Role — Data Detective

Data Detectives translate a retailer's heterogeneous data — POS exports, spreadsheets, system inventories, interview notes, workshop transcripts, invoice archives — into GrowDirect's canonical retail data model. They are the front line of a CATz engagement.

The mandate

Every retailer has data. Most of it doesn't speak a common language. Some of it is in a POS system. Some is in spreadsheets an accounts clerk maintains. Some is in a back-office legacy system nobody has logged into in three years. Some is in the founder's head.

Data Detectives assume nothing, find everything, and structure what they find against the canonical model so the rest of the engagement — diagnostic, option modeling, implementation planning — runs on ground truth.

What a Data Detective delivers

1. Data map. Every source system, spreadsheet, and document cataloged. Format, custodian, refresh cadence, known gaps.
2. Source-to-canonical mapping. For each source, how its entities translate to the canonical retail data model (People × Places × Things × Events × Workflows). Explicit mappings, transformations, defaults for missing values.
3. Evidence pack. Structured JSON inputs for the retail-diagnostic and it-architecture-options skills. Every number in the skill output is traceable to a source.
4. Gap register. Known data gaps with severity (blocks analysis / degrades analysis / nice-to-have) and proposed remediation.

How the role is trained

Data Detectives are agents (or human-agent pairs) trained through:

1. Knowledge cards in the Brain vault — reusable patterns for recognizing source shapes (flat-file POS export vs. database extract vs. receipt text), common entity confusions (item-id vs. SKU vs. UPC), and canonical-model traps (store ≠ location if a store has multiple fulfilment points).
2. Memory chunking against the canonical model — every new retailer's data shape becomes a delta against the model, stored as a reusable translation rule.
3. Progressive exposure to real client evidence packs, with each pack adding new shapes to the translation corpus.

Over time, a Data Detective's capability isn't in knowing one retailer; it's in pattern-matching across many retailers and picking the right translation on first encounter.

Relationship to other roles

• Digital Plumbers (method/roles/digital-plumber) receive the source-to-canonical mappings and wire the runtime integration. Detectives tell them what translation to build; plumbers build it.
• Architects consume the data map during Phase II option modeling — it drives which systems are candidates for replacement vs. integration.
• Engineers consume the evidence pack during the diagnostic skill run.
• Writers reference the gap register in the Phase I diagnostic deliverable.

Anti-patterns

• Assuming the POS is the source of truth. It usually isn't. Receipts, email, spreadsheets, and back-office PDFs carry more operational truth than the POS for most SMB retailers.
• Accepting a source without verifying it. "We have ten years of data in this system" usually means "the system has been running for ten years." Whether the data is consistent, complete, or even read by anyone is a separate question.
• Skipping the gap register. A diagnostic built on incomplete data without an acknowledged gap register overclaims. The gap register is what makes the rest of the engagement defensible.

Related

• method/roles/digital-plumber
• method/retail-diagnostic
• cbm-v2/agent-strategy