This manuscript provides an empirical documentation of the existence and magnitude of heterogeneous bid shading in the discriminatory auction. Unlike existing empirical work that uses aggregate data, this manuscript makes use of a unique data set collected at the individual bidder level containing bidding behaviour and detailed cost information covering a natural experiment in which the format of the auction switches from uniform to discriminatory pricing. Preliminary regression results indicate that bid prices in the discriminatory auction are $1,813 lower than bid prices in the uniform auction. In other words, the magnitude of bid shading in the discriminatory price auction is on average 6.3% of bidders’ willingness to pay. Capacity utilization, housing type and milking system are bidder-specific characteristics identified as determinants in explaining heterogeneity in bid shading across bidders. In terms of allocative efficiency, the uniform auction achieves a higher average efficiency of 91% compared to an average efficiency of 74% reached by the discriminatory auction. The majority of the inefficiency of the discriminatory auction is attributed to the use of bid-spreading strategies while the remaining portion of inefficiency is due to heterogeneous bid shading across bidders. All auction inefficiency identified in the uniform auction is attributed to the use of bid-spreading strategies.