Over time, I have worked on a suite of data infrastructure projects, ranging from small to large and simple to complex. Some have been successful, others less so, and a couple are so notorious they are generally not mentioned in polite company. Sometimes the outcome was predictable—well-organized and well-managed projects are likely to succeed, and those with clear flaws often don’t rise above them—but sometimes it was unexpected. Building effective data infrastructure is not a solved problem, it is an area of research in and of itself, and the larger the project, the more difficult and uncertain it is. For each of my current and past projects, I consider their greatest strength, their showcase best practice, their weaknesses, and their epic fails, as well as the external and internal factors that contributed to positive or negative outcomes. Despite the heterogeneity, certain patterns emerge as lessons learned: Listening to your users is critical, and there are no short-cuts; it requires a long-term investment, including cultivating individual relationships. Good is better than more; hardening infrastructure is time consuming but critical for adoption. Have a clear mission with concrete benefits to a defined user community, and expand thoughtfully from there. In spite of our best intentions, meeting emerging community expectations usually requires a catalyst. External groups identifying best practices, mini-grants for implementation, and multi-project groups collaborating to rise together give needed nudges. Interestingly, I have never been involved with a project that failed because it picked the wrong technology. It can cause pain and consume resources, but I have not seen a terminal impact. Finally, invest in the people behind the infrastructure. Committed, engaged staff supported by ongoing professional development, rational management, and sufficient autonomy can, and regularly do, accomplish the impossible.