Plan a two-arm cluster randomized trial with a continuous outcome and equal allocation. The individually randomized normal-approximation size is inflated by the Eldridge, Ashby & Kerry (2006) design effect 1 + [((CV² + 1)m) − 1]ICC, then by expected participant attrition. The final answer rounds up whole clusters.
This calculator does not apply a small-number-of-clusters degrees-of-freedom correction or cluster-level attrition. When very few clusters are planned, validate with a method tailored to the intended analysis.