We analyse the expected performance of various group testing, or pooling, d
esigns. The context is that of identifying characterized clones in a large
collection of clones. Here we choose as performance criterion the expected
number of unresolved 'negative' clones, and we aim to minimize this quantit
y. Technically, long inclusion-exclusion summations are encountered which,
aside from being computationally demanding, give little inkling of the qual
itative effect of parametric control on the pooling strategy. We show that
readily-interpreted re-summation can be performed, leading to asymptotic fo
rms and systematic corrections. We apply our results to randomized designs,
illustrating how they might be implemented for approximating combinatorial
formulae.