What makes a loyalty platform "predictive"
Most loyalty platforms are reactive — they tell you what happened last month. A predictive loyalty platform forecasts what will happen next month and acts on it before it happens. Fundle ships three predictive scores on every member, refreshed daily: 30-day churn risk, predicted 12-month LTV, and next-best-action propensity. Each score is fed into automated agent decisions — campaigns, tier upgrades, reward selection — without human intervention.
The three predictive models inside Fundle
- Churn model — gradient-boosted classifier on 200+ behavioural features, 30-day horizon
- LTV model — survival + regression hybrid, with category-level lifetime projections
- Next-best-action model — contextual bandit over offer × channel × send-time space
How this translates to revenue
- Win-back triggered at day 30 of declining behaviour, not day 90 of inactivity
- 20-40% LTV uplift on identified vs. anonymous shoppers
- 3x higher campaign ROI vs. blanket sends (matched-control measured)
- 35% reduction in 12-month churn rate in mature deployments
- +22% AOV on AI-personalised offers vs. category baseline
Where this fits in the broader AI loyalty stack
Predictive scoring is one layer. The platform also has to take action: send the message, price the reward, suppress the conflicting campaign, measure the outcome. Fundle ships predictive + agentic + measurement on one platform — the predictive scores are inputs to live agents, not just dashboards.
Related resources
Looking for more? Open the Industries menu to browse playbooks by sector, brand or mall.