliberately leaky paywall that has served as a sandbox of data collection and subscription sale experiments. Reader subscription intent is measured on three levels, as Wells said: “They are cold, warm and hot, like Ronseal, it does what it says on the tin.”
He described its model like this: “We are a dynamic paywall, we can flex based on audience but as far as the consumer sees, we are a freemium paywall.”
The principle is to “sample content to people that we know need it”. By doing so their likelihood of subscribing will rise.
It was once the remit of the newsroom to decide which stories should be opened to non-subscribers. “The decision of whether a piece of content was locked or open was made by the newsroom.”
Wells said this marked a “cultural shift in the business” that has now seen editorial alleviate its control of this section.
