CarbonPilot

02 · Diagnose

What moved, and why

A robust median/MAD test flags the months that moved against their own recent baseline, then attributes each one to the activities and usage patterns behind it.

⚠ High severity Sep 2026 · z-score 42.4

Diesel litre rose 58% in Sep 2026

No single driver isolated — review the contributing activity rows.

Observed
120.9 t
Baseline
76.6 t
median, Mar 2026 – Aug 2026
Excess
+44.2 t
Suggested fix Electrify the delivery fleet −383 t/yr · costs $63K/yr · 94% confidence Simulate it →

Running the attribution…

⚠ High severity May 2026 · z-score 56.8

Cloud compute vCPU hour rose 165% in May 2026

Weekend share of cloud_compute_vcpu_hour usage rose from 15% to 60% (4.1x the previous baseline).

Observed
43.0 t
Baseline
16.2 t
median, Jan 2026 – Apr 2026
Excess
+26.8 t
Suggested fix Right-size weekend cloud workloads −66 t/yr · saves $3.5M/yr · 75% confidence Simulate it →

Running the attribution…

⚠ High severity Mar 2026 · z-score 4.2

Flight passenger km rose 15% in Mar 2026

No single driver isolated — review the contributing activity rows.

Observed
88.0 t
Baseline
76.7 t
median, Jan 2026 – Feb 2026
Excess
+11.3 t
Suggested fix Shift internal travel to virtual meetings −235 t/yr · saves $464K/yr · 81% confidence Simulate it →

Running the attribution…

⚠ High severity Mar 2026 · z-score 4.1

Hotel night rose 17% in Mar 2026

No single driver isolated — review the contributing activity rows.

Observed
7.4 t
Baseline
6.4 t
median, Jan 2026 – Feb 2026
Excess
+1.1 t
Suggested fix Sustainable-hotel travel policy −8 t/yr · saves $134K/yr · 78% confidence Simulate it →

Running the attribution…