Advanced
Lesson 12 of 14 · ~7 min

Calibrated uncertainty: escalate before or after

Sometimes escalate before; sometimes escalate-after-with-documentation. The framing rule resolves edge cases: senior time is cheaper than client-outage time.

Most failed change tickets fail in the gap between I should escalate and I’ll just try it. The two-question test is one lens; calibrated uncertainty about when to escalate is the next.

Sometimes the right move is escalate-before. Sometimes escalate-after-with-documentation. Knowing which keeps you fast without crossing the line into damage.

The two timing choices

Escalate before acting when:

  • The action is in the irreversible / high-blast-radius column (DNSSEC, registrant change, redemption recovery, anything you can’t undo within a TTL or two).
  • The action is outside any runbook you have access to.
  • You don’t understand the why of the action. The senior said do it with no rationale; the client asked for it with no clear end state; the vendor’s instructions say so with no understanding of consequences.

Escalate after acting (with documentation) when:

  • The action was in-scope per the runbook or the explicit instruction you had.
  • Something surprised you mid-action (a panel field labelled differently, a vendor’s verification step revealing additional requirements, a TTL behaviour that didn’t match prediction).
  • The action is recoverable within a TTL or two; the surprise is informational, not destructive.

The document part of escalate-after matters. Write what surprised you, what you decided, what you’d do differently next time. The senior reads the ticket later and either confirms your call or coaches you.

The framing rule

Senior time is cheaper than client-outage time

When you’re hesitating between escalate or proceed on a borderline action:

  • Senior’s time to confirm or correct: minutes.
  • Client outage from a wrong action you didn’t escalate: hours.
  • Multiplied by reputation cost: often days of recovery and lost trust.

Escalation usually costs less than the alternative. Seniors prefer interrupted by a tech with a calibrated question to discovering a tech took an unauthorised action mid-incident. Interruption is the right call.

Choosing the channel

ChannelWhen
Synchronous (chat, phone, walk over)Time-sensitive in-progress actions; senior is busy but you need an answer now.
Asynchronous (ticket comment, email)Actions that can wait an hour or two. Name the decision the senior needs to make and when you need the answer.
Documented escalate-afterActions you took within scope but want a senior to confirm. Ticket note describing the action, the surprise, the decision, the outcome.

What this is NOT

  • “Escalation is a sign of weakness.” Seniors expect escalation on the right work. Hiding uncertainty looks worse than asking.
  • “Documentation-after is paperwork.” It’s how the team’s collective knowledge accumulates. A well-written what surprised me note is more valuable than a clean ticket close.
  • “Ask once and you’re done for the shift.” Each high-stakes action gets its own calibration call.

Decision walkthrough

Same action shape, different calibration
Two scenarios below. The same tech response shape (proceed, ask, document) applies differently to each. First: the senior asks for a TTL drop on a client's A record.
The senior asks you to lower TTL on a client's A record from 86400 to 300, ahead of Friday's planned web host migration. Escalate before or after?
Same shape, different blast radius
Now flip the scenario: the senior asks you to disable DNSSEC and move nameservers in one operation.
DNSSEC disable + nameserver change in one op. Calibration?
Next lesson