Insights
Patterns across all 27 kit missions: where inspections fail, how battery and configuration matter, and how the dispatcher rules would have behaved on real weather.
- 77% vs 91% raw
V1 vs V2 is confounded by batterySee the chart
V2 is better, but part of its raw lead is battery: V2 never ran below 52%. Same battery band: 40-59%: 85% vs 100%; 60-79%: 86% vs 92%.
Measured: Action success by configuration, then again inside 20-point battery bands where both configurations ran.
- 61 of 270
Asset and location disagree
Do not map an asset by the action's location alone. Check assets.csv before sending a crew.
Measured: actions.location_id compared with the asset's location_id in assets.csv.
- 50 of 270
Actions with no reading
A missing reading is not a normal reading. Treat the asset as uninspected.
Measured: Actions with no row in readings.csv, matched on action_id.
- 23 of 27 missions
The replay stream is incomplete
Do not use the live feed as the record. It stops mid-mission (M023). It has 0 MISSION_FAILED events, yet missions.csv lists 1 failed.
Measured: Distinct mission_id in replay.ndjson; last mission checked for MISSION_COMPLETED; event types counted.
- 8 of 40
Anomalies hide in successful actions
A SUCCESS action can still find a problem. Review anomalies, not only failed actions.
Measured: anomalies.csv joined to actions.csv on action_id, counting status SUCCESS.
Color = share of inspections OK there (green ≥ 85%, yellow ≥ 70%, red below). Size = anomaly count. Dashed ring = outdoor leg (assumed).
- Configuration V1
- 77% actions OK · 29 min · 56% battery per mission
- Configuration V2
- 91% actions OK · 22 min · 28% battery per mission
1 of 27 rounds held for weather
- M001 Rain up to 3.5 mm/h before Spot can dock; Spot is rated for light rain and splashes only (IP54)
9 of 27 rounds held for battery (would dock under 20%)
17 of 27 routes include the outdoor leg. Weather: Open-Meteo ERA5 archive, Sep 1-10, 2026. The archive has no thunder data, so this count does not test the thunder rule.