Canary LP · Technical Library

The CRDM

Canonical Reference Data Model

"Our soil is our data model, built at enterprise scale, time-tested, hundreds of billions of rows processed, billions saved."

The CRDM — Canary Retail Data Model

The CRDM is not invented. It descends from a proven enterprise loss prevention data architecture that processed hundreds of billions of rows across the world's largest retailers. Three generations of the same canonical schema pattern, refined across decades of real-world deployment.

Provenance

GenerationSystemDeployed AtEra
TDS (Till Data Store)Bidirectional data store for register configuration, items, price files, and enterprise-to-store metadataTesco worldwide2000s
SMART SystemCanonical abstraction layer for aggressive retail acquisition and consolidation across international marketsWalmart International (Central/South America)2010s
CRDMThird-generation canonical schema targeting Square SMB merchantsCanary LP2026

The orchestration layer that fed TDS was RTI (Real Time Integrator) — a lightweight Windows service agent for real-time data movement deployed across Tesco worldwide, Ross Stores, and Delta Airlines. RTI is the direct ancestor of Canary's webhook listeners and polling workers.

Seven Canonical Data Sources

These are the seven data domains that feed any retail LP platform. They were defined in an enterprise LP data specification that the founder reviewed and signed off on. Every new POS integration must map to these same seven sources.

#SourceWhat It IsKey LP Signals
1Transaction HeaderOne record per POS transaction. The backbone.Voids, cancels, transaction velocity
2Tender / PaymentHow the customer paid — cash, card, crypto, gift cardSplit tenders, unusual payment mixes
3Line ItemsEvery item in every basketSweethearting, selective scanning, quantity manipulation
4Employee & TimecardsWho was on the clock, what they touchedAfter-hours activity, refund velocity by employee
5Product / ArticleItem master — what's in the storeInventory ghosts, high-shrinkage items
6Store / LocationWhere transactions happenCross-location patterns, geographic fraud clusters
7CustomerWho's buying — loyalty, returns historyReturn fraud rings, organized retail crime

Coverage Analysis

The CRDM maps every enterprise ancestor field to a Canary column with explicit gap analysis. Current Square API coverage:

SourceFields MappedCoverage
Transaction Header18 of 2864%
Tender / Payment10 of 1377%
Line Items14 of 2167%
Employee + Timecards9 of 10 + 4 new90%+
Product / Article7 of 1164%
Store / Location12 of 1771%
Customer3 of 743%

The gap analysis guides the roadmap: each missing field is a specific API call or webhook expansion.

The Evidence Chain

The CRDM enforces three tiers of data integrity:

Tier 1: Financial Ledger (Append-Only)

Transactions, refund links, cash drawer events, gift card activities. No UPDATE. No DELETE. Financial data is a ledger — you don't erase entries.

Tier 2: Evidentiary (Insert-Only)

Case evidence, evidence access logs, case timelines. Enforced by database triggers. If the platform accuses someone, the evidence chain must be unbroken.

Tier 3: Audit Trail (Hash-Chained)

Every audit log entry includes the SHA-256 hash of the previous entry. Tamper with one record, the chain breaks downstream. The same principle that secures Bitcoin secures the Canary audit trail.

Chirp Detection Registry

The CRDM powers the Chirp detection engine — 22 rules that watch for loss patterns across the seven canonical data sources:

CategoryExample Chirps
Refund anomaliesExcessive refunds, refund-without-sale, refund velocity spikes
Void patternsPost-sale voids, void-after-cash, high void ratios
Cash handlingCash drawer opened without sale, excessive no-sales
Employee behaviorAfter-hours transactions, discount abuse, timecard anomalies
InventoryGhost items, receiving discrepancies, high-shrinkage SKUs
PaymentSplit tender manipulation, gift card laundering

Every Chirp rule traces to specific CRDM fields. When a new POS integration maps to the canonical schema, all 22 detection rules work automatically.

Why It Matters

The data model is the soil. The detection engine, analytics, and Bitcoin inscription all grow from it.

The founder built this schema once for the largest retailers on earth. This is the third generation — refined, battle-tested, and now Bitcoin-native.

GrowDirect Confidential Patent Pending — Provisional 63/991,596