Demand Sensing for New SKUs

It’s no easy task forecasting demand for new products as there is no sales history to provide guidance. But product innovation is crucial: McKinsey has found that companies typically derive more than 25% of products and sales from new products. But poor forecasting by manufacturers can lead to wasted working capital, stockouts, and strained retailer relationships.

This challenge intensifies as product lifecycles shorten and consumer expectations for novelty rise. Retailers like Target now offer dozens of similar SKUs per category, creating proliferation that overwhelms manual forecasting. Traditional methods often depend on planners’ intuition and are often inaccurate.

AI changes this paradigm by analyzing sales of related products, consumer behavior, and external signals to model demand patterns before launch, enabling confident investment decisions and smoother supply planning.

Modern AI approaches group new or proposed items with historically related products based on features such as brand, price, and packaging, allowing reliable baseline estimates.

Data architectures integrate internal and external signals—including point-of-sale data, social sentiment, search trends, seasonality, and weather—to refine forecasts. Real-time sensing enhances accuracy by continuously updating models with early sales data. While AI enhances precision, organizations must ensure data quality, system integration, and human oversight. Even advanced models can falter in the face of unforeseen “black swan” events, so quick detection of forecast errors and adaptive recalibration are essential.

Results are measurable. An eyewear brand using AI for six-month seasonal forecasts improved accuracy by thirteen percentage points. Food producer Atria achieved 98.1% weekly forecast accuracy and reduced manual changes by 13%, using AI to forecast short shelf-life goods.

AI-driven forecasting can reduce supply chain errors by 20-50% and lost sales and product unavailability by 65%, according to McKinsey. The consulting firm also says automated forecasting can cut workforce management tasks by 50%, saving 10–15% in costs while improving cross-functional collaboration and decision confidence.

Vendors in this space range from integrated enterprise platforms to specialized analytics tools. Selection should emphasize industry fitness, integration capability, and proven model accuracy.

The following list includes the major solution providers:

• E2open – Specializes in demand sensing with full supply chain integration and new product forecasting expertise.
• First Insight – Combines customer feedback with AI-driven predictive scores to de-risk new launches.
• Hypersonix – Generative AI analytics with multi-model forecasting and autonomous recommendations.
• RELEX Solutions – Unified supply chain suite with ML-based demand sensing, claims 98% accuracy for perishables.
• ToolsGroup – Service Optimizer 99+ using machine learning for seasonal and new product forecasting.
• Google Cloud AI – Scalable infrastructure and pre-trained retail forecasting models.
• AWS Forecast – Deep learning-based cloud service for high-accuracy retail and CPG forecasts.
• IBM Watson – Enterprise AI platform combining NLP and computer vision for retail forecasting.
• H2O.ai – Open-source ML platform with demand sensing capabilities for consumer goods.
• Impact Analytics – AI-native forecasting and assortment tools for fashion and apparel.

AI-powered demand sensing gives retailers foresight before launch—helping them position, price, and plan production more accurately. Yet technology alone is not enough. Success depends on integrating human judgment, governance, and rapid response mechanisms assuring that data-driven foresight translates into operational precision.