Archive · 12 issues
Two foundational issues on failure mechanisms that apply everywhere, then one issue per sector. Each is available as a web page and as a PDF suitable for circulating to a board.
Commercial SaaS observability cannot follow classified data across the boundary. The result is that the estates with the highest consequence of failure frequently have the least visibility into it.
This sector has the smallest technology estates in the series and the largest individual transactions. A closing-day failure jeopardises a deal worth more than the entire annual IT budget.
For a carrier, 911 availability is not an SLA — it is a statute with an FCC filing attached. For a broadcaster, the kick-off happens once. Both end at the same place: you have seconds, not minutes.
A missed sort window does not cost one shipment. It cascades through a network for days, and the recovery costs more than the original failure — which is why an hour of diagnosis here is not worth an hour.
Ask a plant manager what a stopped line costs per minute and they will answer from memory. Ask them what a degraded one costs and they will look puzzled — because there is no such thing.
Supply must match demand continuously, IT failures propagate into physical systems, and a regulator counts. In energy, the most dangerous telemetry signal is not an alarm — it is silence.
These companies sell reliability to everyone else. Their outages come with a status page, a postmortem the industry reads, and SLA credits — which makes honest positioning harder and more necessary.
Black Friday does not move. A retailer’s entire year of reliability engineering is graded across roughly five days, and the cost of failure is a basket that goes to a competitor one tab away.
A payment API written this year calls a core system written in 1987. Neither team can see the other’s telemetry. That seam is where your most expensive incidents live — and where your regulator’s deadline starts running.
A retailer’s outage costs a basket. A hospital’s outage diverts an ambulance. Healthcare is the one sector where the right answer to a low-confidence diagnosis is to refuse to act.
Your automated investigation queried the last sixty minutes. The incident started four hours ago. Every signal came back flat, the agent reported no anomaly, and it was right — about an hour in which nothing happened.
An AI investigator never says “I don’t know.” It says something plausible, in complete sentences, with no tremor in its voice — and at 3 a.m., under pressure, that is indistinguishable from knowing.
Next issue
Subscribe and it will be in your inbox on a Tuesday morning. No gated content, and no more than one product mention per issue.
Fortnightly · Tuesday 07:00 ET · Unsubscribe in one click