A mechanism with low direct cost of use may be preferred to alternatives implementing more efficient allocations. We show this experimentally by giving pairs of subjects the option to agree on a single average price for a sequence of trades - in effect pooling several small bargains into a larger one. We make pooling costly by tying it to some inefficient trades, but subjects nevertheless reveal strong tendencies to pool, particularly when more bargains remain to be struck and when bargaining is face to face. The results suggest that implementation costs could play a significant role in the use of many common trading practices, including the employment relationship.