We budgeted six months for consent design on the South Asia genomics project. It took two years, and we'd do it again.
The problem with one-time consent
Standard "sign once, data is ours forever" consent doesn't hold up well over a multi-year public dataset. We wanted something revocable.
What we built instead
A tiered consent model: participants choose what their data can be used for, and can withdraw future use at any point, developed jointly with regional ethics boards rather than imported wholesale from elsewhere.
The cost
Slower recruitment, more paperwork, harder logistics. We think a dataset people can trust is worth more than a faster one they can't.