Completions

For subscription retention leadership

Save flow that picks the right offer for each subscriber

Not the same 10% discount for everyone. Propensity-scored offer selection, waterfall logic, per-location overrides, and a compliance gate — so save-rate climbs from 8-12% to 18-30% without burning unit economics.

By Jay Christopher10 min read

What this gets you

  • Propensity-scored offer selection — discount, pause, swap, gift, or downgrade chosen per subscriber based on churn signals + historical save-rate data.
  • Save-flow waterfall — highest-propensity offer first, fall through to next on rejection. The subscriber never sees the cheapest offer first by accident.
  • Per-location save-offer overrides — corporate retention policy plus location-specific augmentation. Franchise units stay inside corporate guardrails while testing their own variants.
  • A/B testable variants — measure save-rate per offer per cancellation reason per location. The next best offer surfaces from data, not from a meeting.
  • Winning-offer feedback to retention — the offers that save subscribers at cancellation become the proactive retention campaigns that prevent the next cohort from clicking cancel at all.

Same offer for everyone. Save-rate stuck at 8%.

Most subscription operators run a single generic save flow. The subscriber clicks cancel; the system offers a 10% discount; the subscriber either accepts or completes the cancellation. Same offer for everyone. Save-rate hovers around 8-12% regardless of who the subscriber is or why they are leaving.

Save-flow propensity scoring picks the right offer for each subscriber. A long-tenured high-LTV subscriber facing budget pressure gets a pause flow, not a discount. A short-tenured subscriber facing product-fit issues gets a swap flow, not a pause. A churn-risk subscriber with low engagement gets a win-back gift, not a downgrade. The offer matches the cancellation reason and the subscriber’s propensity to respond to each offer type.

The infrastructure is three components: a propensity-score model trained on cancellation-reason × offer-type × historical save-rate, a save-flow waterfall that picks the highest-propensity offer first, and A/B testable variants per location. Save-rate lifts to 18-30% — a 2-3x improvement over generic offer-spray, without escalating discount magnitude.

The feedback edge is what most save-flow stacks leave on the table. The offers that win at the cancellation moment are the offers that work proactively two weeks earlier. Wire that signal back into comm-broadcast and the next cohort does not arrive at cancel intent.

What is in market — and what each category leaves to you

The cancellation-flow canvas is mature. The propensity model and the feedback wiring are operator-side.

Save-flow specialists — Brightback (ChurnKey), ChurnKey

Strong cancellation-flow editor, offer library, A/B framework, analytics. They ship with the canvas; the propensity model that picks the offer is operator-side.

Billing-platform retention — Recurly Retention, Chargebee, Stripe Billing

Bundled with the billing platform, so the integration is cheap. Editor and offer library are lighter than the specialists. Propensity model still operator-side.

Retention analytics — Profitwell Retain, Baremetrics Cancellation Insights

Excellent for understanding cancellation patterns after the fact. Diagnostic, not in-flow decisioning.

Customer-success platforms — ChurnZero, Catalyst, Vitally, Gainsight

Strong on proactive retention before cancel intent. Not pointed at the in-flow cancellation moment.

The single generic save flow

One screen, one 10% discount, no propensity logic. Most operators sit here for years because the savings math looks fine in the dashboard and the harder waterfall is not a single-engineer project.

The pipeline, end to end

  1. Cancellation-reason classifier. Capture exit-survey responses, click signals, and behavior signals at cancel intent. The reason is structured, not free text.
  2. Propensity-score model.Trained on historical save-rate by cancellation-reason × offer-type × subscriber-segment. Updated weekly as new save outcomes land in the data layer.
  3. Save-offer library. Discount, pause, swap, gift, downgrade, win-back. Each canonical offer type has a template, an eligibility rule, and a compliance overlay.
  4. Per-location override workflow. Corporate offer library is the default; per-location augmentation is allowed through a reviewable workflow that respects the discount ceiling.
  5. Save-flow waterfall logic. Highest-propensity offer first. On rejection, fall through to the next-highest offer. Two-to-three offers max per session before the flow releases to confirm-cancel.
  6. A/B test framework.Variants measured at offer × reason × location × cohort. Statistical significance gates promotion of any variant.
  7. Per-channel UI library. The cancellation experience renders on web, app, email, and phone with the same logic underneath. The propensity model is channel- agnostic.
  8. Compliance gate. Save offers comply with per-vertical regulatory overlays — healthcare, financial, cannabis, telecom. The gate blocks any offer that would breach a per-jurisdiction rule.
  9. Offer-fatigue tracking. Each subscriber carries a save-attempt counter inside a rolling window. The model deprioritizes save offers once fatigue threshold is reached.
  10. Discount-ceiling enforcement. Hard cap on discount magnitude per subscriber per period. Breaks the escalation race that lets subscribers train themselves to churn for bigger offers.
  11. Save-rate observability. Per-offer, per-reason, per-location, per-cohort save-rate dashboards. The model retrains on this same data.
  12. Winning-offer feedback edge. Top-performing save offers feed into proactive retention campaigns sent two-to-four weeks before predicted churn. The save flow stops being the only retention surface; it becomes the last one.

Frequently asked

What is a cancellation flow?

A cancellation flow is the sequence of screens, offers, and decisions a subscriber encounters between clicking "cancel" and either churning or staying. The flow can capture a reason, offer a save (discount, pause, swap, downgrade, gift), route to a human, or simply confirm the cancellation. The shape of the flow is the difference between save-rate at 8% and save-rate at 25%.

Why does the generic 10% save offer fail?

Most save flows offer one thing — a 10% discount — to everyone. A long-tenured high-LTV subscriber facing budget pressure does not want 10% off; they want to pause for two months. A short-tenured subscriber facing product-fit issues does not want a discount; they want to swap into a different plan. Spraying the same offer at everyone produces save-rate around 8-12% and trains subscribers to expect the discount.

What is propensity scoring in a save flow?

A propensity score predicts how likely a given subscriber is to respond to a given offer type. The model is trained on historical save-rate data segmented by cancellation reason × subscriber segment × offer type. At the cancellation moment, the save flow picks the offer with the highest propensity for that subscriber, falls through to the next-highest on rejection, and captures the outcome for future calibration.

How is this different from Brightback (ChurnKey), Recurly Retention, Profitwell Retain, or Chargebee?

Those platforms provide the canvas — the cancellation flow editor, the offer library, the A/B framework, the analytics dashboard. They do not ship with a propensity model trained on your data, the per-location override workflow, the compliance gate per vertical, or the feedback edge that turns winning offers into proactive retention campaigns. That wiring is operator-side.

What is offer fatigue in a save flow?

A subscriber who has already received three save offers in the last six months is less likely to respond to a fourth. Offer-fatigue tracking caps the number of save attempts per subscriber in a rolling window so the save flow does not burn the relationship while trying to save it.

What is the discount-escalation race anti-pattern?

When a save flow lets each rejected offer trigger a bigger discount, subscribers learn to click cancel and watch the offers stack until they hit the magnitude they wanted. Unit economics collapse, and the next acquisition cohort enters with a baked-in expectation of high discounting. The correct shape is propensity-scored selection inside a hard discount ceiling, not waterfall-by-magnitude.

Hire the agent that runs the save flow

The subscription-lifecycle agent owns the propensity model, the save-offer library, the waterfall logic, the compliance gate, and the feedback edge into proactive retention. It runs the full save-flow pipeline across every subscriber, every location, every channel.

Hire the subscription-lifecycle agent

We scope on the call and send a private checkout link after.

Related reading: Multi-location reporting · Auto-PR for API drift