CS-08
Global Fashion Retailer — AI Personalisation at Scale
A FTSE 100 global fashion retailer operating in 40+ countries with $3.8B in annual revenue was leaving significant growth on the table: 8% email open rates, 1.2% conversion rates, and $120M in annual inventory markdown losses. The retailer's customer data platform was siloed from its commerce engine, making 1:1 personalisation operationally impossible at global scale. Anicalls' AI Personalisation Agent unified the data estate and deployed real-time 1:1 recommendations — delivering 340% email revenue growth and $85M in markdown reduction.
340%Email Revenue Growth
4.8×Conversion Rate Lift
$85MMarkdown Reduction
28%CLV Increase
Business Challenge
Why Generic Campaigns Were Destroying Margin
8% Email Open Rate — Declining
The retailer sent 42M marketing emails weekly — at an 8% open rate and 1.2% click-through rate that was declining quarter-on-quarter. Batch-and-blast campaigns sent the same promotional message to every subscriber regardless of category affinity, purchase history, or channel preference. Unsubscribe rate was 2.4× the industry average, with 6M subscribers lost in 18 months — representing $180M in lifetime value permanently lost.
$120M Annual Markdown Losses
Inventory markdown losses reached $120M annually — driven by poor demand forecasting that failed to match stock allocation to local demand signals. End-of-season clearance campaigns discounted at 50–70% were the only mechanism for clearing inventory, destroying margin. The retailer was simultaneously experiencing stockouts in high-demand sizes and colours while sitting on excess inventory in slow-moving lines — a mismatch costing 8 margin points.
Siloed Data and Technology
The retailer operated 7 separate data systems — Salesforce Commerce Cloud (e-commerce), Salesforce Marketing Cloud (email), SAP (ERP/inventory), a legacy CDP, Bazaarvoice (reviews), Klaviyo (SMS), and a DMP for digital advertising. None were integrated in real time. The data science team spent 80% of their time on data pipeline maintenance rather than model development. Personalisation required 6-week campaign build cycles that were outdated by launch.
40-Country Complexity
Operating across 40+ countries meant 40+ local inventory positions, 40+ localised catalogues with different assortments, 8 currencies, and marketing regulations spanning GDPR, UK PECR, CAN-SPAM, CASL, and PDPA — each requiring different consent capture and data handling approaches. The personalisation team of 12 could not manually manage 40-country campaign variants — defaulting to single global campaigns that performed poorly in every market.
Solution Delivered
Anicalls AI Personalisation Agent — Global Retail Deployment
Recommendations
1:1 Product Recommendation Engine
Deep learning recommendation models process each customer's full purchase history, browsing behaviour, returns data, wishlist signals, and peer affinity patterns to generate truly individualised product recommendations — refreshed in real time with every new interaction. Deployed across email, website homepage, product detail pages, basket, checkout, and post-purchase flows simultaneously, with consistent personalisation across every touchpoint.
- Real-time recommendation refresh (sub-100ms)
- Cross-channel consistency (email + web + app)
- Returns-aware recommendations (size/fit intelligence)
- New visitor cold-start handled by trend/segment models
Email AI
Dynamic Email Personalisation at Scale
Every email sent to 42M subscribers is now individually assembled in real time at the moment of opening — with AI-selected subject line, hero product, secondary recommendations, offer mechanics, and send-time optimisation calibrated to each individual's historical engagement patterns. Subject line AI tests 8 variants per subscriber cohort per week, continuously learning optimal phrasing for each preference segment.
- Real-time email assembly at open-time
- AI send-time optimisation per subscriber
- Dynamic subject line testing (8 variants/cohort)
- GDPR / PECR / CAN-SPAM consent enforcement
Inventory AI
Markdown Optimisation & Demand Intelligence
AI demand forecasting at SKU-store-week level predicts sell-through rates and triggers proactive inventory rebalancing — redistributing stock between warehouses and stores before markdown becomes necessary. When markdown is unavoidable, the AI calculates optimal timing, depth, and targeting — showing markdown products only to customers with high affinity for that category and price point, maximising revenue recovery.
- SKU-store-week demand forecasting
- Proactive stock rebalancing recommendations
- AI-optimised markdown timing and depth
- Affinity-targeted markdown audience selection
AI Workforce Deployment
The Anicalls AI Personalisation Operations Team
AI Campaign Operations Agents
20 AI Campaign Operations Agents manage the 40-country email programme — localising content, enforcing regulatory consent requirements per jurisdiction, optimising send scheduling across time zones, and continuously testing subject lines, hero images, and offer mechanics. What previously required a 12-person campaign team with 6-week build cycles is now executed by AI agents in 48 hours, enabling weekly campaign refreshes per country.
AI Inventory Intelligence Agents
12 AI Inventory Intelligence Agents monitor sell-through velocity at SKU-store-week level across 40 markets — generating daily rebalancing recommendations, early markdown triggers, and demand signals for the buying team's next-season purchasing decisions. Each agent monitors 50,000 SKUs across 8 markets, providing the buying team with actionable intelligence that was previously impossible to generate manually at this granularity.
AI Data Science Agents
6 AI Data Science Agents handle model retraining, A/B test analysis, attribution modelling, and customer lifetime value scoring — running continuously rather than requiring data scientist intervention. The human data science team (now freed from pipeline maintenance) focuses exclusively on new model development, new data source integration, and strategic measurement framework design — their output tripled in quality and volume within 6 months.
Technologies Used
The AI Technology Stack Deployed
Platform
Customer Intelligence Engine™ v4.0
Anicalls' retail personalisation platform — pre-integrated with Salesforce Commerce Cloud, Salesforce Marketing Cloud, SAP S/4HANA, and Klaviyo via pre-built connectors. Unified customer profiles built from all 7 data sources with real-time event streaming. GDPR/PECR consent management native — with jurisdiction-aware consent propagation across all marketing channels simultaneously.
- Salesforce CC + MC + SAP + Klaviyo connectors
- Real-time unified customer profile
- GDPR / PECR / CAN-SPAM / CASL consent engine
- Sub-100ms recommendation API
Recommendation AI
Deep Learning Recommendation Models
Transformer-based sequential recommendation models processing the full customer interaction sequence — purchases, browses, saves, returns, reviews — to predict next-best product with category, colour, and size intelligence. Separate cold-start model for new visitors using trend signals, peer affinity, and geographic demand patterns. Models retrained weekly on new interaction data across 40 markets.
- Transformer sequential recommendation model
- Category + colour + size preference intelligence
- Cold-start model for new visitors
- Weekly retraining on 2B+ interaction events
Forecasting
AI Demand Forecasting & Inventory AI
Hierarchical time-series forecasting models (LightGBM + neural prophet ensemble) operating at SKU-store-week granularity — incorporating seasonality, weather, local events, promotions, and cross-SKU cannibalisation effects. Connected to SAP ERP for real-time inventory position data and automated rebalancing recommendation generation. Integrated with the buying system for next-season open-to-buy optimisation.
- LightGBM + NeuralProphet ensemble forecasting
- SKU-store-week granularity (40 markets)
- SAP ERP real-time inventory integration
- Automated rebalancing recommendation engine
Quantified ROI
The Commercial Impact at 12 Months
340% Email Revenue Growth
Email open rate improved from 8% to 28%; click-through rate from 1.2% to 6.8%; email-attributed revenue grew 340% year-on-year. The channel generated an additional $124M in annual revenue — from the same subscriber base, with 18% lower email volume (more targeted sends reduced unsubscribe rate from 2.4× to 0.6× industry average). Unsubscribe trend reversed: the retailer regained 2.1M previously lapsed subscribers through re-engagement campaigns.
$85M Markdown Reduction
AI demand forecasting reduced end-of-season inventory requiring markdown from $120M to $35M annually — a $85M improvement. Average markdown depth reduced from 52% to 31% (because AI-targeted markdown campaigns reached buyers with genuine affinity, driving higher sell-through at shallower discounts). Gross margin improved by 3.2 percentage points — equivalent to $122M incremental profit on $3.8B revenue.
4.8× Conversion Rate Lift
Website conversion rate improved from 1.2% to 5.8% — driven by AI-personalised homepages and product detail page recommendations showing customers items they are statistically most likely to purchase. Average order value increased 24% (AI cross-sell recommendations surfacing complementary items). Customer lifetime value across the active customer base increased 28% — an estimated $340M improvement in 3-year CLV across the 42M subscriber base.
Business Outcomes
Strategic Outcomes Beyond Revenue
Sustainability
$42M Sustainability Impact
The $85M markdown reduction also delivered significant sustainability benefits — with fewer unsold garments entering the clearance/destruction pipeline. The retailer estimated 4.2M fewer garments required discarding or clearance-sale disposal annually, reducing Scope 3 emissions by an estimated 18,000 tonnes CO₂e. The CFO cited the AI-driven inventory efficiency as a key contribution to the retailer's 2030 sustainability commitments.
Speed
6-Week → 48-Hour Campaign Cycle
The campaign production cycle collapsed from 6 weeks to 48 hours — enabling the marketing team to respond to trend signals, competitor activity, and cultural moments in near-real-time. The team launched 14 reactive campaigns in the first 6 months that would have been impossible under the previous build cycle. One trend-reactive campaign (responding to a viral fashion moment on TikTok within 48 hours) generated $8.2M in 72-hour revenue.
Tech Stack
Legacy Tech Decommissioned
The Anicalls integration layer unified the 7-system data estate — enabling the retailer to decommission the legacy CDP (saving $1.8M in annual licensing fees) and retire the DMP (saving $2.4M). The data science team's infrastructure maintenance burden fell 80%, with 4 of 5 data engineering FTEs redeployed to new AI model development. Technology cost savings contributed $4.2M to the ROI calculation.
Executive Testimonial
"We had tried personalisation three times before. Each attempt failed because the data wasn't unified and the technology couldn't operate at global scale. Anicalls solved both problems simultaneously. Within 6 months, every customer across 40 countries was seeing a uniquely assembled experience — and our email open rates tripled. The $85M markdown reduction was perhaps even more impressive than the revenue growth: it tells us we are finally matching supply to real demand. That's a fundamental transformation of how a global retailer operates."
Metrics Dashboard
12-Month Performance Scorecard
28%Email Open Rate (was 8%)
340%Email Revenue Growth
5.8%Conversion Rate (was 1.2%)
$85MMarkdown Reduction
+24%Average Order Value
+28%Customer Lifetime Value
48hrsCampaign Cycle (was 6 weeks)
$124MIncremental Email Revenue
Personalise Every Customer Interaction — Globally
See how the Anicalls Customer Intelligence Engine™ can transform your email performance, cut your markdown losses, and grow customer lifetime value across every market you operate in.