The error pattern tells you more than the error size does
Whether a part comes out too big, too small, or just wrong in a specific spot, the pattern of that error across the part is more diagnostic than the raw percentage. An error uniform across every dimension points toward a global cause, calibration or shrinkage. An error isolated to holes specifically points toward the well-known FDM hole-undersizing effect rather than a machine problem at all. An error that is inconsistent from print to print, even with identical settings, points toward a mechanical issue like backlash rather than a clean, fixable number.
How the triage works
Your reported pattern routes the triage directly: holes-only points to the Hole and Slot Printability Check's orientation-based undersizing, inconsistent-across-part points to a mechanical cause like belt tension. For a uniform error, a recent firmware or steps/mm change strongly suggests scale calibration; a small error (under about 2%) with no recent change is more likely ordinary material shrinkage; a larger unexplained uniform error splits weight between scale and flow-rate calibration.
Worked example
A uniform 1.5% error, no recent firmware or steps/mm change: since the error is small and nothing was recently changed, the tool ranks material shrinkage as the top hypothesis and points to the Shrinkage Allowance Planner for comparison, rather than assuming a machine calibration problem exists at all.
A common mistake
A frequent error is recalibrating steps/mm in response to a single measurement that turns out to be an outlier, then discovering the "fixed" printer is now off in the other direction on the next print. Measure the same calibration cube two or three times, ideally across separate prints, before adjusting a global setting based on what could just be normal print-to-print variance.
Limitations
This tool ranks plausible causes from the pattern and measurements you describe using measured data, not a guess. It cannot access your printer's actual firmware settings or measure your parts directly, so treat the ranked list as a place to start testing, not a confirmed root cause.