We introduce novel metrics to evaluate production process heterogeneity using both machine learning (ML) and traditional kernels. ML kernels, particularly through economically motivated transfer learning models, enhance M&A forecasting accuracy. A wider gap in firms' production processes predicts fewer M&As, lower success rates, reduced returns, diminished post-M&A growth, and increased divestiture. Dynamic learning among repeat acquirors mitigates adverse effects of production process dissimilarity on post-M&A growth. The adoption of Right-to-Work laws, reducing employees' bargaining power, significantly alleviates detrimental effects of heterogeneous production processes. Our findings underscore the pivotal role of technology heterogeneity in shaping integration synergy and firm boundary decisions.