◇Hover any dotted term or ⓘ badge — every claim is hover-explainable.

GENPACT · Retail Hub

OVERVIEW ◆ SHOPPER DNA MERCHANDISING MARGIN ADOPTION ◆ AGENT WORKFLOWS

PRIMARK · FASHION DATA: ERP · BLUE YONDER · POS · LOYALTY Data sources (mocked) The hub pulls from your ERP (inventory, costs, vendors), Blue Yonder (demand forecast + replenishment), POS (every receipt in every store), and Loyalty (the named shoppers + their app behavior). In production we sync nightly + stream POS in real time. Synthetic in this prototype — connectors are real in production builds. ● LIVE

SENIOR MERCHANT · CLUSTER SE-04 · WEEK 21

Good morning, Sarah.
3 things to do today. £1.8M on the table.

Sales lift · last 8 weeks

i Sales lift 8-week revenue change for this store cluster vs the baseline period (matched seasonality). Measured against control stores not on Shopper DNA recommendations. Industry benchmark: +5-7% from cohort-driven assortment programs. Yours is at +5.2%.

+5.2%

▲ vs cluster baseline

Margin we could capture

i Margin opportunity Sum of unrealized margin from: (1) repricing actions not yet taken, (2) assortment swaps the team hasn't approved, (3) vendor renegotiations flagged. Refreshed nightly. "Captured" = approved + executed. Cuts the BS of vanity dashboards.

£2.1M

▲ £340K this week

Cost alerts on raw materials

i Commodity / raw material alerts Active changes in commodity prices (cotton, polyester, wheat, etc.) that affect your specific SKUs. Each alert comes with vendor exposure + a draft negotiation talking point. Industry term: "commodity opportunity tracking." Covers 20,000+ raw materials globally.

▲ 2 new since Friday

£ from selling shopper data to brands

i Retail Media Network revenue Brands pay Primark to advertise to specific shopper groups — using Primark's own first-party data, not Meta's or Google's. Margin-rich revenue stream that didn't exist five years ago. Retail Media Network revenue: brands pay retailers for access to first-party shopper audiences.

£840K

this quarter

FOUR JOBS · ONE HUBWHAT MERCHANDISERS ACTUALLY DO ALL DAY

i Merchandising Inside this domain: Assortment Optimization (which SKUs per store), Customer Decision Tree (how shoppers actually choose), Item Attribution (consistent product tagging at scale), and Store Cluster Fit. The four most-used surfaces in Sarah's day. Industry-standard merchandiser toolkit, all cohort-aware.

Merchandising

"What do I stock, where, and why?"

3 actions · 12 SKUs flagged

i Pricing & Value Price elasticity simulator (slider: change price → see sales/margin impact), competitor benchmark, markdown cadence (when to start a sale on each SKU), and AI-generated vendor negotiation packs. Standard industry naming: "pricing & value" — covers both revenue and customer-perceived value.

Pricing

"What do I charge, and when do I discount?"

£840K opportunity

i Cost Optimization True Dead Net Profit (the real margin after every rebate, return, shipping cost — most retailers can't see this), Commodity tracker, Vendor scorecard, and GNFR spend (goods not for resale — the hidden cost line). DNP visibility unlocks $100M+ margin opportunities for most retailers.

Cost

"What does this actually cost me — and which vendors are gaming me?"

7 raw-material alerts

i Change & Adoption Five-pillar readiness dashboard (Vision · Stakeholder · Enablement · Governance · Sustainability). Token ROI and Inference TCO widgets. The 70% in the 10/20/70 model — most AI programs under-invest here, and that's where transformations quietly die. Change management rendered as a live measured screen, not a slide.

Adoption

"Is my team actually using this — or is it shelfware?"

70% of value lives here

▲ EVERYTHING ABOVE RUNS ON

i Shopper DNA A single customer graph fed by loyalty, POS, browse, and returns. Powers assortment decisions, dynamic newsletters, and 1P audience exports for ads. The merchant team and the marketing team query the same source of truth — no more parallel datasets in different tools. Value-centric operating model: data layer feeds every domain.

◆ FOUNDATION · SHOPPER DNA

4.2M shoppers.
We know who they are, what they bought, what they'll buy next.

Loyalty + receipts + browse + returns, fed into one customer graph. Every domain above pulls from this. Every ad campaign pulls from this. Every newsletter pulls from this.

Shopper groups we've spotted Cohorts (industry term) Groups of shoppers behaving the same way — life-stage ("parents of 6-month-olds"), seasonal ("school-uniform buyers in August"), or behavioral ("returns everything but the cheapest item"). Refreshed nightly from loyalty + POS.

e.g. "parents of 6-month-olds", "school-uniform Q2"

What they'll buy next Lifecycle predictions For each cohort: the 3 next-likeliest purchases in the next 4–12 weeks. Powers store assortment, newsletter content, and ad targeting. Example: "infant 3mo" cohort → 78% probability of buying 6mo within 90 days.

12 predictions

per group, refreshed nightly

Stores getting a layout fix Store-cluster assortment shifts When the cohort mix of a store changes (e.g., young parents moving in to a neighborhood), the recommended SKU mix changes. The hub flags it; the merchandiser approves. Beats waiting for the annual planogram cycle.

because the shoppers there changed

Lists sent to Meta · Google · TikTok First-party (1P) audiences Lists of named shoppers (hashed, GDPR-safe) exported to ad platforms for targeting. Cookieless world made these gold. Same lists power your Retail Media Network revenue line above — brands pay Primark to reach these audiences. "1P" = first-party. Industry shorthand in adtech since ~2022.

first-party audiences, exportable now

THE CLOSED LOOP — HOVER EACH STEP

Who bought what POS + Loyalty signal Receipts (every line item, every store, every minute) joined to the loyalty identifier. The raw input. Streaming from in-store POS in production builds. → Who are they (groups) Cohort engine ML clustering on purchase patterns + lifecycle signals + return behavior. Outputs named, durable shopper groups the merchandiser can reason about. → What to stock next Assortment recommendation Each store's recommended SKU mix recalculated based on which cohorts shop there. Surfaced in the Merchandising domain. Approved or rejected by the merchandiser — never auto-applied. → What to show them Dynamic content Newsletters, app push, in-app banners — different content per cohort. Generated from a single template + cohort-aware variables. CMO team's work, but powered by the merchandiser's data. → Lists for ad targeting 1P audience export Hashed shopper IDs sent to Meta · Google · TikTok · CDP for paid ad targeting. Cookieless-safe. Refreshes nightly. → £ from brands paying to reach them Retail Media Network revenue Brands pay Primark for the privilege of advertising to Primark's own audiences. Margin-rich. The KPI strip's £840K comes from here. ↻

i Agentic Workflow Builder Drag-and-drop interface to wire "if this happens, do that" rules across the hub. Genpact's tangible take on agentic AI — agents that act on merchandiser-defined triggers, not chatbots that just answer questions. Agentic AI made operable: agents that act on triggers, not chatbots that just answer questions.

Agentic Workflow Builder →

Wire up "if this happens, do that" automatically. No more rule-running in spreadsheets.

IF 6-month baby group grows 8% in a week THEN propose 9-month clothes · draft a newsletter · send list to Meta