Skip to main content

The first ten minutes of a supplier outage

Full volume: Sovereignty Is Measured in Response Time

A small exercise to locate the delay between recognising a failure and being able to act.

Consider a small exercise before the next supplier review. A service your organisation depends on becomes unavailable on a Tuesday morning. The supplier has acknowledged the incident but cannot give a recovery time. Nothing else about the scenario is known. Give the people who run the service ten minutes to show what they can actually do.

The ten minutes are an exercise window, not a universal recovery target. A payroll system and an emergency communication service have different tolerances. The useful question is whether the organisation can make a bounded decision with the information and authority already available.

Put one decision on the table

Choose a concrete action: pause incoming work, switch to a prepared alternative, or continue temporarily with a stated limit. Name the person who can authorise it and the person who can execute it. A supplier contact is useful; it does not establish who inside the organisation can accept the consequences of waiting.

Ask the team to open the relevant instructions. A document title in an inventory is not enough for this exercise. Can the person on duty reach the instructions when the supplier's login service is unavailable? Can their deputy? If emergency credentials are required, demonstrate access through the approved process without revealing those credentials to the room.

Keep the exercise within a controlled environment. Do not disconnect production to create a convincing demonstration. A prepared test account, a synthetic transaction and a separate recovery environment are enough to expose many gaps in the proposed sequence. Mark steps that could only be described rather than demonstrated.

Record where the minutes went

Separate recognition, decision, execution and confirmation. Note when someone establishes that the service is affected, when an authorised person chooses an action, when the action begins, and when the team has evidence that it worked. Do not combine these into one reassuring response-time number.

Waiting for a decision and waiting for a database restore need different corrections. The former may require delegated authority or a reachable deputy. The latter may require a smaller recovery set, a tested restore procedure or a revised expectation of what can be restored first. Buying another monitoring dashboard will not resolve an absent decision maker.

Record uncertainty as well. If nobody can establish how much work was lost, write that down. A successful login does not prove that records are complete or that queued work will execute exactly once. Confirmation should include the business result the recovery was meant to protect.

End with one assigned correction and a date chosen by its owner for repeating the exercise. Keep the original observations so the second run can be compared with the first. Improving the appearance of the procedure is less useful than removing the delay the team actually encountered.

This brief extends the argument in Sovereignty is measured in response time: operational autonomy becomes inspectable when people must act under a constraint. The exercise does not certify resilience. It reveals one dependency that can now be changed.