Dynamic paywall:
Every reader gets their gate
A fixed meter treats the first-time visitor like the loyal reader on the verge of subscribing. A propensity model decides per reader and article when the paywall kicks in - and pulls far more subscriptions from the same contacts.
01 The problem - The fixed meter gives away value in both directions
Too early it scares off reach, too late it gives away sales
Every rigid rule - three free articles, then the gate - is wrong for most readers: the fleeting visitor is scared off before any bond forms; the highly engaged reader keeps reading free for months although they would have paid long ago. Both cost subscriptions.
Willingness to subscribe is in your data: visit frequency, section mix, reading depth, device, newsletter status. NZZ built a model with 100 to 150 data points from these signals and increased its conversion rate fivefold within three years. The principle transfers to any media house.
Not more traffic, not more ad pressure: just the right gate for the right reader.
02 The model - Subscription probability per reader and article
Gradient boosting on session history and article features, served in real time at the paywall
{'kalt': 0.71, 'warm': 0.24, 'heiss': 0.05}
Erwarteter Uplift vs. fixe Meterung (Backtest): +31%In practice the score yields three zones: Cold - the paywall stays open, build a bond first (newsletter, registration). Warm - a measured gate with a trial offer. Hot - a hard gate with a full-price offer, because these readers convert anyway. Piano measures a factor of 174 in subscription probability between the highest and the lowest propensity segment - equal treatment is the most expensive option here.
03 Business impact - More sales from the same traffic
The lever: a 30 percent conversion uplift on today's paywall conversion rate
No round number: every assumption comes from the sample publisher and is stored centrally. With your real figures only the input changes, not the method.
| Item | Value |
|---|---|
| Paywall contacts / month | 150,000 |
| Conversion rate today (0.6%) → new subscriptions / month | 900 |
| Baseline: Additional subscriptions / year at +30% conversion | 3,240 |
| Lever: Additional subscriptions × 7 paid months in year 1 × ARPU | 3,240 × 7 × €10.50 |
| Result: added revenue / year | €238,140 |
Assumptions of a sample publisher - in a real project your data replaces these values.
An INMA benchmark with propensity steering documents +77 to +166 percent higher conversion, Schibsted +75 percent more subscription sales from ML-selected front-page articles. The model calculation deliberately starts at the lower edge with 30 percent.
04 Next steps in your publishing house
From the score to a learning paywall rulebook
We reconstruct from your web analytics and paywall data who stands at the gate today - and who bounces there.
The model first runs in parallel with the existing meter. You see on real sessions where it would have decided differently - and better.
The dynamic gate starts on a share of traffic and proves its uplift in a controlled test before it scales.