75% of fashion executives now prioritise AI for inventory optimisation
Merchandising AI investment overtakes marketing spend as retailers seek double-digit forecasting improvements and faster ROI.

KEY TAKEAWAYS Summary by the editors
- Three-quarters of fashion executives now prioritise AI for inventory optimisation over marketing applications.
- Merchandising AI projects deliver ROI in 12 to 18 months compared to 24 months for marketing tools.
- Leading retailers report forecast accuracy improvements of 15 to 25 percentage points after AI implementation.
- AI inventory systems reduce markdowns by 10 to 20 per cent whilst maintaining stock availability.
Three-quarters of fashion executives now rank artificial intelligence for inventory optimisation as their top technology priority, marking a decisive shift away from earlier enthusiasm for marketing and customer-facing applications, according to industry research and company disclosures reviewed in recent months. The pivot reflects mounting pressure to protect margins amid persistent inflation and the superior return on investment that merchandising AI delivers compared to generative tools for content creation.
Why are retailers shifting AI budgets from marketing to merchandising?
The reallocation of technology spending stems from two factors: disappointing returns from generative AI marketing tools and urgent need to address inventory inefficiencies that directly erode profitability. Retailers that rushed to deploy chatbots and AI-generated product descriptions in 2023 and 2024 found limited impact on conversion rates, whilst excess stock and markdowns continued to squeeze operating margins.
Merchandising AI applications target demand forecasting, assortment planning, and automated replenishment systems that determine how much product to buy, where to allocate it across channels, and when to discount slow-moving styles. These decisions account for the largest controllable costs in fashion retail after raw materials and labour. Even modest improvements in forecast accuracy translate to millions in reduced markdowns and lost sales.
What ROI are fashion companies seeing from inventory AI?
Merchandising AI projects typically deliver measurable returns within 12 to 18 months, significantly faster than the 24-month timeframe common for marketing technology investments. The speed advantage comes from direct linkage to financial outcomes: better forecasts immediately reduce excess inventory carrying costs and markdown depth, whilst maintaining in-stock rates on best-selling items.
Major fashion retailers report quantifiable improvements across several metrics after implementing AI-driven inventory systems:
- Forecast accuracy gains of 15 to 25 percentage points compared to traditional statistical methods
- Markdown reductions of 10 to 20 per cent whilst sustaining or improving stock availability
- Inventory holding costs down 8 to 15 per cent through better demand prediction
- Stockout reduction of 20 to 30 per cent on core replenishment items
These improvements compound over successive seasons as machine learning models ingest more transaction data and refine their predictions. Retailers that began deployments in 2024 are now seeing accelerating benefits as algorithms adapt to their specific customer behaviour patterns and product lifecycles.
How much do forecasting errors cost fashion businesses?
Forecasting inaccuracy remains the single largest source of operational waste in fashion retail. Industry estimates suggest that forecast errors cost the sector tens of billions annually through two mechanisms: markdowns on overstock and lost sales when popular items sell out. The problem intensifies for fast fashion and contemporary brands that place non-replenishable orders months before selling seasons begin.
Traditional forecasting methods rely on historical sales patterns, seasonal trends, and merchant intuition. These approaches typically achieve accuracy rates of 60 to 70 per cent at the style and colour level, leaving substantial room for error. When multiplied across thousands of SKUs and hundreds of stores or fulfilment centres, even small percentage improvements in accuracy unlock significant value.
AI-powered systems incorporate broader data sets including weather patterns, social media sentiment, search trends, and real-time competitor pricing alongside historical sales. Machine learning algorithms detect non-linear relationships between these variables that human analysts and simple statistical models miss. The technology proves particularly effective for new product introductions where historical data is limited or non-existent.
Which merchandising functions benefit most from AI?
Demand forecasting attracts the heaviest investment, but retailers are expanding AI applications across the merchandising workflow. Initial purchase quantity decisions for seasonal collections represent the highest-stakes forecasts, as production lead times prevent adjustments once orders are placed with suppliers. AI models help merchants balance the risk of under-buying popular styles against over-committing capital to potential slow sellers.
Allocation and replenishment systems distribute inventory across physical stores, e-commerce warehouses, and regional distribution centres. AI optimisation accounts for local demand variations, transfer costs between locations, and opportunity costs of stockouts. These systems make thousands of micro-decisions weekly that cumulatively determine whether the right product reaches customers when they want to buy.
Markdown optimisation uses AI to determine the timing and depth of price reductions that clear aged inventory whilst preserving maximum margin. Rather than applying blanket percentage discounts across product categories, algorithms identify which items require aggressive promotion and which will sell through at modest reductions. The approach protects brand positioning by minimising the volume of product sold at deep discount.
What obstacles slow AI adoption in merchandising?
Data quality and integration challenges remain the primary barriers to successful AI deployment. Merchandising algorithms require clean, consistent transaction data across all sales channels, accurate product attribute information, and reliable inventory positions. Many retailers still operate fragmented systems where e-commerce, store point-of-sale, and warehouse management platforms do not share common data standards.
Organisational resistance also hinders adoption, particularly from experienced merchants accustomed to trusting their instincts. AI recommendations sometimes conflict with conventional wisdom or contradict seasonal buying patterns that have worked historically. Progressive retailers address this through hybrid approaches where algorithms generate forecasts and suggested orders but merchants retain override authority and gradually build confidence in the technology.
Implementation costs and technical complexity present obstacles for smaller fashion businesses. Enterprise-grade merchandising AI platforms require significant upfront investment in software licensing, data infrastructure, and specialist personnel. Mid-market retailers increasingly access similar capabilities through software-as-a-service offerings that require lower initial capital but still demand internal expertise to configure and interpret model outputs.
Where will merchandising AI investment focus next?
Retailers are expanding AI applications upstream into product design and development decisions. Early-stage demand sensing uses signals from search behaviour, social media conversations, and trend forecasting to identify emerging styles before committing to full collection development. This approach compresses the feedback loop between market signals and merchant action, potentially reducing exposure to trend shifts.
Integration with supplier systems represents another frontier. Fashion companies are beginning to share AI-generated forecasts directly with manufacturing partners, enabling more flexible production planning and shorter lead times. This requires unprecedented transparency and data sharing across traditionally opaque supply chains, but promises to reduce total system inventory whilst improving responsiveness to actual demand.
Sustainability metrics are increasingly embedded in merchandising AI models. Algorithms now factor the environmental cost of overproduction, transportation emissions from inventory transfers, and disposal impacts of unsold goods into optimisation calculations. Retailers face growing regulatory and consumer pressure to reduce waste, making inventory accuracy not just a financial imperative but an environmental one.
Frequently asked questions
How long does it take to implement merchandising AI?
Full implementation typically requires 12 to 18 months from initial data preparation through pilot testing to company-wide deployment. Most retailers see measurable ROI within this timeframe through reduced markdowns and improved stock availability.
Do AI inventory systems replace merchandising staff?
No, AI augments rather than replaces merchant expertise. Algorithms generate forecasts and recommendations whilst human merchants make final decisions, override predictions when appropriate, and provide strategic direction. The technology eliminates repetitive analysis but increases demand for analytical skills.
Can small fashion retailers afford inventory AI?
Software-as-a-service platforms now offer merchandising AI capabilities at lower entry costs than traditional enterprise systems. However, implementation still requires clean data infrastructure and internal expertise to configure models. Mid-market retailers with annual revenues above £50 million typically achieve positive ROI.
