I study theoretically and experimentally the performance of four different first-price auction designs (treatments): the Product-Mix auction, standard simultaneous auctions, and two sequential auction formats. Two differentiated varieties are for sale, and buyers wish to buy at most one of them. While the Product-Mix auction offers an intuitive design for this environment, sequential auctions allow greater strategic flexibility. Efficiency, revenue, and bidder surplus rankings are theoretically ambiguous and thus particularly interesting to examine in the lab. Subjects decide how to shade their bids, facing additional exposure to winning more than one unit in some treatments. Bidding behaviour is compared to Bayes-Nash equilibrium predictions.