The Business

An independent homeware and gift retailer operating four stores across Greater Manchester, plus an online shop. 22 employees. Annual turnover around £1.8 million. Approximately 3,500 active SKUs across stores, with different product mixes in each location.

The Problem

Inventory management was killing their margins. The owner and one part-time buyer managed stock across all locations using spreadsheets and gut instinct. The results were predictable:

• Overstock — seasonal items unsold, tying up cash in dead stock. End-of-season markdowns were eating 8% of gross margin.
• Stockouts — popular items running out, especially during peak periods. Estimated £2,200/month in lost sales from out-of-stock situations.
• Manual reordering — 12-15 hours per week spent checking stock levels, generating purchase orders, and coordinating between stores.
• No demand visibility — ordering was reactive, not predictive. They knew what sold last month, but not what would sell next month.

What We Built

Phase 1: Data Unification (Week 1)

Connected their EPOS system, e-commerce platform, and supplier spreadsheets into a single data pipeline. Cleaned 18 months of historical sales data and standardised product categorisation across all stores.

Phase 2: Demand Forecasting (Week 2)

Built an AI forecasting model trained on their historical sales data, accounting for:

• Seasonal patterns (Christmas, Valentine's, Mother's Day, etc.)
• Day-of-week and payday effects
• Weather impact on footfall
• Local events (football matches, festivals) that affected specific stores
• Product lifecycle patterns (new product uptake curves, long-tail decline)

Phase 3: Automated Reordering (Week 3)

Created an intelligent reorder system that:

• Generates purchase orders automatically based on forecast demand and current stock
• Accounts for supplier lead times and minimum order quantities
• Suggests inter-store transfers to balance stock before reordering
• Flags slow-moving inventory for markdown decisions

Phase 4: Dashboard and Alerts (Week 4)

Built a simple dashboard the owner checks daily — stock health by store, upcoming reorder recommendations, and margin alerts. Weekly AI-generated reports highlighting trends and anomalies.

Results After 3 Months

• 32% reduction in overstock value — £18,000 less cash tied up in dead inventory
• 45% fewer stockout incidents — capturing an estimated £1,400/month in previously lost sales
• 12 hours per week saved on manual inventory management
• End-of-season markdown reduced from 8% to 4.5% of gross margin
• Forecast accuracy of 87% on weekly demand predictions

ROI

Implementation cost: £4,200. Monthly platform and maintenance: £380.

Monthly benefit: approximately £3,100 (recovered sales + margin improvement + time savings valued at staff cost).

Payback period: 6 weeks.

What Made It Work

• Started with data. The forecasting model was only as good as the data feeding it. Spending a full week on data quality paid off.
• Human review on purchases. The system recommends, the buyer approves. Trust builds over time as predictions prove accurate.
• Store-specific models. Each location has different customers and patterns. A city-centre store behaves differently from a retail park location.
• Phased approach. The owner saw value from the data unification alone before the AI forecasting even kicked in.

Related: Learn more about our AI data analysis and process automation services. Read about measuring ROI on AI integration, or see other case studies: professional services, manufacturing, hospitality.