Expansion of e-commerce presents new opportunities for small and medium enterprises (SMEs) to enter broader market at lower costs, but the SMEs face barriers to growth after entry. To facilitate new entrants to overcome these barriers, we implement a training program as a randomized controlled experiment with over two million new sellers on a large e-commerce platform. The training focuses on practical skills specific to online business operations. Treated new sellers with access to the training earn higher revenues. These sellers improve marketing skills and attract more consumers to their online stores. Leveraging detailed consumer-seller matched search and browsing data, we find that consumers have higher purchase probability when they encounter new sellers. When consumers make purchases, they choose treated new sellers over incumbents; moreover doing so does not lower the quality of their purchases. We use a structural model to characterize consumer demand and recover sellers' underlying quality. Both treated and control new sellers have higher quality compared to incumbents. The training increases new sellers' likelihood of being encountered by consumers, which improves the matching quality between consumers and sellers. The counterfactual exercise shows that training leads to higher consumer surplus and sellers' total revenues due to market expansion. The platform could benefit in both short and long run because of the training.