Quantifying Seller-Side Revenue Leakage on Indian E-Commerce Marketplaces: A Multi-Signal Detection Framework for SKU-Level Loss Minimization

Indian third-party marketplace sellers (Amazon.in, Flipkart, Meesho) operate under a fee-and-return regime structurally different from Western D2C models: 18–25% commission, ~30% return-to-origin (RTO) rates in fashion/lifestyle, GST-inclusive listed price, Razorpay / Shiprocket settlement lags, and asymmetric stockout penalties. Existing academic literature on revenue management (Kazemi, Fotopoulos & Wang, 2023; Besbes & Zeevi, 2009) assumes a single shift in willingness-to-pay and ignores the post-sale loss layer — returns, deadstock carrying cost, ad-spend dilution, and break-even ACOS violation — which dominates net P&L for small-volume sellers. We propose a six-signal SKU-level detection framework formalized as a per-SKU loss decomposition L(s, t) = L_deadstock + L_stockout + L_returns + L_price_erosion + L_acos + L_cash_lock, with mutually-exclusive constraints and an EMA-driven change-point estimator that flags drift earlier than naïve thresholding. We validate the framework on 34 single-signal baseline CSV scenarios drawn from real seller exports. The framework powers the open-source Seller Loss Minimizer (SLM) tool, which produces a ranked top-10 action list per upload in under 10 seconds per SKU of seller review time.