Investors actively optimize for trading costs when making portfolio decisions. Yet, prominent asset pricing factors aiming to summarize the investor opportunity set follow fixed rebalancing rules that overlook the cost of trading. I propose a class of transaction-cost-aware (TCA) factors --- explicitly optimized to explain net-of-cost returns --- that bridge this gap. TCA construction controls how aggressively factor portfolios adjust when the underlying characteristics change. Standard factors rebalance in full, irrespective of whether marginal expected returns cover trading costs incurred. My methodology finds the optimal balance between these two forces. TCA factors generate higher risk-adjusted returns net of costs than unoptimized factors based on the same characteristic. Spanning regressions of TCA factors on their unoptimized counterparts consistently deliver positive alphas. Further, models that employ TCA factors come closer to spanning the feasible efficient frontier for investors facing non-zero transaction costs. These performance gains carry over out-of-sample and are robust to cost-mitigation techniques proposed in the previous literature. Construction inefficiencies addressed by TCA factors significantly impact asset pricing inference. The benefits of TCA construction are heterogeneous and most apparent for high-turnover factors, such as momentum. Such factors appear unprofitable after costs if constructed suboptimally but regain prominence in the TCA setting. Therefore, asset pricing tests on unoptimized factors can reach incorrect conclusions on which economic characteristics matter for the cross-section of expected returns.