Accuracy benchmark
How often MessyAPI gets it right
Every case has a known correct answer, taken from published reference values, real public datasets (GeoNames, US Census, Social Security, IANA), answers computed from the specs, or checked by hand — never from our own output. The suite runs against the live tools once a day. Failures are shown, not hidden.
Running benchmark…