Despite the attention given to venture capital (VC) and venture capital fund performance over the past decades, our understanding of what drives VC fund performance remains limited. This is mainly caused by a lack of data on VC fund performance, and the fact that, by consequence, mainly indirect measures have been used to assess it, including the number of IPOs realized by a fund or the growth of portfolio companies managed by the fund. This paper aims at complementing and fine graining the previous findings by using direct measures of VC performance in a large dataset of 4,222 funds established worldwide. Using the Preqin dataset, uniting both measures of realized and forecasted returns, we analyze to which extent VC funds’ learning affects fund performance. We find that follow-on funds outperform first time funds, and find that sector and stage specific learning is beneficial to VC fund performance. Specializing at both levels simultaneously however negatively affects VC fund performance. Our research has implications for practitioners, including VC fund managers and VC fund investors, industry and academia.