Own the signal.Rent the platform.
Observability strategy, incident leadership and AI operations for the people who are accountable when it breaks. Written by Allan Mann, with nothing to sell you but the thinking.
Allan Mann
Twenty-five years running IT operations and observability across five tier-1 banks and one UAE government, the environments where "it looked fine on the dashboard" ends up in front of a regulator.
Author of Metrics & Mayhem: A CTO's Guide to Observability That Actually Works. Host of the Metrics & Mayhem podcast. Vendor-neutral by construction: no platform reselling, no vendor sponsorship, no referral fees.
Most observability advice is written to sell a platform.
This is not. No vendor sponsorship, no reference architecture that happens to need three seats of something. Just what actually holds up in a large enterprise, and what quietly does not.
It was never the tooling
OpenTelemetry is not what is blocking you. Ownership boundaries, funding models and the team that will not go first are. Those are the pieces nobody writes about.
Cost is a design decision
Cardinality, retention and sampling get chosen by accident, then defended for two years. Decide them on purpose and the invoice stops being a surprise.
AI needs a gate, not a leash
Autonomous SRE agents will run queries you would never approve and spend budget while they think. The question is where accountability actually sits.
What separates the teams that recover fast from those that do not
Chapter 4 of Metrics & Mayhem. The chapter that answers the question every engineering leader is afraid to ask out loud: when it breaks, who actually owns getting it back?
- Why one airline recovered from the CrowdStrike outage in a day and another took five, from the identical software fault.
- The three-layer ownership model: outcome owner, indicator owner, platform owner.
- Why most organisations build only the platform layer, then wonder why their P1s take an hour.
- One concrete Monday action you can take before you close the page.
Recent writing
The thinking, in public
Three ways in
The Signal
One email every Friday since February 2024, issue 103 and counting. One idea, argued properly, for people who have to make the call on Monday.
Subscribe →Metrics & Mayhem
The podcast. Short audio notes from the front line of IT operations, with deep dives when a topic earns the longer cut.
Listen →The book
A CTO's Guide to Observability That Actually Works. Eleven chapters. Chapter 4 is free, and it stands on its own.
Read chapter 4 free →More tools. More dashboards. Worse recovery.
The Observability Assessment
A fixed-scope review of one part of your observability: one platform, one service domain, or a defined set of questions. Not the whole estate.
-
01
Prework, before I arriveAn intake pack: access, current dashboards and alerting, recent incident data, and the people I need to speak to. The work does not start until this is in.
-
02
Review and interviewsI work through what you have supplied, and I talk to your people. The conversations matter as much as the data.
-
03
Findings and roadmapA written assessment of what is working, what is theatre, and a prioritised roadmap, plus a readout session with your team.
Ongoing advisory from £1,250 per day, scoped and quoted per engagement. No retainer you have to justify.
One email a week.
No vendor pitch at the end of it.
Join 561 IT and engineering leaders who get The Signal every Friday.