# KNX to Control4 processing benchmark 2026

Published: 2026-08-18  
Dataset cutoff: 2026-08-18 10:11 UTC  
Canonical HTML: https://coduworks.com/research/knx-control4-processing-benchmark-2026

This anonymized production benchmark includes 132 completed KNX imports, 248 completed AI mapping/translation jobs, nine completed KNX reprocess analyses, 97 active KNX project files, and 7,393 active generated Control4 devices. It measures backend time from job start to completion; it does not claim manual-work time savings.

| Workflow | Completed | Median | p90 | Maximum |
|---|---:|---:|---:|---:|
| KNX import | 132 | 35.7 s | 130.7 s | 522.8 s |
| Location mapping | 113 | 9.4 s | 20.4 s | 55.1 s |
| Device mapping generation | 103 | 182.8 s | 408.0 s | 687.6 s |
| Thermostat processing | 12 | 146.4 s | 279.8 s | 307.9 s |
| Control4 translation | 10 | 49.7 s | 92.4 s | 98.0 s |
| KNX reprocess analysis | 9 | 50.0 s | 149.3 s | 318.6 s |

Active KNX project file size: median 13.29 MB, p90 21.04 MB, maximum 80.47 MB. Active device population: 6,388 lights, 798 blinds, 109 thermostats, 94 IP gateways, 2 keypad buttons, 1 relay, and 1 motion sensor.

Method: a read-only PostgreSQL transaction queried aggregate production counts and timestamps. Durations use `completed_at - started_at`; percentiles use PostgreSQL `percentile_cont`. No user, project, filename, address, mapping, prompt, or output content was selected or published. The population is operational rather than randomized, and durations exclude queue waiting, human review, Composer work, and commissioning.