Recent sociophonetic analysis typically involves a pipeline of procedures to turn a spreadsheet of formant measurements into interpretable numbers and figures. However, little has been said about the order that these procedures should be applied. To explore the effect that order of operations has on sociophonetic analysis, a dataset containing formant measurements from 53 speakers was analyzed 5,040 different ways, each representing a different permutation of seven common steps in the analysis pipeline. The Low Back Merger and the Low-Back-Merger Shift were examined in these 53 speakers using three different metrics. The results indicate that changing the order of operations produces variation similar in magnitude to sociolinguistically meaningful variation and can influence the interpretation of any one speaker’s data and the entire dataset. This paper concludes with a recommended order that researchers can use in their analyses: classify allophones, remove outliers, normalize, and then subset.