Completions

Data-layer swarm · Customer-Data-Graph Agent · Ltv-math-primitives skill · Build pillar · Published July 11, 2026

How to build per-location per-channel per-brand CLV at runtime

This guide explains how to architect the ltv-math-primitives skill on the customer-data-graph agent end-to-end at multi-location multi-channel multi-brand franchise scale: per-portfolio per-location per-channel per-brand per-customer per-canonical-per-customer-cohort-assignment-spec + per-per-cohort-CLV-model-ensemble-spec + per-per-customer-embedding-spec + per-per-customer-RFM-spec + per-per-cohort-recalibration-spec + per-per-customer-runtime-lookup-spec + per-per-customer-CLV-confidence-interval-spec + per-per-customer-CLV-explainability-spec + per-per-customer-CLV-cross-skill-handoff + per-per-customer-CLV-FBC-feedback-loop + per-per-customer-CLV-audit-trail + per-portfolio audit-trail.

What you will build

  • Per-portfolio per-canonical-per-customer-cohort-assignment-spec across per-location + per-channel + per-brand + per-vertical + per-tenure (0-30 + 30-90 + 90-180 + 180-365 + 365-day-plus) + per-RFM + per-seasonality + per-loyalty-tier + per-subscription-status + per-acquisition-source + per-first-order-SKU + per-multi-LLM-classification + per-confidence + per-cohort-membership-versioning + per-behavioral-cohort-computation-handoff.
  • Per-canonical-per-cohort-CLV-model-ensemble-spec — per-BG-NBD-Beta-Geometric-NBD + per-Pareto-NBD + per-Gamma-Gamma-monetary-value + per-MBG-NBD-Modified-Beta-Geometric-NBD + per-CDNOW-Schmittlein-Morrison-Colombo + per-Lifetimes-Python-library + per-deep-learning-LSTM + per-Transformer + per-Convolutional-Sequence-to-Sequence + per-stacking-blending + per-cross-validation + per-hyperparameter-tuning + per-A/B-test + per-MLflow + per-Weights-Biases + per-Comet-ML.
  • Per-canonical-per-customer-embedding-spec + per-RFM + per-cohort-recalibration — per-purchase-history-embedding + per-clickstream-embedding + per-engagement-embedding + per-SKU-embedding + per-Pinecone + per-Weaviate + per-Qdrant + per-Milvus + per-Chroma + per-pgvector + per-Vertex-AI-Vector-Search + per-embedding-drift-monitoring + per-recency + per-frequency + per-monetary-value + per-RFM-quintile + per-RFM-percentile + per-RFM-cross-cohort-percentile + per-Platt-scaling + per-isotonic-regression + per-temperature-scaling + per-Bayesian-binning + per-expected-calibration-error + per-reliability-diagram + per-Brier-score + per-daily + per-weekly + per-monthly recalibration cadence + per-FBC-pattern-learning.
  • Per-canonical-per-customer-runtime-lookup-spec + per-confidence-interval + per-explainability — per-Redis + per-DynamoDB + per-Cassandra + per-ScyllaDB + per-Aerospike + per-TTL + per-warm + per-invalidation + per-real-time-recompute + per-batch-recompute + per-streaming-recompute (Apache Flink + Spark Streaming + Materialize + RisingWave) + per-p50-p95-p99-latency + per-runtime-FBC-pattern-learning + per-Bayesian-posterior + per-frequentist + per-bootstrap + per-4-tier-confidence + per-SHAP + per-LIME + per-feature-importance + per-counterfactual + per-attribution-graph + per-multi-LLM-narrative + per-chain-of-thought.
  • Per-canonical-per-customer-CLV-cross-skill-handoff + per-FBC-feedback + per-audit-trail — per-handoff-to-per-member-monthly-clv + per-handoff-to-per-member-next-best-action + per-handoff-to-churn-prediction-per-subscriber + per-handoff-to-tier-transition-timing + per-handoff-to-save-flow-propensity-scoring + per-handoff-to-cross-location-offer-coordination + per-handoff-to-foot-traffic-integration + per-handoff-to-cross-touchpoint-identity-resolution + per-handoff-to-attribution-event-emission + per-realized-vs-predicted-feedback + per-model-recalibration + per-cohort-reassignment + per-embedding-recalibration + per-feature-importance-recalibration + per-A/B-test-promotion-demotion + per-per-customer-CLV-canonical-audit-record + per-CCPA-CPRA-DSAR + per-GDPR-DPIA + per-PIPEDA-OPC + per-LGPD-ANPD + per-EU-AI-Act-Article-13-transparency-disclosure + per-EU-AI-Act-Article-14-human-oversight + per-NIST-AI-RMF-attestation + per-immutable-WORM-storage.

Why per-vendor-Klaviyo-Predictive-CLV-account-level-CLV breaks at multi-location multi-channel multi-brand franchise scale

Per-vendor-Klaviyo-Predictive-CLV-canonical-account-level-CLV ships per-account per-customer per-90-day-prediction primitive. Per-vendor-Lifetimes-Python-library + Theta-Equity + Retention-Science + Optimove + Custora-Optimove + Glew + Polar-Analytics + Tresl + Daasity + Twilio-Engage + Segment-Customer-Profiles + Amplitude-Audiences + Hightouch-canonical-account-level-CLV ship per-vendor per-native account-level CLV primitives.

At 1-account-1-channel-1-brand scale per-account per-customer per-90-day-prediction primitive is enough. At multi-location multi-channel multi-brand franchise scale per-per-customer-cohort-assignment-spec + per-per-cohort-CLV-model-ensemble-spec + per-per-customer-embedding-spec + per-per-customer-RFM-spec + per-per-cohort-recalibration-spec + per-per-customer-runtime-lookup-spec + per-per-customer-CLV-confidence-interval-spec + per-per-customer-CLV-explainability-spec + per-per-customer-CLV-cross-skill-handoff + per-per-customer-CLV-FBC-feedback-loop + per-per-customer-CLV-audit-trail.

Per-account-customer-90-day-prediction + per-per-customer-cohort-assignment-blind + per-per-cohort-CLV-model-ensemble-blind + per-per-customer-embedding-blind + per-per-cohort-recalibration-blind + per-per-customer-runtime-lookup-blind + per-per-customer-CLV-confidence-interval-blind + per-per-customer-CLV-explainability-blind + per-per-customer-CLV-cross-skill-handoff-blind + per-per-customer-CLV-FBC-feedback-loop-blind + per-multi-format-audit-trail-blind + per-EU-AI-Act-Article-13-Article-14-blind + per-NIST-AI-RMF-blind.

The operator-side architecture above per-vendor-CLV-prediction primitive is canonical-per-customer-cohort-assignment-spec + per-per-cohort-CLV-model-ensemble-spec + per-per-customer-embedding-spec + per-per-customer-RFM-spec + per-per-cohort-recalibration-spec + per-per-customer-runtime-lookup-spec + per-per-customer-CLV-confidence-interval-spec + per-per-customer-CLV-explainability-spec + per-per-customer-CLV-cross-skill-handoff + per-per-customer-CLV-FBC-feedback-loop + per-per-customer-CLV-audit-trail + per-portfolio-audit-trail.

What is in market today

Per-platform per-CLV-prediction-vendor

Klaviyo Predictive CLV, Lifetimes Python library, Theta Equity, Retention Science, Optimove, Custora (Optimove), Glew, Polar Analytics, Tresl, Daasity, Twilio Engage, Segment Customer Profiles, Amplitude Audiences, Hightouch, Northbeam, Rockerbox, HockeyStack. Per-account per-customer per-90-day-prediction primitive. Per-canonical-per-customer-cohort-assignment-canonical-per-cohort-CLV-model-ensemble-canonical-per-customer-embedding-canonical-per-customer-RFM-canonical-per-cohort-recalibration-canonical-per-customer-runtime-lookup-canonical-per-customer-CLV-confidence-canonical-per-customer-CLV-explainability-canonical-per-customer-CLV-cross-skill-handoff-canonical-per-customer-CLV-FBC-feedback-canonical-per-customer-CLV-audit-trail is not the primitive.

Per-platform per-vector-store-vendor

Pinecone, Weaviate, Qdrant, Milvus, Chroma, pgvector, Vertex AI Vector Search, AWS OpenSearch Serverless, Azure AI Search, Elasticsearch Vector Search, Vespa. Per-account per-index per-embedding primitive. Per-canonical-per-customer-purchase-history-embedding-canonical-per-customer-clickstream-embedding-canonical-per-customer-engagement-embedding-canonical-per-customer-SKU-embedding-canonical-per-customer-embedding-drift-monitoring is not the primitive.

Per-platform per-runtime-cache-vendor

Redis, DynamoDB, Cassandra, ScyllaDB, Aerospike, Memcached, KeyDB, Dragonfly, ElastiCache, MemoryDB. Per-account per-key per-TTL primitive. Per-canonical-per-customer-CLV-cache-canonical-per-customer-CLV-cache-TTL-canonical-per-customer-CLV-cache-warm-canonical-per-customer-CLV-cache-invalidation-canonical-per-customer-CLV-runtime-latency-p50-p95-p99 is not the primitive.

Per-platform per-stream-processing-vendor

Apache Flink, Apache Spark Streaming, Materialize, RisingWave, Apache Beam, Kafka Streams, Decodable, Estuary Flow, ksqlDB, Apache Pulsar Functions. Per-account per-stream per-job primitive. Per-canonical-per-customer-CLV-real-time-recompute-canonical-per-customer-CLV-batch-recompute-canonical-per-customer-CLV-streaming-recompute-canonical-per-customer-CLV-runtime-FBC-pattern-learning is not the primitive.

How the architecture is built

  1. Per-portfolio per-canonical-per-customer-cohort-assignment-substrate. Per-11-cohort-dimensions + per-multi-LLM-classification + per-confidence + per-cohort-membership-versioning + per-behavioral-cohort-computation-handoff canonical-cohort.
  2. Per-portfolio per-canonical-per-cohort-CLV-model-ensemble-spec. Per-9-CLV-model-types + per-stacking-blending + per-cross-validation + per-hyperparameter-tuning + per-A/B-test + per-MLflow-Weights-Biases-CometML canonical-ensemble.
  3. Per-portfolio per-canonical-per-customer-embedding-spec. Per-purchase-history + per-clickstream + per-engagement + per-SKU + per-7-vector-stores + per-drift-monitoring canonical-embedding.
  4. Per-portfolio per-canonical-per-customer-RFM-spec. Per-recency + per-frequency + per-monetary + per-quintile + per-percentile + per-cross-cohort-percentile canonical-RFM.
  5. Per-portfolio per-canonical-per-cohort-recalibration-spec. Per-Platt + per-isotonic + per-temperature + per-Bayesian + per-ECE + per-reliability + per-Brier + per-3-cadence-tiers + per-FBC-pattern-learning canonical-recalibration.
  6. Per-portfolio per-canonical-per-customer-runtime-lookup-spec. Per-5-cache-vendors + per-TTL + per-warm + per-invalidation + per-real-time-recompute + per-batch-recompute + per-streaming-recompute + per-4-stream-vendors + per-p50-p95-p99-latency + per-FBC-pattern-learning canonical-runtime-lookup.
  7. Per-portfolio per-canonical-per-customer-CLV-confidence-interval-spec. Per-Bayesian-posterior + per-frequentist + per-bootstrap + per-4-tier-confidence canonical-confidence.
  8. Per-portfolio per-canonical-per-customer-CLV-explainability-spec. Per-SHAP + per-LIME + per-feature-importance + per-counterfactual + per-attribution-graph + per-multi-LLM-narrative + per-chain-of-thought canonical-explainability.
  9. Per-portfolio per-canonical-per-customer-CLV-cross-skill-handoff. Per-9-sibling-skills canonical-handoff.
  10. Per-portfolio per-canonical-per-customer-CLV-FBC-feedback-loop. Per-realized-vs-predicted + per-model-recalibration + per-cohort-reassignment + per-embedding-recalibration + per-feature-importance-recalibration + per-A/B-promotion-demotion canonical-FBC.
  11. Per-portfolio per-canonical-per-customer-CLV-audit-trail + per-portfolio-audit-trail. Per-customer-CLV-canonical-audit-record + per-CCPA-CPRA-DSAR + per-GDPR-DPIA + per-PIPEDA-OPC + per-LGPD-ANPD + per-EU-AI-Act-Article-13 + per-EU-AI-Act-Article-14 + per-NIST-AI-RMF + per-immutable-WORM canonical-audit.
  12. Per-portfolio per-customer-data-graph-agent-canonical-bundle. Per-ltv-math-primitives + per-versioned-customer-history + per-cross-touchpoint-identity-resolution + per-deterministic-probabilistic-hybrid-identity-resolution + per-behavioral-signal-ingestion-customer-graph + per-per-field-conflict-resolution-policy + per-customer-data-orchestration + per-identity-resolution-software + per-master-record-sync + per-dsar-software + per-customer-journey-tracking canonical-bundle.
  13. Per-portfolio per-canonical-end-to-end-SLA. Per-cohort-assign-to-model-score-to-calibrate-to-cache-to-explain-to-handoff-SLA canonical-end-to-end-SLA.

Frequently asked questions

What is per-location per-channel per-brand CLV at runtime for multi-location franchises?

Per-location per-channel per-brand CLV at runtime runs per-portfolio per-location per-channel per-brand per-customer per-canonical-per-customer-cohort-assignment-spec (per-location-cohort + per-channel-cohort + per-brand-cohort + per-vertical-cohort + per-tenure-cohort + per-RFM-cohort + per-seasonality-cohort + per-loyalty-tier-cohort + per-subscription-status-cohort + per-acquisition-source-cohort + per-first-order-SKU-cohort per-canonical-cohort) + per-canonical-per-cohort-CLV-model-ensemble-spec (per-BG-NBD + per-Pareto-NBD + per-Gamma-Gamma + per-MBG-NBD + per-CDNOW + per-Lifetimes-Python-library + per-deep-learning-LSTM + per-deep-learning-Transformer + per-deep-learning-Convolutional-Sequence-to-Sequence per-canonical-model) + per-canonical-per-customer-embedding-spec + per-canonical-per-customer-RFM-spec + per-canonical-per-cohort-recalibration-spec + per-canonical-per-customer-runtime-lookup-spec + per-canonical-per-customer-CLV-confidence-interval-spec + per-canonical-per-customer-CLV-explainability-spec + per-canonical-per-customer-CLV-cross-skill-handoff + per-canonical-per-customer-CLV-FBC-feedback-loop + per-canonical-per-customer-CLV-audit-trail + per-portfolio audit-trail.

Why does per-vendor-Klaviyo-Predictive-CLV-canonical-account-level-CLV break at multi-location multi-channel multi-brand franchise scale?

Per-vendor-Klaviyo-Predictive-CLV-canonical-account-level-CLV ships per-account per-customer per-90-day-prediction primitive. Per-vendor-Lifetimes-Python-library + per-Theta-Equity + per-Retention-Science + per-Optimove + per-Custora-Optimove + per-Glew + per-Polar-Analytics + per-Tresl + per-Daasity + per-Twilio-Engage + per-Segment-Customer-Profiles + per-Amplitude-Audiences + per-Hightouch-canonical-account-level-CLV ship per-vendor per-native account-level CLV primitives. At 1-account-1-channel-1-brand scale per-account per-customer per-90-day-prediction primitive is enough. At multi-location multi-channel multi-brand franchise scale per-canonical-per-customer-cohort-assignment-spec + per-canonical-per-cohort-CLV-model-ensemble-spec + per-canonical-per-customer-embedding-spec + per-canonical-per-customer-RFM-spec + per-canonical-per-cohort-recalibration-spec + per-canonical-per-customer-runtime-lookup-spec + per-canonical-per-customer-CLV-confidence-interval-spec + per-canonical-per-customer-CLV-explainability-spec + per-canonical-per-customer-CLV-cross-skill-handoff + per-canonical-per-customer-CLV-FBC-feedback-loop + per-canonical-per-customer-CLV-audit-trail.

How does per-portfolio per-canonical-per-customer-cohort-assignment + per-cohort-CLV-model-ensemble work?

Per-portfolio per-canonical-per-customer-cohort-assignment-spec runs per-portfolio per-canonical-per-customer-location-cohort-spec + per-canonical-per-customer-channel-cohort-spec + per-canonical-per-customer-brand-cohort-spec + per-canonical-per-customer-vertical-cohort-spec + per-canonical-per-customer-tenure-cohort-spec (per-0-30-day + per-30-90-day + per-90-180-day + per-180-365-day + per-365-day-plus per-canonical-tenure) + per-canonical-per-customer-RFM-cohort-spec + per-canonical-per-customer-seasonality-cohort-spec + per-canonical-per-customer-loyalty-tier-cohort-spec + per-canonical-per-customer-subscription-status-cohort-spec + per-canonical-per-customer-acquisition-source-cohort-spec + per-canonical-per-customer-first-order-SKU-cohort-spec + per-canonical-per-customer-cohort-multi-LLM-classification + per-canonical-per-customer-cohort-confidence-spec + per-canonical-per-customer-cohort-membership-versioning + per-canonical-per-customer-cohort-membership-handoff-to-behavioral-cohort-computation-skill (sibling at /behavioral-cohort-computation). Per-canonical-per-cohort-CLV-model-ensemble-spec runs per-portfolio per-canonical-per-BG-NBD-Beta-Geometric-NBD-spec + per-canonical-per-Pareto-NBD-spec + per-canonical-per-Gamma-Gamma-monetary-value-spec + per-canonical-per-MBG-NBD-Modified-Beta-Geometric-NBD-spec + per-canonical-per-CDNOW-Schmittlein-Morrison-Colombo-spec + per-canonical-per-Lifetimes-Python-library-spec + per-canonical-per-deep-learning-LSTM-spec + per-canonical-per-deep-learning-Transformer-spec + per-canonical-per-deep-learning-Convolutional-Sequence-to-Sequence-spec + per-canonical-per-CLV-model-stacking-blending-spec + per-canonical-per-CLV-model-cross-validation-spec + per-canonical-per-CLV-model-hyperparameter-tuning-spec + per-canonical-per-CLV-model-A/B-test-spec + per-canonical-per-CLV-model-MLflow-Weights-Biases-CometML-experiment-tracking-spec.

What does per-portfolio per-canonical-per-customer-embedding + per-RFM + per-cohort-recalibration do?

Per-portfolio per-canonical-per-customer-embedding-spec runs per-portfolio per-canonical-per-customer-purchase-history-embedding-spec + per-canonical-per-customer-clickstream-embedding-spec + per-canonical-per-customer-engagement-embedding-spec + per-canonical-per-customer-SKU-embedding-spec + per-canonical-per-customer-embedding-store-spec (per-Pinecone + per-Weaviate + per-Qdrant + per-Milvus + per-Chroma + per-pgvector + per-Vertex-AI-Vector-Search per-canonical-vector-store) + per-canonical-per-customer-embedding-drift-monitoring-spec. Per-canonical-per-customer-RFM-spec runs per-portfolio per-canonical-per-customer-recency-spec + per-canonical-per-customer-frequency-spec + per-canonical-per-customer-monetary-value-spec + per-canonical-per-customer-RFM-quintile-spec + per-canonical-per-customer-RFM-percentile-spec + per-canonical-per-customer-RFM-percentile-cross-cohort-spec. Per-canonical-per-cohort-recalibration-spec runs per-portfolio per-canonical-per-cohort-Platt-scaling-spec + per-canonical-per-cohort-isotonic-regression-spec + per-canonical-per-cohort-temperature-scaling-spec + per-canonical-per-cohort-Bayesian-binning-spec + per-canonical-per-cohort-expected-calibration-error-spec + per-canonical-per-cohort-reliability-diagram-spec + per-canonical-per-cohort-Brier-score-spec + per-canonical-per-cohort-recalibration-cadence-spec (per-daily + per-weekly + per-monthly per-canonical-cadence) + per-canonical-per-cohort-recalibration-FBC-pattern-learning.

What does per-portfolio per-canonical-per-customer-runtime-lookup + per-confidence-interval + per-explainability do?

Per-portfolio per-canonical-per-customer-runtime-lookup-spec runs per-portfolio per-canonical-per-customer-CLV-cache-spec (per-Redis + per-DynamoDB + per-Cassandra + per-ScyllaDB + per-Aerospike per-canonical-cache) + per-canonical-per-customer-CLV-cache-TTL-spec + per-canonical-per-customer-CLV-cache-warm-spec + per-canonical-per-customer-CLV-cache-invalidation-spec + per-canonical-per-customer-CLV-real-time-recompute-spec + per-canonical-per-customer-CLV-batch-recompute-spec + per-canonical-per-customer-CLV-streaming-recompute-spec (per-Apache-Flink + per-Apache-Spark-Streaming + per-Materialize + per-RisingWave per-canonical-stream) + per-canonical-per-customer-CLV-runtime-latency-p50-p95-p99-spec + per-canonical-per-customer-CLV-runtime-FBC-pattern-learning. Per-canonical-per-customer-CLV-confidence-interval-spec runs per-portfolio per-canonical-per-customer-CLV-Bayesian-posterior-spec + per-canonical-per-customer-CLV-frequentist-confidence-spec + per-canonical-per-customer-CLV-bootstrap-confidence-spec + per-canonical-per-customer-CLV-confidence-tier-spec (per-high-confidence + per-medium-confidence + per-low-confidence + per-no-confidence per-canonical-confidence-tier). Per-canonical-per-customer-CLV-explainability-spec runs per-portfolio per-canonical-per-customer-CLV-SHAP-spec + per-canonical-per-customer-CLV-LIME-spec + per-canonical-per-customer-CLV-feature-importance-spec + per-canonical-per-customer-CLV-counterfactual-explanation-spec + per-canonical-per-customer-CLV-attribution-graph-spec + per-canonical-per-customer-CLV-multi-LLM-narrative-explanation + per-canonical-per-customer-CLV-chain-of-thought-extraction.

What does per-portfolio per-canonical-per-customer-CLV-cross-skill-handoff + per-FBC-feedback + per-audit-trail + per-customer-data-graph-agent-canonical-bundle do?

Per-portfolio per-canonical-per-customer-CLV-cross-skill-handoff runs per-portfolio per-canonical-per-customer-CLV-handoff-to-per-member-monthly-clv (sibling at /per-member-monthly-clv) + per-canonical-per-customer-CLV-handoff-to-per-member-next-best-action (sibling at /per-member-next-best-action) + per-canonical-per-customer-CLV-handoff-to-churn-prediction-per-subscriber (sibling build-pillar at /how-to-build-per-location-signal-aware-churn-model) + per-canonical-per-customer-CLV-handoff-to-tier-transition-timing (sibling at /tier-transition-timing) + per-canonical-per-customer-CLV-handoff-to-save-flow-propensity-scoring (sibling at /save-flow-propensity-scoring) + per-canonical-per-customer-CLV-handoff-to-cross-location-offer-coordination (sibling build-pillar at /how-to-build-cross-location-offer-coordination-for-multi-banner-franchise-loyalty) + per-canonical-per-customer-CLV-handoff-to-foot-traffic-integration (sibling build-pillar at /how-to-build-foot-traffic-integration-for-multi-location-attribution) + per-canonical-per-customer-CLV-handoff-to-cross-touchpoint-identity-resolution (sibling build-pillar at /how-to-build-cross-touchpoint-identity-resolution-at-multi-location-scale) + per-canonical-per-customer-CLV-handoff-to-attribution-event-emission (sibling at /attribution-event-emission). Per-canonical-per-customer-CLV-FBC-feedback-loop runs per-portfolio per-canonical-FBC-per-customer-CLV-realized-vs-predicted-feedback + per-canonical-FBC-per-customer-CLV-model-recalibration + per-canonical-FBC-per-customer-CLV-cohort-reassignment + per-canonical-FBC-per-customer-CLV-embedding-recalibration + per-canonical-FBC-per-customer-CLV-feature-importance-recalibration + per-canonical-FBC-per-customer-CLV-A/B-test-arm-promotion-demotion. Per-canonical-per-customer-CLV-audit-trail runs per-portfolio per-canonical-per-customer-CLV-canonical-audit-record (per-customer-ID + per-location-pointer + per-channel-pointer + per-brand-pointer + per-cohort-assignment-record + per-model-ensemble-decision + per-CLV-score + per-confidence-interval + per-confidence-tier + per-embedding-pointer + per-RFM-record + per-recalibration-record + per-runtime-cache-pointer + per-explainability-record + per-cross-skill-handoff-record + per-FBC-feedback per-canonical-audit-record) + per-canonical-CCPA-CPRA-DSAR-export + per-canonical-GDPR-DPIA-export + per-canonical-PIPEDA-OPC-export + per-canonical-LGPD-ANPD-export + per-canonical-EU-AI-Act-Article-13-transparency-disclosure + per-canonical-EU-AI-Act-Article-14-human-oversight + per-canonical-NIST-AI-RMF-attestation + per-canonical-immutable-WORM-storage. Per-customer-data-graph-agent-canonical-bundle integrates the ltv-math-primitives skill with sibling skills on the same agent: per-canonical-ltv-math-primitives (parent commercial pillar at /ltv-math-primitives) + per-canonical-versioned-customer-history (sibling build-pillar at /how-to-build-versioned-customer-history-for-multi-location-dsar-compliance) + per-canonical-cross-touchpoint-identity-resolution (sibling build-pillar at /how-to-build-cross-touchpoint-identity-resolution-at-multi-location-scale) + per-canonical-deterministic-probabilistic-hybrid-identity-resolution (sibling build-pillar at /how-to-build-deterministic-probabilistic-hybrid-identity-resolution) + per-canonical-behavioral-signal-ingestion-customer-graph (sibling build-pillar at /how-to-build-behavioral-signal-ingestion-customer-graph) + per-canonical-per-field-conflict-resolution-policy (sibling build-pillar at /how-to-build-per-field-conflict-resolution-policy) + per-canonical-customer-data-orchestration (sibling commercial pillar at /customer-data-orchestration) + per-canonical-identity-resolution-software (sibling commercial pillar at /identity-resolution-software) + per-canonical-master-record-sync (sibling commercial pillar at /master-record-sync) + per-canonical-dsar-software (sibling commercial pillar at /dsar-software) + per-canonical-customer-journey-tracking (sibling commercial pillar at /customer-journey-tracking).

Engage the customer-data-graph agent

Per-portfolio per-location per-channel per-brand per-customer per-canonical-per-customer-cohort-assignment-spec + per-per-cohort-CLV-model-ensemble-spec + per-per-customer-embedding-spec + per-per-customer-RFM-spec + per-per-cohort-recalibration-spec + per-per-customer-runtime-lookup-spec + per-per-customer-CLV-confidence-interval-spec + per-per-customer-CLV-explainability-spec + per-per-customer-CLV-cross-skill-handoff + per-per-customer-CLV-FBC-feedback-loop + per-per-customer-CLV-audit-trail + per-portfolio audit-trail shipped as the orchestration layer above your existing per-CLV-prediction-vendor + per-vector-store-vendor + per-runtime-cache-vendor + per-stream-processing-vendor primitive.