Coronavirus testing using binary search method

Would getting fluid samples from 2 groups of say 50 people, combining the samples and testing that, then keep halving and testing recursively the groups that return positive, be a viable way of testing large sums of people with far fewer tests than individual person by person tests.

I'm not very familiar with all the tests run, but surely something like this could be applied to the antibody one?

Would love to hear if anyone with more knowledge of the testing process and capabilities could give their thoughts...