Our objective was to add an orthogonal dimension of milk urea nitrogen (MUN) concentrations to an existing orthogonal (for fat, true protein, and anhydrous lactose) milk calibration sample set used by the USDA Federal Milk Marketing orders. MUN data is used to improve protein feeding efficiency, nitrogen utilization, and reproductive efficiency. However, even though multivariate statistical tools are available for prediction of MUN through MIR analysis, calibration samples with a wide, well-distributed range of MUN concentrations (from 12 to 25 mg/100 g milk) to adjust slope and intercept to local conditions are needed. The calibration sample set was produced using a combination of pasteurization, gravity separation, and ultrafiltration to make milk-based ingredients, plus pure urea, pure lactose monohydrate and water to formulate the orthogonal design 14-sample set. Once the calibration milks including MUN were successfully developed and produced once every 4 wk, they were sent to 9 laboratories for determination of MUN using a chemical reference method (spectrophotometric enzyme-based assay expressed as mg/100 g milk) and were used to calibrate their MIR instruments starting in October of 2018. The MUN reference method had excellent performance with relative standard deviations (RSD) for within and between laboratories of <1%. Within (S_r) and between (S_R) laboratory method performance statistics for the MUN reference method in January 2019 were S_r = 0.069, S_R = 0.104, RSD_r = 0.395, RSD_R = 0.595, r-value = 0.193, R-value 0.291. MIR instruments use partial least squares (PLS) models to predict MUN from the MIR spectra for each milk. Different PLS models from one manufacturer to another may perform differently. Predicted MIR data for MUN produced by analysis of the orthogonal milk calibration sample set can be used to run a diagnostic evaluation of how well a MUN prediction model compensates for background variation in fat, true protein, and anhydrous lactose from milk to milk.