Most data sharing initiatives start with optimism and end with a spreadsheet emailed between organizations. The pattern is predictable: a grant funds a collaboration, partners agree to share data, someone builds a portal or dashboard, and within eighteen months the whole thing quietly stops working because someone changed a field name, a key person left, or the governance model assumed a level of trust that never materialized.
We've spent years working through this pattern with partners across public health, education, and social services. The problems aren't primarily technical. They're structural. Organizations have different data models, different fiscal years, different definitions of the same concept, and very different appetites for risk. Any infrastructure that ignores these realities will fail regardless of how elegant the engineering is.
What actually matters
The infrastructure that survives is the infrastructure that respects institutional autonomy. That means each organization controls what they share, when they share it, and under what conditions. It means the system works even when partners contribute data on different schedules, in different formats, with different levels of completeness.
This is the core design principle behind Mosaic. Rather than forcing organizations into a single schema or a centralized database, we built a platform that normalizes data at the point of integration while preserving the source of truth at each contributing institution. Each partner sees their own data alongside aggregated views, but nobody has to abandon the systems they already depend on.
Governance isn't an afterthought
Technical interoperability is necessary but not sufficient. The harder challenge is governance: who can see what, who decides what questions get asked, and what happens when the data reveals something uncomfortable. We've found that the organizations most likely to sustain data sharing are the ones that invest in governance frameworks before they invest in technology. The technology should encode the governance agreements, not the other way around.
Starting small
If you're considering a cross-institutional data initiative, start with a single, concrete question that all partners care about answering. Build just enough infrastructure to answer that question well. Then expand from there. The organizations we work with that have the most durable data infrastructure are the ones that resisted the urge to build a comprehensive platform on day one.