FULLFACT Inc. has released a free practical guide (white paper) titled 'Beyond Forecast Accuracy: Identifying the SKUs That Benefit Most from AI.' This guide helps businesses determine which SKUs will benefit from AI-driven demand forecasting and how to design procurement authority, before getting swayed by vendor claims like 'forecast accuracy improved by XX%.' Background: Accuracy Metrics Alone Won't Get Approval Implementing AI for demand forecasting isn't just about high model accuracy. While staple products can be managed based on experience, new products, seasonal items, items around promotional periods, and weather-sensitive goods often lead to forecasting errors—resulting in both waste and stockouts that erode gross profit. Simply trying to 'reduce stock' or 'hold more inventory' tends to worsen one issue or the other, so a structured approach is needed: operations must be designed separately by SKU type. Vendor claims of 'XX% improvement in forecast accuracy' reflect model evaluation metrics, not actual profit impact. More decisive factors for successful implementation—beyond accuracy—are identifying which SKUs benefit most from AI, preparing the right data, and integrating forecast outputs into procurement authority. This guide organizes those critical decisions and implementation steps. What You'll Learn from This Guide - Five criteria to judge whether AI 'works / doesn't work / works conditionally' for each of five SKU types: staples, new products, seasonal items, items around promotions, and weather-sensitive products - Prioritization of data preparation for POS, weather, campaigns, and competitor pricing, including realistic timeframes required - How to choose among three options—building in-house, using cloud-based forecasting services, or integrating with accounting/inventory SaaS (e.g., freee, YAYOI)—based on your company's scale - Three-tier authority design for when AI predictions conflict with frontline buyers: AI-led, human override, or mutual a