Done-for-you offer · Fractional CMO with AI Swarm · churn-prediction-per-subscriber skill
Completions builds the per-location churn-prediction stack with local-signal features and uplift modeling
Operators running multi-location subscription portfolios work above a strong subscription-billing + churn-modeling + uplift + per-source- signal + lifecycle-orchestration primitives layer (Recharge + Bold Subscriptions + Stripe Billing + Chargebee + Zuora + Maxio + Ordergroove + Skio + Smartrr + Stay AI for ecom subscription; Mariana Tek + Mindbody + Glofox + Pike13 + Zenoti + Booker + Vagaro + WellnessLiving + ClubReady + ABC Financial + Daxko + EZFacility + Hapana + Clubessential for fitness/wellness subscription; XGBoost + LightGBM + CatBoost + TabNet + TabTransformer + FT-Transformer + lifelines + DeepSurv + DeepHit + Random Survival Forest for churn modeling; CausalML + DoubleML + EconML + PyMC + Stan for uplift modeling; Yext + Birdeye + Podium for review/NPS; Weather.com for weather; Google Maps API for proximity; Eventbrite for local events; Klaviyo + Iterable + Braze + Customer.io + Twilio + Bandwidth + MessageBird for lifecycle orchestration). The orchestration layer that sits above those primitives — per-subscriber churn probability, per-location local- signal feature engineering, per-cohort uplift modeling, per- subscriber pre-emptive intervention engine, per-intervention treatment-assignment protocol, per-vertical compliance overlay, closed-feedback-loop coordination across the subscription- management swarm — is operator-side architecture. Completions builds the churn-prediction-per-subscriber skill on the subscription-management agent end-to-end and coordinates it with the peer-cohort-computation, per-location-rollup-reporting, master-record-canonicalization, and multi-stream-subscription siblings. Operator owns every artifact: subscriber data, per-location signal data, churn model, uplift model, intervention engine, compliance overlay, audit trail. Operator can in-house at any time.
Published September 24, 2026
What we operate every week
Per-subscriber churn probability via a gradient-boosted tree ensemble (XGBoost + LightGBM + CatBoost) plus deep-tabular models (TabNet + TabTransformer + FT-Transformer) plus survival analysis (Cox proportional hazards + DeepSurv + DeepHit + Random Survival Forest) with per-subscriber 7-day, 14-day, 21-day, 30-day, 60-day, and 90-day churn-probability forecasts that operator data-science maintains.
Per-location local-signal feature engineering with a per-location feature catalog operator data-science maintains — per-location visit recency, visit frequency, class attendance, service usage, staff turnover, facility maintenance, community NPS, competitor proximity, weather, local-event calendar, seasonality, demographic shift, commute time, parking availability, and amenity availability.
Per-cohort uplift modeling via the CATE meta-learner ensemble (T-learner + S-learner + X-learner + DR-learner + CausalML + DoubleML + EconML + Bayesian-treatment-effect + counterfactual-prediction + causal-forest) emitting per- subscriber per-intervention uplift estimates against operator- maintained holdouts.
Per-subscriber pre-emptive intervention engine routes 7-21-day- ahead at-risk subscribers to per-intervention treatment (per-subscriber outreach, save offer, retention call, location transfer, pause membership, downgrade, concierge) with per- intervention budget cap operator finance maintains and per- vertical compliance overlay (HIPAA when health-adjacent subscription + FDA OPDP when pharma + DEA when controlled- substance + FTC Negative Option Rule + FTC Click-to-Cancel + state subscription auto-renewal laws + ROSCA + CCPA + GDPR + TCPA + CAN-SPAM + state CARD Act + 10DLC) expressed as a policy- as-code gate (OPA Rego, AWS Cedar, Casbin, Cerbos, Oso) that operator counsel reviews.
Where the orchestration above subscription, churn-modeling, uplift, and lifecycle primitives compounds at portfolio scale
The vendor primitives are strong. Subscription-billing vendors handle billing and cancellation flows. Churn-modeling libraries handle survival analysis and gradient-boosted classification. Uplift-modeling libraries handle CATE estimation. Per-source signal vendors expose review, NPS, weather, proximity, and local-event data. Lifecycle vendors handle multichannel delivery. The orchestration above those primitives is what compounds at multi-location subscription scale. Per-subscriber churn probability sits above the churn-modeling ensemble with operator-data-science-maintained forecast horizons. Per-location local-signal feature engineering sits above the per-source ingestion layer with operator-versioned feature catalog. Per- cohort uplift modeling sits above the CATE meta-learner ensemble with operator-maintained holdouts. Per-subscriber pre-emptive intervention engine sits above lifecycle vendors with operator-counsel-approved treatment-assignment protocol. Per-vertical compliance overlay sits above a policy-as-code gate (OPA Rego, AWS Cedar, Casbin, Cerbos, Oso) that counsel reviews. Closed-feedback-loop coordination sits above the swarm orchestration layer.
Completions coordinates this orchestration layer under a Tier 3 Fractional CMO with AI Swarm engagement (1-2 days/wk embedded), tying the churn-prediction-per-subscriber skill to the peer-cohort-computation, per-location-rollup-reporting, master-record-canonicalization, and multi-stream-subscription siblings.
How the engagement progresses
Tier 1 AI Readiness Assessment (2-3 weeks, diagnostic). Completions audits the current operation across seven axes. Deliverable: gap-pack report with per-subscriber per-intervention expected-uplift estimate + projected ARR retention gain.
Tier 2 AI Swarm Setup Sprint (4-8 weeks, build with 30-day operating tail). Completions builds the per-location churn-prediction stack on operator infrastructure — per-location-signal-aware-churn-model + per-subscriber-churn-probability + pre-emptive-churn-intervention-triggers + per-subscriber-uplift-modeling on the subscription-management agent + multi-platform-subscription-billing-ingestion on multi-stream-subscription + save-offer-library-management on save-flow + subscriber-lifecycle-cadence on lifecycle-flow.
Tier 3 Fractional CMO with AI Swarm ( month, 6-month minimum, 1-2 days/wk embedded). Continues operating the stack with weekly per-subscriber churn probability refresh + monthly per-location feature engineering refresh + quarterly per-cohort uplift modeling refresh + per-event compliance overlay refresh + cross-agent swarm coordination.
Frequently asked
What does "Completions builds the per-location churn-prediction stack — local-signal features + uplift modeling + intervention" actually deliver?
Completions builds and coordinates per-subscriber churn probability + per-location local-signal feature engineering + per-cohort uplift modeling + per-subscriber pre-emptive intervention across the operator multi-location subscription base, above a strong subscription-billing + churn-modeling + uplift + per-source-signal + lifecycle-orchestration primitives layer (Recharge + Bold Subscriptions + Stripe Billing + Chargebee + Zuora + Maxio + Ordergroove + Skio + Smartrr + Stay AI for ecom subscription; Mariana Tek + Mindbody + Glofox + Pike13 + Zenoti + Booker + Vagaro + WellnessLiving + ClubReady + ABC Financial + Daxko + EZFacility + Hapana + Clubessential for fitness/wellness subscription; XGBoost + LightGBM + CatBoost + TabNet + TabTransformer + FT-Transformer + lifelines + DeepSurv + DeepHit + Random Survival Forest for churn modeling; CausalML + DoubleML + EconML + PyMC + Stan for uplift modeling; Yext + Birdeye + Podium for review/NPS signal; Weather.com for weather signal; Google Maps API for proximity signal; Eventbrite for local-event signal; Klaviyo + Iterable + Braze + Customer.io + Twilio + Bandwidth + MessageBird for lifecycle orchestration). Per-subscriber churn probability via gradient-boosted tree ensemble (XGBoost + LightGBM + CatBoost) + deep-tabular models (TabNet + TabTransformer + FT-Transformer) + survival analysis (Cox proportional hazards + DeepSurv + DeepHit + Random Survival Forest) with per-subscriber 7-day + 14-day + 21-day + 30-day + 60-day + 90-day churn-probability forecasts that operator data-science maintains. Per-location local-signal feature engineering with a per-location feature catalog operator data-science maintains (visit recency + visit frequency + class attendance + service usage + staff turnover + facility maintenance + community NPS + competitor proximity + weather + local-event calendar + seasonality + demographic shift + commute time + parking availability + amenity availability). Per-cohort uplift modeling via the CATE meta-learner ensemble (T-learner + S-learner + X-learner + DR-learner + CausalML + DoubleML + EconML + Bayesian-treatment-effect + counterfactual-prediction + causal-forest) emitting per-subscriber per-intervention uplift estimates against operator-maintained holdouts. Per-subscriber pre-emptive intervention engine routes 7-21-day-ahead at-risk subscribers to per-intervention treatment (outreach + save offer + retention call + location transfer + pause membership + downgrade + concierge) with per-intervention budget cap operator finance maintains and per-vertical compliance overlay (HIPAA when health-adjacent subscription + FDA OPDP when pharma + DEA Schedule II-V when controlled-substance + FTC Negative Option Rule + FTC Click-to-Cancel + state subscription auto-renewal laws + ROSCA + CCPA + GDPR + TCPA + CAN-SPAM + state CARD Act + 10DLC) expressed as a policy-as-code gate (OPA Rego, AWS Cedar, Casbin, Cerbos, Oso) on every intervention. Operator owns every artifact: subscriber data, per-location signal data, churn model, uplift model, intervention engine, compliance overlay, audit trail. Completions owns the swarm orchestration on the subscription-management agent and its coordination with the peer-cohort-computation, per-location-rollup-reporting, master-record-canonicalization, and multi-stream-subscription siblings.
Where does the orchestration layer above subscription + churn-modeling + uplift + lifecycle primitives compound at multi-location scale?
Subscription-billing vendors (Recharge + Bold Subscriptions + Stripe Billing + Chargebee + Zuora + Maxio + Ordergroove + Skio + Smartrr + Stay AI for ecom subscription; Mariana Tek + Mindbody + Glofox + Pike13 + Zenoti + Booker + Vagaro + WellnessLiving + ClubReady + ABC Financial + Daxko + EZFacility + Hapana + Clubessential + Compete for fitness/wellness subscription), churn-modeling tooling (XGBoost + LightGBM + CatBoost + TabNet + TabTransformer + FT-Transformer + lifelines + scikit-survival + DeepSurv + DeepHit + Random Survival Forest), uplift-modeling tooling (CausalML + DoubleML + EconML + PyMC + Stan + EconomicSML), per-source signal vendors (Yext + Birdeye + Podium for review/NPS; Weather.com + OpenWeather for weather; Google Maps API for proximity; Eventbrite for local events; per-location POS for visit/usage), and lifecycle orchestration (Klaviyo + Iterable + Braze + Customer.io + Twilio + Bandwidth + MessageBird) ship strong primitives. The orchestration layer above them is operator-side architecture, and at multi-location subscription scale it spans seven workstreams: (1) per-subscriber churn probability above a gradient-boosted + deep-tabular + survival ensemble with per-subscriber forecast horizons (7-day, 14-day, 21-day, 30-day, 60-day, 90-day) operator-counsel-reviewed; (2) per-location local-signal feature engineering above the per-source ingestion layer with operator-versioned per-location feature catalog; (3) per-cohort uplift modeling above the CATE meta-learner ensemble (T-learner + S-learner + X-learner + DR-learner + CausalML + DoubleML + EconML + Bayesian-treatment-effect + counterfactual-prediction + causal-forest) with operator-maintained holdouts; (4) per-subscriber pre-emptive intervention engine above lifecycle vendors with operator-counsel-approved treatment-assignment protocol and budget cap operator finance maintains; (5) per-vertical compliance overlay (HIPAA when health-adjacent subscription + FDA OPDP when pharma + DEA when controlled-substance + FTC Negative Option Rule + FTC Click-to-Cancel + state subscription auto-renewal laws + ROSCA + CCPA + GDPR + TCPA + CAN-SPAM + state CARD Act + 10DLC) expressed as a policy-as-code gate (OPA Rego, AWS Cedar, Casbin, Cerbos, Oso) that counsel reviews; (6) per-intervention treatment-assignment protocol (random-assignment + propensity-score-matching + counterfactual-validation) above operator-data-science-team-maintained experimental design; (7) cross-agent swarm coordination across peer-cohort-computation, per-location-rollup-reporting, master-record-canonicalization, and multi-stream-subscription siblings. Completions coordinates this orchestration layer under a Tier 3 Fractional CMO with AI Swarm engagement.
What does the engagement look like across Tier 1 → Tier 2 → Tier 3?
Tier 1 AI Readiness Assessment (2-3 weeks, diagnostic): Completions audits the operator current churn-prediction operation across seven axes — per-subscriber churn probability coverage + per-location local-signal feature engineering maturity + per-cohort uplift modeling + per-subscriber pre-emptive intervention + per-vertical compliance overlay + per-intervention treatment-assignment-protocol + cross-agent swarm coordination. Deliverable: gap-pack report with per-subscriber per-intervention expected-uplift estimate + projected ARR retention gain. Tier 2 AI Swarm Setup Sprint (4-8 weeks, build with 30-day operating tail): Completions builds the per-location churn-prediction stack on operator infrastructure — per-location-signal-aware-churn-model + per-subscriber-churn-probability + pre-emptive-churn-intervention-triggers + per-subscriber-uplift-modeling on subscription-management agent + multi-platform-subscription-billing-ingestion on multi-stream-subscription + save-offer-library-management on save-flow + subscriber-lifecycle-cadence on lifecycle-flow + per-location-signal-feature-engineering on per-location-rollup-reporting + peer-cohort-computation on location-benchmarking. Tier 3 Fractional CMO with AI Swarm (6-month minimum, 1-2 days/wk embedded): Completions continues operating the stack with weekly per-subscriber churn probability refresh + monthly per-location feature engineering refresh + quarterly per-cohort uplift modeling refresh + per-event compliance overlay refresh + cross-agent swarm coordination.
Who owns the subscriber data, local-signal features, uplift models, and intervention engine?
Operator owns 100% of every artifact: subscriber data (in operator data infrastructure — Snowflake + Databricks + BigQuery + Redshift + Postgres operator data warehouse), per-location signal data (visit data + staff data + facility data + community NPS data + competitor data + weather data + local-event data + demographic data in operator data warehouse), per-source ingestion credentials (Mariana Tek + Mindbody + Glofox + Pike13 + Zenoti + Booker + Vagaro + WellnessLiving + ClubReady + ABC Financial + Daxko + EZFacility + Hapana + Clubessential + Compete subscription-billing system + per-location POS + Yext + Birdeye + Podium operator NPS + Weather.com + Google Maps API + Eventbrite under operator billing + operator credentials), churn model code (in operator repo with operator-controlled deploy pipeline), per-location feature engineering code (operator-owned), uplift model code (operator-owned + operator-data-science-team-aligned), per-intervention treatment-assignment-protocol (operator-owned + operator-counsel-aligned), per-intervention budget cap config (operator-owned + operator-finance-team-aligned), per-vertical compliance overlay (rule library in operator repo with attorney-approved updates), HIPAA + FDA + DEA + FTC Negative Option + FTC Click-to-Cancel + state auto-renewal + ROSCA + CCPA + GDPR + TCPA + CAN-SPAM + CARD Act + 10DLC disclosure register (operator-owned + operator-counsel-maintained), brand spec (versioned in operator repo), LLM prompts (in operator repo), audit trail (retention infrastructure on operator cloud account with WORM-storage when HIPAA + FDA + FTC + state auto-renewal retention required). Completions owns: the orchestration knowledge — how to design per-location signal-aware churn models + how to tune per-cohort uplift modeling + how to debug per-intervention treatment-assignment cascades + how to coordinate the stack with peer-cohort + per-location-rollup + master-record-canonicalization + multi-stream-subscription siblings. The operator can in-house at any time; Completions credentials revoke immediately on engagement-end.
What does Completions commit to reporting on a Tier 3 engagement?
Tier 3 engagements report against a pre-engagement baseline that the Tier 1 assessment establishes for the operator stack. The reporting cycle covers six workstreams: (1) per-subscriber churn probability prediction observed via AUC + Brier score + calibration plots against rolling holdouts that operator-side analysts maintain, with per-cohort breakdowns reported; (2) per-location local-signal feature ingestion coverage observed across the per-source surface, with per-source event-emission completeness and ingestion lag reported; (3) per-cohort uplift modeling surfacing observed via the CATE meta-learner ensemble against operator-labeled holdouts, with per-treatment confidence intervals and propensity-score diagnostics reported; (4) per-subscriber pre-emptive intervention coverage observed for at-risk subscribers in the 7-21-day-ahead window, with per-treatment-assignment protocol diagnostics (random-assignment + propensity-score-matching + counterfactual-validation) reported; (5) per-intervention budget cap adherence observed at the per-cohort grain against operator-finance-team-maintained budget allocations; (6) per-vertical compliance overlay pass rate observed across HIPAA + FDA OPDP + DEA + FTC Negative Option Rule + FTC Click-to-Cancel + state subscription auto-renewal laws + ROSCA + CCPA + GDPR + TCPA + CAN-SPAM + state CARD Act + 10DLC scope. Monthly reports surface to the C-suite + CS leadership + counsel + finance with per-cohort breakdowns. Caveats: per-source vendor API rate limits + per-source ingestion completeness + data-warehouse query budget + LLM-vendor availability + per-statute retention windows shifting with operator counsel policy sit outside Completions control and are reported alongside observed performance; attorney-client privilege on counsel-reviewed intervention rules is preserved through every layer of the reporting cycle. Completions does not commit to fixed numeric SLAs on prediction AUC, ingestion coverage, uplift confidence, intervention coverage, churn-rate reduction, or ARR retention when those KPIs depend on vendor performance, data infrastructure availability, or counsel policy decisions.
How does engagement end and what is the operator transition path?
Tier 3 engagements are 6-month minimum with 90-day notice. At engagement end, Completions transitions the per-location churn-prediction operation back to operator in-house in 30-60 days: operating-playbook hand-off + in-house staff training across 3-5 operator team members covering per-subscriber churn probability modeling + per-location local-signal feature engineering + per-cohort uplift modeling + per-subscriber pre-emptive intervention engine + per-intervention treatment-assignment-protocol + per-vertical compliance overlay management + cross-agent coordination + subscriber data infrastructure hand-off + per-source ingestion credentials hand-off + churn model code hand-off + LLM prompts hand-off + audit trail hand-off with WORM-storage operator-account-ownership confirmation; Completions credentials revoke immediately on engagement-end. Operator can re-engage Completions at any time on Tier 1 or Tier 2 cadence.
Engage Completions
Start with the AI Readiness Assessment (Tier 1, 2-3 weeks). Hand off to Tier 2 (4-8 weeks) for the build. Continue under Tier 3 Fractional CMO with AI Swarm (6-month minimum, 1-2 days/wk embedded). Operator owns every artifact at every tier. Operator can in-house at any time.