Scholarlink

When launching new products, firms must determine whether the new offering will cannibalize existing sales or expand overall portfolio performance. This requires distinguishing genuine cannibalization from coincidental shifts caused by external market conditions. This study presents a three-phase framework that combines multivariate time series forecasting with counterfactual analysis to quantify cannibalization in a rigorous way. Phase 1 identifies key external drivers such as macroeconomic indicators, market trends and competitive actions using correlation and multicollinearity analysis. Phase 2 builds counterfactual forecasts based on pre-launch data to estimate how the incumbent product would have performed in the absence of the new product. Phase 3 converts forecast deviations into business metrics such as cannibalization rate, displaced revenue and net portfolio impact. Five forecasting methods (SARIMAX, Prophet, Random Forest, XGBoost and LightGBM) were evaluated using rolling-window and growing-window validation. In the medical devices case study, the SARIMAX model with rolling-window validation showed the greatest consistency. Results indicate that Product B cannibalized 28% of Product A’s sales while still generating a positive net revenue effect, which supports the launch decision. Overall, this framework provides a robust and replicable approach for assessing product interactions and enabling evidence-based decisions in multi-product portfolios.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An End-to-End Framework for Measuring Product Cannibalization Using Multivariate Time Series Forecasting

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Daniela Martins,
  • Luís Miguel Matos

摘要

When launching new products, firms must determine whether the new offering will cannibalize existing sales or expand overall portfolio performance. This requires distinguishing genuine cannibalization from coincidental shifts caused by external market conditions. This study presents a three-phase framework that combines multivariate time series forecasting with counterfactual analysis to quantify cannibalization in a rigorous way. Phase 1 identifies key external drivers such as macroeconomic indicators, market trends and competitive actions using correlation and multicollinearity analysis. Phase 2 builds counterfactual forecasts based on pre-launch data to estimate how the incumbent product would have performed in the absence of the new product. Phase 3 converts forecast deviations into business metrics such as cannibalization rate, displaced revenue and net portfolio impact. Five forecasting methods (SARIMAX, Prophet, Random Forest, XGBoost and LightGBM) were evaluated using rolling-window and growing-window validation. In the medical devices case study, the SARIMAX model with rolling-window validation showed the greatest consistency. Results indicate that Product B cannibalized 28% of Product A’s sales while still generating a positive net revenue effect, which supports the launch decision. Overall, this framework provides a robust and replicable approach for assessing product interactions and enabling evidence-based decisions in multi-product portfolios.