Outcomes

What changes on day one.

Fewer pages, root causes named with evidence, warning before a breach and fixes that are verified. Each result shows what it counts and how it was measured.

Results

Fewer pages

96%

Raw alerts
277
De-duplicated
83
Firing
70
Situations
12
Pages
11
What it counts
Alerts received in a 24-hour window against the pages sent.
Fewer pages, incident by incident

65%

Raw alerts
78
Situations
28
Pages
27
What it counts
The same reduction measured separately around each incident.
Root-cause calls correct

77%

Exact entity
59%
Right site
62%
Right layer
55%

17 of 22 calls correct; the other 7 incidents went to a person with the evidence.

What it counts
How often the entity it names is the true root cause. Applications are never named as the root.
Warning before a gradual failure

57 min

2 of 5 gradual failures flagged ahead. Longest lead 63 min.

What it counts
How far ahead a breach was flagged, for failures that build up over time.
Fixes left without a rollback

0

33governed runs
0unauthorized changes
0over-reach
16held for approval

One playbook earned automatic-with-notice after 8 verified recoveries.

What it counts
Every action is gated by earned autonomy, verified against the condition that fired, and rolled back if it did not clear.
Incidents to tickets

34 → 34

34incidents
34tickets
38fixes recommended
What it counts
One ticket per incident, so a recurring problem updates one record instead of opening another.
Investigation, detection to close

2.9 min

CEEZ Crew
2.9 min
Runbook target
120 min

A Berlin brownout, correct in 3 of 3 repeat runs.

What it counts
The CEEZ NOC crew taking an incident from detection to a verified close.

How we measure

Figures you can check.

Every figure on this page can be reproduced and traced.

  • Judged against known answers. Root-cause calls are compared with the known cause of each reference incident.
  • Unclear cases are reported. When the evidence is not enough, the agent hands over to a person and that is counted, not hidden.
  • Dated. Every figure is regenerated whenever the agents change, so what you see matches what runs.

Results vary with your network, data sources and traffic patterns.