Title: METHODOLOGY FOR USING ARTIFICIAL INTELLIGENCE TO ASSESS CUSTOMER NEEDS IN LOW- AND MEDIUM-COMPETITION PRODUCT NICHES
Author:
Artem Korshun
Abstract:
The study proposes an Intelligent Demand Mapping Framework (IDMF) that integrates a Multimodal Temporal Fusion Transformer (MTFT) to evaluate customer need in low- and medium-competition Amazon niches. IDMF first clusters ultra-sparse search queries with lightweight transformer sentence embeddings and product-image encodings, then quantifies “behavioural gaps” between clustered intent and on-shelf offers through a Behavioural Gap Index (BGI) and a review-weighted Herfindahl score. A Monte-Carlo engine, seeded by MTFT quantile forecasts, converts those diagnostic signals into risk-adjusted return projections. Experiments on 7 600 long-tail queries across three home-organisation sub-categories yield 45 intent clusters where BGI > 0.40 and competitive friction < 0.25; in those clusters the mean absolute percentage error of monthly-sales predictions improves by 11 percentage points versus a text-only baseline, and simulated twelve-month return on invested capital reaches 31 % under conservative advertising assumptions. These results demonstrate that fusing sparse textual and visual cues with temporal forecasting narrows uncertainty around demand, enabling smaller sellers to prioritise profitable launch opportunities without GPU-heavy infrastructure. The framework is fully reproducible on consumer-grade hardware and invites future extensions such as live A/B validation, cross-domain transfer, and sustainability-aware profit metrics.
Keywords: Multimodal prognosis; Temporary merger transformer; Index behavioral gaps; Amazon niches; AI calls for mapping.
PDF Download