Save flow · Propensity scoring · Subscription DTC

Your generic 10% save offer is teaching every subscriber to cancel-then-discount. Score the offer per subscriber at the cancel moment.

You run a subscription DTC business. Your save flow shows a 10% discount to every subscriber clicking cancel. Subscribers who would have paused for two months take the 10% and keep going. Subscribers who needed a swap into a different plan take the 10%, churn next month, and now expect the discount. Gift-givers who wanted to convert into a one-time gift cancel because the discount does not fit their actual need. The cancellation surface has become the cheapest way to buy from you, and the data is training subscribers on the wrong incentive. Propensity-scored save flow picks per-subscriber offer (discount, pause, swap, downgrade, gift) from cancellation reason × subscriber segment × historical save signal.

Audit your save flow (Tier 1)Explore Fractional CMO operation

Published May 30, 2026

The five offer shapes a generic discount cannot reach

Discount. 10-25% off the next billing cycle. Works for price-sensitive subscribers who would have stayed at a lower price point. Wrong answer for long-tenured subscribers facing budget pressure or short-tenured subscribers facing fit issues.

Pause. 1-3 month suspension of billing + delivery. Works for budget-pressure, life-event, travel-related cancellations. Long-tenured high-LTV subscribers convert at much higher rates on pause than on discount.

Swap. Different plan, different SKU, different frequency. Works for fit issues — the subscriber chose wrong at signup and a different configuration would have worked. Short-tenured subscribers benefit disproportionately.

Downgrade. Lower-tier plan at lower price. Works for active-usage subscribers who use less than the tier they bought into. Captures revenue you would have lost while keeping the subscriber in the ecosystem.

Gift. Convert recurring subscription into a one-time gift purchase. Works for gift-givers and gift-receivers who signed up for the wrong shape. Often produces a healthier customer relationship than the cancellation would have.

The right offer per subscriber depends on cancellation reason × subscriber segment × historical save signal. The propensity model picks; the waterfall falls through on rejection; the A/B framework keeps the model honest.

We’ve built save-flow propensity scoring for subscription DTC operators. Here’s what we know.

You probably already use a save-flow vendor — Loop, Recharge, Smartrr, Stay AI, ProsperStack, Churnkey. Each is good at the cancellation-flow execution primitive. The gap is the propensity model + waterfall fall-through + A/B framework that converts the per-subscriber input into the per-subscriber best offer. We bring the propensity model architecture, the per-segment offer catalog template, and the A/B framework playbook.

We have built this for subscription DTC operators across categories (consumables, apparel, supplements, content subscriptions). We know which cancellation reasons dominate per category (price for consumables; fit for apparel + supplements; engagement for content). We bring the per-category reason taxonomy starter and the per-segment offer prior so the first 30 days of propensity scoring produces calibrated output rather than random selection.

How we get from generic 10% to per-subscriber propensity

Step 1 — Tier 1 AI Readiness Assessment ($10k, 2-3 weeks). We audit your current save flow. We sample 30-90 days of cancellations against the offers presented and the outcomes. Output: the propensity model specification, the per-segment per-reason offer catalog map, the per-category reason taxonomy, and the A/B framework plan.

Step 2 — Tier 2 AI Swarm Setup Sprint ($25-50k, 4-8 weeks). We build the save flow end-to-end: propensity model, per-segment offer catalog wiring, waterfall logic, per-state compliance overlay integration, A/B framework, outcome capture into the propensity-training data store. Your engineering + growth team receives the running system, all source code, all credentials.

Step 3 — Tier 3 Fractional CMO with AI Swarm ($15-25k/month, 6-month minimum, 1-2 days/wk). We operate the save flow in production. Propensity model retrained weekly. Per-segment waterfall reviewed monthly with growth + retention leadership. A/B framework continuously running. Roll up weekly per-cohort save trend reports.

What changes for you

You stop watching the cancellation surface train subscribers to expect a discount. The propensity-scored save flow presents pause to subscribers who want pause, swap to subscribers who want swap, gift to gift-givers, and discount only to subscribers for whom discount is the right answer.

You stop guessing at the next quarter’s save flow A/B. The continuous A/B framework runs on a documented portion of traffic; the propensity model learns from the outcomes; the next month’s offers compound from this month’s learning.

You can answer the question your CFO asks every monthly business review: how much LTV did the save flow preserve, segmented by cancellation reason and by offer type. The per-flow outcome record is the answer.

You can onboard a new product line or vertical with the per-category reason taxonomy extended rather than re-architecting the save flow from scratch.

Frequently asked

Why does the generic 10% save offer fail at scale?

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. A gift-giver wants to convert into a one-time gift purchase rather than a recurring sub. Spraying the same offer at everyone produces flat save rates in the category and trains subscribers to expect the discount. Within 6-12 months, the cancellation surface becomes the cheapest way to buy from you.

What is propensity scoring in a save flow, and how is it different from a static rule?

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. A static rule (if budget-pressure then 15% off) is a special case where every subscriber in a segment sees the same offer. Propensity scoring extends the rule with cross-segment learning, falls through on rejection, and updates the model from outcomes — three behaviors a static rule cannot do.

What is a save-flow waterfall, and how is it structured?

The waterfall is the ordered sequence of offers presented during a single save flow. First offer is the highest-propensity offer for that subscriber. On rejection, the second offer is the next-highest propensity (often a different offer type — if discount was first, pause is second). On rejection of the second, a third may surface (often a route-to-human path for high-LTV subscribers). Each step is logged. The model learns from the full waterfall (which offer at which step saved or did not save) rather than just the final-step outcome. The waterfall is what makes the propensity model improve cycle over cycle.

What does Completions commit to on Tier 3 if we run the save flow in production?

Tier 3 process commitments include: propensity model retrained weekly on the latest save-flow outcomes; per-segment per-cancellation-reason waterfall reviewed monthly with your growth + retention leadership; per-state compliance overlay applied to every offer at evaluation time; A/B framework continuously running on a documented portion of the save-flow traffic; weekly per-cohort save-flow trend report routed to growth + finance leadership. We commit to the operating discipline. Per-cohort save-rate precision is tuned per stack and recorded as engagement KPIs.

Who owns the propensity model, the offer catalog, and the credentials post-engagement?

Your team owns the propensity model weights, the cancellation-reason taxonomy, the per-segment offer catalog, the per-state compliance overlay, the A/B configuration, and the credentials. Completions owns the orchestration knowledge: the propensity model retraining runbook, the per-segment waterfall tuning history, the A/B framework playbook. At engagement end we transition operational ownership back to your team over 30-60 days with documented handover.

How does the save flow connect to the rest of the retention + revenue stack?

The save flow subscribes upstream to the pre-emptive intervention triggers (at-risk customers get pre-cancellation outreach; only the customers who still click cancel land in the save flow), to the save-offer library management (the canonical offer catalog), to the per-jurisdiction overlay (per-state compliance gate on every offer), and to the change-event-emission stream (customer-state changes that affect propensity). It publishes downstream: per-flow outcome events to the recovery-rate dashboard and the lifecycle-flow architecture (saved subscribers get a different follow-up than churned). Four upstream feeds, two downstream consumers, one save contract.

Start with the audit

Tier 1 AI Readiness Assessment ($10k, 2-3 weeks): we audit your current save flow, sample 30-90 days of cancellations against offers + outcomes, and produce the propensity model spec + per-segment per-reason offer catalog map + per-category reason taxonomy + A/B framework plan. If you decide to build, Tier 2 ships the save flow. If you decide to operate it with us, Tier 3 runs the weekly retraining + monthly waterfall review in production. You choose the next step at each gate.