Opens in a new tab
Commentary

A blameless review starts with what the responder saw

In brief

Two workers share a name. A blameless review can show why one looked like the other and test whether a change helps the next responder choose.

4 min read

Is this the right starting point?
Before you read
No specialist technical background required; a real or hypothetical incident can supply the decision to examine.
What you will take away
Reconstruct a mistake in context and choose a testable follow-up while separating learning reviews from conduct processes.
Sources
Three illustrated engineers reveal large TEST and PRODUCTION labels beneath two identical export-worker names.
On this page4 sections

Both entries in the restart menu say export-worker. One runs tests. The other builds reports for customers. In this hypothetical incident, the responder meant to restart the test worker and chose the production one. Clicking either entry starts the restart. The resulting activity log shows the account and region, when those details can explain the mistake but can no longer help prevent it.

A blameless review keeps the wrong selection in the record. It also reconstructs the decision around it, including what the responder could see, what they expected to happen and what could still stop the command. That gives the team something it can change before someone else opens the same menu.

What the responder could see before clicking

Start with the screen the responder actually used. An audit log may give the exact time of the production restart, yet say nothing about how the targets appeared or what someone said on the call. A screenshot from the next day helps only if the relevant interface was still the same. Ask the people involved to correct the sequence, and leave an open question where their accounts or the artifacts disagree.

The difference matters when the answer has become obvious to everyone in the room. Once someone has put a red box around the production account in a timeline, it can be hard to imagine missing it. Google’s guidance on blameless postmortems asks teams to study the conditions that shaped the choice. Here, that means getting the account label back into the review at the moment it became visible, even when that makes the account of the incident less tidy.

Reconstruct the decision with the information available then
A sequence from the information and options available at the time to the action, safeguards and a follow-up test.
Work from what was visible before the restart to a change you can test in the same kind of choice.
Read diagram description

Read the diagram from the information available to the responder, through the available options and constraints, to the action, result and safeguards. Use that record to choose a follow-up hypothesis you can check.

Give the two workers names people can tell apart

Put the environment, account and region beside each worker before selection. Then give someone the original task in a safe exercise and watch where they look. Can they identify the test worker before activating either entry? That is the protection the interface change is supposed to add. A completed ticket establishes that the labels shipped. The exercise tells you whether someone can use them at the point of choice.

Training would make sense if the responder could see the account labels and needed help interpreting them. Training is a poor way to reveal a label the screen has yet to show. Even a successful selection exercise covers this particular confusion. A saved command pointed at the wrong account might bypass the menu altogether and need a different safeguard.

Tell people which conversation they are in

The review still needs someone to say, plainly, “I selected the production worker.” People can do that more candidly when they know the purpose of the meeting. Explain how the learning review will be used. Handle a separate concern about conduct or performance through its proper process, and describe any limits on how the incident account can be used. A promise of immunity the team cannot make would give participants the wrong picture too.

Give the menu repair a test and an owner

The team that owns the interface needs time to change it and a place to repeat the selection exercise safely. If the repair has to wait, record who can accept the remaining risk and when that choice comes back for review. A detail that helps explain the incident can stay in the record without becoming its own ticket. That leaves the work with a clear connection to the choice it should change.

The 5 Whys guide can help follow the clues behind a decision. The guide to responsibility and accountability covers who has the authority to act on what the review finds.

Go back to those two identical entries. The next responder should be able to point to the test worker and say which account and region make it the right one before touching either entry. If they can do that in the exercise, the review has shown exactly what changed for the person facing that screen.

Sources & context

Sources linked in this article. Read alongside the author’s analysis; a citation does not independently verify a publisher’s claims.

Report an error or outdated detail

Keep reading

Where to go next