This paper demonstrates that in a simple setting with managerial concern for reputation and asymmetric information on ability, most managers may refrain from undertaking innovations which stochastically dominate an industry standard. Common components of uncertainty in outcome lead the market to form inferences of managerial ability based on relative performance. Managers who undertake the industry standard are in turn evaluated with a more accurate benchmark than those innovating. Discontinuities in compensation when performance is low (due to firings) lead managers to have differing valuations of an accurate benchmark, depending on type. We find that very high and very low ability managers are more likely to undertake superior innovations than those of average ability. When the expected improvement under the new action is small, most managers with access to the innovation choose to forgo it in favor of the industry standard.