ANS-100/P — Retail & Consumer Sector Calibration Note

Drafted 30 August 2026 · Ratification logged; date not separately stated · written before the first retail and consumer company was scored · Rater 1 · Companion to ANS-100/P v2.1 (frozen anchor set) · Sector: Titan, Trent, Avenue Supermarts (DMart), ITC, Asian Paints, Britannia

Ratification log: E11 ratified — Q32 classes N for healthcare; no score changes.

1. Why this sector needs its own note

Retail and consumer sits between two patterns already seen. Like consumer internet, it is customer-facing and generates transaction data at scale; unlike it, these are physical-first businesses where AI lives in demand forecasting, assortment planning, inventory and replenishment, store operations, supply chain and marketing — not in an app that is the product. And like manufacturing, several of these companies (ITC, Asian Paints, Britannia) run large factories, so the D-rulings' hard-won distinctions must carry over intact.

The sector also carries a specific risk the Index has not yet faced: retail is the sector most likely to describe ordinary merchandising analytics as AI. Demand forecasting, planogram optimisation and customer segmentation predate machine learning by decades. The rulings below cut both ways deliberately.

2. Proposed rulings (ratify individually: F1 ☐ … F10 ☐)

F1 — Q25 economic pair. Revenue per employee, read with the sector's own productivity metrics where disclosed: same-store sales growth (SSSG), sales per square foot, inventory turns, gross margin, and shrinkage. A1 unchanged — level 3 needs a disclosed number linking named AI systems to one of these. Named confounds: store-opening cadence, commodity and gold-price movements, festive-season timing, and GST changes.

F2 — Q27, the sector's strongest opportunity. Retail admits: forecast-accuracy improvement attributed to AI, auto-replenishment share, shrinkage reduction, markdown-optimisation rates, and AI-assisted customer-service resolution rates. Excluded (the B2/C2/D3/E2 analogue): e-commerce or omni-channel sales share, app download counts, loyalty-programme membership, and general "digitisation of stores" — these measure channel and scale, not AI.

F3 — "Analytics is not necessarily AI." Demand forecasting, assortment planning, customer segmentation and planogram optimisation score nothing in D2 without evidence of learning systems — models trained on data that improve with use. Statistical forecasting and rules-based replenishment are the retail analogue of PLCs and SCADA. Where a company says only "data-driven" or "analytics-led", that is level 1.

F4 — Q6/Q7 autonomy anchor. Production autonomy means AI systems taking or directly triggering commercial actions: automated replenishment orders, dynamic pricing or markdown execution, autonomous assortment decisions, algorithmic supply-chain allocation. Dashboards and recommendations to merchandisers are level 2 at most. B2-4 applies — predictive systems at disclosed scale in a delivery-critical path may reach level 3 without actuation authority.

F5 — Q33 product-embedded AI. Level 3 requires customer-facing AI shipped with disclosed volumes: virtual try-on with usage figures, AI styling or recommendation engines with attributed sales contribution, conversational commerce at scale. For manufacturers in this group, AI-derived product development or formulation with disclosed output counts as the analogue. Vendor partnership announcements are level 2.

F6 — Q31 exposure re-anchored. Retail's AI exposure is agentic commerce — AI shopping agents transacting on customers' behalf and bypassing brand and store discovery — plus LLM answer-engine erosion of search-led discovery for brands. This is the C4 exposure applied to physical retail, and C4's precedent holds: silence on it is a scored absence. Mitigation = proprietary customer data, private-label economics, agent-ready product feeds and APIs, or an owned demand channel.

F7 — Q18. No sector-specific regulator baseline exists here. DPDP-readiness plus published customer-data-handling specifics = 2; audited or certified AI management systems, or AI surfaces in a public bug-bounty scope = 3. Consumer-data governance disclosures count where they name AI or automated decision-making specifically; generic privacy policies do not.

F8 — Q24. The A4 ≥5%-of-TTM-revenue line applies. Store-expansion capex is never AI investment. Technology or digital capex counts toward level 3 only where an AI-specific component is identifiable, per B8 and D8.

F9 — Entity rulings (verified before ratification, per the D9 lesson). Titan: consolidated, spanning jewellery, watches, eyecare and Taneira; Titan Engineering & Automation is a subsidiary and its automation business must not be read as Titan's own AI adoption. Trent: consolidated, Westside and Zudio, with Trent Hypermarket noted. Avenue Supermarts: DMart plus DMart Ready, with the e-commerce arm's evidence recorded separately. ITC: consolidated — flag for the file that ITC Infotech is a separately-run IT services subsidiary and its evidence must not cross over into ITC's own score, exactly as Tech Mahindra was excluded from M&M and LTM/LTTS from L&T. Asian Paints and Britannia: consolidated.

F10 — Signal library additions. Integrated annual reports, BRSR filings, investor-day decks, supply-chain and technology-partner case studies, app release notes and changelogs where a consumer app exists, retail-technology analyst assessments, and store-format or format-innovation disclosures.

3. Proposed promotion to a universal floor item (ratify separately)

F-U — Full-text annual-report retrieval becomes an A9 floor item for every remaining sector, not just pharma. The healthcare synthesis makes the case: full-text retrieval moved four of four pharma files, and in Cipla's case would have cost roughly twenty points had it been skipped. Search snippets systematically miss AI disclosure buried in operational, quality and technology sections — precisely where physical-first businesses put it. Retail and infrastructure companies have the same document structure. Cost: one retrieval per company. Benefit: the difference between a defensible score and a wrong one. Ratify: F-U ☐

4. What does not change

All v2.1 anchors, P-caps, the evidence floor, the 18-month window, A1–A10, Batch-2 as applied, ceiling arithmetic per Appendix A, the pack-based confidence rule, and the evidence-language rule. Application, not amendment.

5. On ratification

Score in universe order: Titan → Trent → Avenue Supermarts → ITC → Asian Paints → Britannia. Expect the D-sector pattern for the manufacturers and the C-sector pattern for the retailers, with F2 the live test of whether retail discloses AI-attributed operational rates. A Q32 ruling (F11) is anticipated at the sector synthesis.

Published 9 September 2026 under CC BY-SA 4.0 as part of the Edition One replication material. Status changed from "proposed" to "ratified" on publication: the note was drafted on 30 August 2026 and written before the sector was scored; the ratification record for this set is described on the rulings index. No ruling text is changed by this publication.