How Torfs achieved 50% sales uplift with AI allocation for reallocated shoes

Belgian footwear retailer Torfs has achieved sales increases of up to 50% on reallocated inventory by deploying a proprietary AI system that optimises stock distribution across its store network. The family-owned retailer built the allocation model in-house to address persistent challenges of matching inventory to local demand across its physical locations.

What problem was Torfs trying to solve with AI allocation?

Traditional merchandise allocation in multi-store retail relies heavily on manual decisions and static rules that struggle to account for local variations in customer preferences and buying patterns. Torfs faced the common industry challenge of having popular styles out of stock in some locations whilst slow-moving inventory accumulated in others, leading to both lost sales and markdowns.

The retailer recognised that footwear presents particular allocation complexity due to size curves, seasonal demand shifts and regional style preferences. Rather than adopt a third-party solution, Torfs chose to develop a system tailored to its specific business model and data infrastructure, allowing greater control over the underlying logic and ongoing refinement.

How does the Torfs allocation model work?

The AI system analyses historical sales data, current stock positions and customer purchase patterns to generate allocation recommendations for moving inventory between stores. The model evaluates which locations are most likely to sell specific products based on local demand signals, then suggests transfers to maximise sell-through rates before items require discounting.

Key inputs to the model include:

• Store-level sales velocity by style, colour and size
• Current inventory positions across the network
• Historical conversion rates for similar products at each location
• Seasonal and trend-based demand patterns
• Transfer logistics and timing constraints

The system operates continuously, updating recommendations as new sales data becomes available and inventory positions change. This allows Torfs to respond dynamically to emerging demand patterns rather than waiting for periodic manual reviews of stock distribution.

What results has Torfs achieved with AI allocation?

Torfs reported sales uplifts of up to 50% on inventory that was reallocated using the AI system compared to items that remained in their original locations. This improvement reflects the model's ability to identify stores where specific products have higher probability of selling at full price based on local customer preferences.

Beyond the headline sales increase, the retailer also achieved improvements in inventory turnover and reduction in time spent on manual allocation decisions. Store managers who previously dedicated significant hours to requesting transfers or analysing stock reports now receive data-driven recommendations that streamline the process.

The system has also reduced markdown rates by moving slow-selling inventory to locations where it can sell at full price before being discounted, though Torfs has not disclosed specific figures on margin improvement. The ability to extract more value from existing inventory without additional purchasing represents a direct contribution to profitability.

What challenges did Torfs face during implementation?

Initial resistance from store staff presented the primary adoption hurdle for the AI allocation system. Store managers accustomed to making their own merchandising decisions viewed algorithmic recommendations with scepticism, particularly when suggestions contradicted their intuition about local customer preferences.

Torfs addressed this resistance through a combination of training, transparency about how the model generates recommendations and early demonstrations of tangible results. Store teams needed to see concrete evidence that AI-driven transfers outperformed traditional methods before trust in the system developed.

Technical integration also required careful planning to ensure the allocation model could access clean, timely data from existing point-of-sale and inventory management systems. The retailer invested in data infrastructure improvements to support the AI system, recognising that model performance depends heavily on input data quality.

What does this mean for retail merchandise allocation?

The Torfs case demonstrates that mid-sized retailers can develop effective AI allocation capabilities in-house rather than relying exclusively on enterprise software platforms. This approach requires technical expertise and data infrastructure investment but offers customisation advantages and avoids ongoing licensing costs.

The substantial sales uplift on reallocated inventory highlights the persistent inefficiency in traditional allocation methods, particularly for categories like footwear where size and style variation creates complex matching challenges. Even modest improvements in directing inventory to the right locations can generate significant revenue impact.

However, the adoption challenges Torfs encountered underscore that technology deployment alone does not guarantee success. Retailers must address organisational change management, staff training and trust-building to realise the potential value of AI allocation systems. The human element of implementation often proves as critical as the technical capabilities of the model itself.

As inventory optimisation becomes increasingly central to retail profitability amid margin pressure and shifting consumer behaviour, more retailers are likely to explore AI-driven allocation. The Torfs experience suggests that success requires not only sophisticated algorithms but also careful attention to data quality, system integration and staff engagement throughout the deployment process.