Articles

What we measured.

Benchmarks we ran ourselves, and how to use what they found. Every number comes with the data behind it.

Benchmark

One row, several buildings

Finding the schedule rows that bundle a campus or a complex, so you can split them before you model. What a decision model catches, and what it misses.

92.4%of bundled rows found on real public schedules
Benchmark

What a decision model gets right and wrong on 1,430 exposure rows

Three slices, two fields, one keyword baseline. It reads typos and other languages, and it does not know codes from other systems.

98.4%construction codes right on real public schedules
Practice

Confidence gates for exposure data

One threshold for every field is either too strict or too loose. How to pick one per field, and what to count when you do.

7.7wrong occupancy codes per 1,000 rows at our gate
Hands-on

Code a schedule of values with one API call

Three real rows, one POST, and how to read what comes back: the code, the confidence, and the rows left for you.

1 callrows in, OED codes out

Try it on your own rows in the SOV coder, or read the API reference. Get a free API key.