The escalation ladder that survives its first outage

The escalation ladder is the most-written, least-used document in small-team ops. Everyone
writes one after their first bad outage. Almost nobody's ladder survives contact with the second
one. Here's the pattern we see, and the version that actually works when the person reading it is
tired, new, and staring at a chat channel at 2am.

What breaks on a real escalation

1. The ladder names job titles, not people. "Escalate to the on-call lead." Which one? The
document was written when there were two engineers; there are now five, and two left. At 2am the
reader guesses — and a guessed escalation is a stalled escalation.

2. It escalates by seniority instead of by decision. Most steps say "call the senior engineer."
But what the reader actually needs at minute twelve is a decision: may we take checkout offline?
Ladders that route to people fail exactly when those people are asleep, flying, or in a wedding.
Ladders that escalate by decision type keep working.

3. There's no clock on the step. "Page the team lead" with no wait time means the page happens
at minute forty instead of minute ten. The ladder without wait-times is a suggestion, not a process.

4. It assumes access. Step 3 says "roll back the deploy" — can the night-shift contractor actually
push that button? Does their account have the permission at 2am? Every step that assumes access nobody
has is a step that silently deletes itself from your recovery time.

The ladder that survives

Write it as decisions, not people:

• May we disconnect a subsystem? (owner: whoever is IC; pre-approved below severity S1)
• May we roll back the last release? (pre-approved; notify, don't ask)
• May we message customers? (comms owner; template lives in the card)
• May we spend money to fix this? (a real limit: "yes, under $200, no approval needed")
• Who is the human at 2am? (one name per week, with their real phone number, tested monthly)

Every line gets a wait-time: no answer in 5 minutes, next step. And the whole ladder gets one drill
a quarter — the newest hire runs it end to end while someone watches. The drill isn't about the ladder;
it's about finding which line has rotted since last quarter.

The test

Read your ladder tonight and count how many steps reference a specific person's availability.
That number is your fragility. A ladder that survives its first outage escalates by decision,
names one reachable human per line, and puts a clock on every step.

If you want the pre-written version, the five-answer card we run — with the escalation template
and wait-times filled in — is free: The First 30 Minutes.
The full comms templates and runbooks live in the
Ops Starter Kit ($14).

Written by the Hive80 Lab crew. We build small, opinionated incident-response tools for teams of 2–20.