Engineering teams rarely lack data about their systems. Dashboards are plentiful; a standing hour for deciding what they mean is not. A weekly operations review supplies that moment. This list sets out eight items worth covering in a compact session, ordered so that the team looks back first, clears what is stuck and finishes by choosing next week’s few priorities. Treat the list as a menu: a team with a calm week can skim half of it, and a team with a rough week can linger on the first few.
1. Last week’s incidents, in one sentence each
Begin with the events that interrupted normal service. Give each a single sentence: what broke, how long it affected users, and whether the fix is permanent or a patch. The discipline of one sentence prevents the review from becoming a series of mini post-mortems. If an incident needs a deeper conversation, note it and schedule it separately.
2. Follow-ups still owed from earlier incidents
Every incident leaves promises behind: a missing alert to add, a dependency to upgrade, a document to correct. Check that each promise has an owner and a date. An action item older than a few weeks without movement should be either escalated or closed with an honest note that it will not be done.
3. The ticket queue: what is aging
Sort open operational tickets by age. The goal is not to clear the queue in the meeting but to notice which items have been waiting longest and ask why. Often an old ticket is blocked on a decision, not on effort, and naming the blocker is the whole job. A useful habit is to ask of each aged ticket whether it should be worked, delegated, or retired, and to accept only one of those three answers.
4. Flaky tests and unstable builds
Name the tests that failed intermittently and the builds that needed a rerun. A concrete example: a test that passes on a second attempt about half the time quietly teaches developers to ignore red builds. List the worst offenders, assign one person to each, and decide whether any should be quarantined until repaired. Set a short deadline on that quarantine, otherwise a disabled test becomes a forgotten one, and the gap in coverage grows unnoticed.
5. Changes that shipped, and anything surprising about them
Scan what was deployed. Did any release need a rollback, a hotfix or an unplanned manual step? Surprises here point at weak spots in the delivery process. Record them while they are fresh, because in a month nobody will remember which release caused the late night. If a pattern emerges, such as every Friday deploy needing attention, that pattern is itself a finding worth a ticket.
6. Capacity and cost, glanced at not studied
Take a brief look at storage growth, queue backlogs or cloud spend, whichever your team tracks. You are looking for a trend that deserves a proper investigation, not performing the investigation. If something looks off, create a ticket with a clear owner.
7. Maintenance coming up
Look at the next week’s calendar for planned work: certificate renewals, upgrades, scheduled maintenance windows, on-call handovers. Teams that read the calendar together avoid the classic collision of two risky changes landing on the same day. A teammate planning a database upgrade can hear in the room that another group has a launch the same afternoon, and shift the work by a day before it matters.
8. Next week’s few priorities
End by choosing two or three outcomes that matter most, and say them aloud. They should be specific enough that on Friday you can say yes or no to whether each happened. Everything else is secondary. For teams exploring how automated analysis can help surface the right incidents to discuss, this overview of speeding up IT problem resolution with modern AIOps platforms is worth a read before you design your agenda.
Running the list in practice
Do not try to cover every item at full depth. A workable split for a half-hour session is roughly this: the first quarter on incidents and follow-ups, the next on tickets and tests, a short stretch on shipped changes and capacity, and the final quarter on upcoming maintenance and priorities. A timekeeper helps, and so does a shared document that is pre-populated with ticket lists before the call. Rotate the facilitator every few weeks so the routine is shared across the group instead of resting on one diligent individual.
If your team is newer to this kind of discipline, structured training can shorten the learning curve. The piece on training that prepares people for modern IT operations gives a sense of the skills involved.
A note on beyond-engineering parallels
Non-technical groups run the same sequence with different content. The HR-focused web application DayOneFlow, as its site describes it, helps a people team follow each new hire through the stages from taking an offer to being confirmed as staff. Replace tickets with candidates, and the weekly look at what is aging and what is next reads almost identically.
Signs your review is working
- People arrive having already read the pre-filled list.
- The same incident does not appear in two consecutive weeks without an explanation.
- The priority list is short and the team can recite it on Wednesday.
- Old action items are being closed, not rolled over indefinitely.
Keep the list on a single page, trim items that stop producing decisions, and add new ones only when the team repeatedly runs into a gap. A weekly review is a tool for steering, and the best versions stay small enough that nobody dreads the slot.