Propensity modelling
Propensity modelling uses data on a reader's behaviour — how often they visit, what they read, whether they've hit the wall — to estimate how likely they are to subscribe.
Publishers use it to power dynamic paywalls and target offers.
Why it matters for publishers
Not every reader is equally close to subscribing. Scoring propensity lets you spend your best offers and most aggressive prompts on the readers most likely to convert, and leave everyone else free to keep reading — which protects reach while lifting conversion.
How publishers use it in practice
- The strongest signals are behavioural: sessions in the last 30 days, articles read, recency of last visit and whether the reader is registered.
- You do not need machine learning to start — a simple engagement score bucketed into low/medium/high captures much of the value.
- Validate against actual conversions monthly and retrain; reader behaviour shifts with the news cycle.
- Use scores for more than the wall: prioritise sales outreach, retention emails and win-back offers too.
Frequently asked questions about propensity modelling
- Is propensity modelling GDPR-compliant?
- It can be. Using your own first-party behavioural data with a lawful basis and transparent privacy notice is standard practice; buying third-party profiles is where risk arises.
- How much traffic do I need?
- Enough conversions to learn from — as a rule of thumb, several hundred subscriptions a month before a trained model beats simple engagement buckets.