Retail Data Infrastructure Audit

Purpose

This recipe executes a structured audit of retail data infrastructure across 7 sub-dimensions: demand signal inventory, POS-to-analytics pipeline latency, supply chain data integration, product knowledge graph maturity, AI retrieval readiness (GEO), real-time vs batch decision-making ratio, and overall maturity benchmarking. It produces a scored data infrastructure report that feeds Dimension 1 of the Retail AI Diagnostic Engagement. [src1, src2]

Prerequisites

• POS system access — read-only API credentials or trailing 90-day transaction data export
• Product catalog feed — structured product data (API endpoint, XML feed, or CSV export)
• Technology stack inventory — POS, ERP, WMS, CRM, e-commerce, analytics, CDP
• Supply chain system list — EDI partners, supplier portals, logistics platforms
• Data processing agreement signed
• IT point of contact identified for access questions

Constraints

• POS system access must be read-only — never write to production POS databases. [src1]
• Customer PII must be anonymized before analysis — hash customer IDs, strip names, addresses, payment details.
• Latency measurements require minimum 7-day observation window to capture weekly seasonality. [src4]
• Knowledge graph scoring requires product catalog API or structured feed — cannot be scored from public website alone. [src3]
• GEO audit results are model-version-dependent — document AI model versions and test date. [src5]

Tool Selection Decision

Which path?
├── Retailer has modern cloud POS with API (Shopify POS, Square, Lightspeed)
│   └── PATH A: API-Direct — real-time latency measurement, automated profiling
├── Retailer has legacy POS (Oracle MICROS, NCR Voyix, Toshiba)
│   └── PATH B: Export-Based — CSV/flat file analysis, manual latency estimation
├── Retailer has multiple POS systems (acquisitions, regional variance)
│   └── PATH C: Multi-Source — audit each system, reconciliation analysis
└── Retailer has only dashboard access (no raw data)
    └── PATH D: Interview-Based — structured questionnaire, proxy metrics

Execution Flow

Step 1: Demand Signal Source Inventory

Duration: 0.5-1 day · Tool: Spreadsheet + IT stakeholder interview

Inventory every data source that produces demand signals. Catalog each with source name, signal type, refresh rate, data format, current usage, and integration status. Target: 15-30 sources for a mid-size retailer. [src2]

Verify: Inventory complete with 15+ sources, each with refresh rate and integration status. · If failed: Supplement with stakeholder interviews and public documentation.

Step 2: POS-to-Analytics Pipeline Latency Measurement

Duration: 1-2 days (plus 7-day observation) · Tool: POS API or export + timestamp analysis

Measure end-to-end latency from POS transaction to analytics availability. For API-accessible POS: record test transactions, measure API and analytics latency across 7 days. For export-based: document export schedule and ETL processing time. Benchmark: Level 1 (> 24h) through Level 5 (< 1 min). [src4]

Verify: Latency measured or estimated, maturity level assigned with evidence. · If failed: Use export schedules and ETL logs to estimate, document as “estimated.”

Step 3: Supply Chain Data Integration Assessment

Duration: 0.5-1 day · Tool: System inventory + integration mapping

Map data flows across 6 supply chain domains: inventory positions, supplier data, logistics/shipping, warehouse operations, returns/reverse logistics, and demand planning signals. Score each for completeness, integration, and timeliness (1-5).

Verify: All 6 domains mapped with scores. · If failed: Use IT inventory and interviews to estimate.

Step 4: Product Knowledge Graph Maturity Assessment

Duration: 0.5-1 day · Tool: Product catalog analysis + knowledge graph tools

Evaluate 5 knowledge graph dimensions: taxonomy depth, attribute coverage, relationship mapping, semantic richness, and entity resolution. Benchmark from Level 1 (flat catalog) to Level 5 (full knowledge graph with embeddings). [src3]

Verify: KG maturity level assigned (1-5) with evidence per dimension. · If failed: Audit public website, note reduced accuracy.

Step 5: AI Retrieval Readiness (GEO Audit)

Duration: 0.5-1 day · Tool: ChatGPT, Perplexity, Google AI Overviews, Bing Copilot

Test 20-30 queries across product discovery, brand/store, and comparison categories. Record mention presence, position, data accuracy, and source attribution per query. Calculate GEO visibility score. [src5]

Verify: GEO audit spreadsheet complete with 20+ queries and visibility score. · If failed: Use Google AI Overviews as primary source.

Step 6: Real-Time vs Batch Decision-Making Ratio

Duration: 0.5 day · Tool: Decision inventory + system analysis

Classify 10 key retail decisions (pricing, replenishment, markdown, scheduling, targeting, fraud, routing, assortment, rerouting, personalization) as real-time or batch. Calculate real-time ratio and assign maturity level. [src4]

Verify: All 10 decisions classified with evidence, ratio calculated. · If failed: Classify from stakeholder interviews and system capabilities.

Step 7: Maturity Level Benchmarking and Report Generation

Duration: 0.5-1 day · Tool: Scorecard template + analysis synthesis

Synthesize all 6 steps into a weighted composite Data Infrastructure Maturity Score (1-5). Produce report with per-dimension breakdown, critical gaps, quick win recommendations, and infrastructure investment recommendations. [src2]

Verify: All sub-dimensions scored, composite calculated, report produced. · If failed: Mark incomplete dimensions with confidence flag.

Output Schema

{
  "output_type": "retail_data_infrastructure_scorecard",
  "format": "PDF + JSON",
  "sections": [
    {"name": "composite_maturity_score", "type": "number", "description": "Weighted composite score 1-5"},
    {"name": "sub_dimension_scores", "type": "array", "description": "7 sub-dimension scores with evidence"},
    {"name": "pipeline_latency_report", "type": "object", "description": "Latency per pipeline with bottlenecks"},
    {"name": "knowledge_graph_assessment", "type": "object", "description": "5-dimension KG maturity with coverage %"},
    {"name": "geo_audit_results", "type": "object", "description": "AI visibility score with per-query results"},
    {"name": "gap_analysis", "type": "array", "description": "Ranked gaps by AI readiness impact"},
    {"name": "recommendations", "type": "array", "description": "Quick wins + infrastructure investments"}
  ]
}

Quality Benchmarks

If below minimum: Extend audit by 1-2 days. Prioritize measured over estimated metrics. Request additional system access.

Error Handling

Cost Breakdown

Anti-Patterns

Wrong: Measuring latency once and extrapolating

Single POS latency measurement on a Tuesday afternoon. Result: misses 3x latency spike during Saturday peak and overnight batch delay. [src4]

Correct: Measure across a full weekly cycle

Sample at minimum 4 time points per day across 7 days. Report P50, P95, and P99 latency.

Wrong: Scoring knowledge graph from website alone

Evaluating product data from the public e-commerce site. Result: overestimates maturity — website shows curated data while backend has 40% missing attributes. [src3]

Correct: Audit the source catalog, not the presentation layer

Request raw product feed or catalog API export. Calculate attribute coverage across all SKUs.

Wrong: Ignoring uncollected demand signals

Only inventorying data already in the analytics stack. Result: misses 5-10 high-value signal sources that would unlock the next maturity level. [src2]

Correct: Inventory what should exist, not just what does

Start with a reference list of 25-30 retail demand signal types. Check each one. Uncollected signals often have the highest marginal value.

When This Matters

Use when an agent needs to audit retail data infrastructure as a standalone engagement or as Dimension 1 of the Retail AI Diagnostic Engagement Playbook. Produces a scored, evidence-based assessment of 7 data infrastructure sub-dimensions that determine AI deployment readiness.