Somewhere in every operations team there is a story about a name. A scheduled job called “cleanup” that deleted more than anyone expected. A chat channel called “urgent” that stopped being urgent two years ago. A service everyone calls “the old thing” because nobody remembers its real name. These stories get told with a laugh, and then the same naming habits carry on. This FAQ collects the questions teams actually ask when they decide to do better, with answers rooted in the way incidents unfold on a real shift.
Why does a name matter so much during an incident?
Because an incident compresses time. In calm moments you can ask around, open three tabs and compare. In the first ten minutes of an outage, every extra question is a cost, and unclear names generate questions. A responder who reads “Orders API: checkout errors rising” knows which system to look at and which symptom to chase. A responder who reads “Alert 31” has to look it up before doing anything useful.
Think of the name as a label on a fire extinguisher. Nobody reads the manual while the room is smoking; they read the label.
What should an incident title contain?
Two things and, when there is room, a third: the affected system, the visible symptom, and optionally the scope. “Customer login: timeouts for some users” works. “Login problem” is the minimum, and “URGENT!!!” is not a title at all. Avoid putting the guessed cause in the title, because early guesses are frequently wrong, and a title that says “database issue” may mislead the next hour of work. Update the title when the cause is confirmed.
How should we name the files and notes that come out of an incident?
Use the same pattern every time: system, type of document, then a plain description. “checkout-api-review-timeout-after-config-change” sorts with its siblings and says what happened. Avoid putting only a date or a ticket number in the name, because neither tells a future reader whether the document is worth opening. A date is useful as a secondary detail, written year first, but it should never be the entire label.
Is it worth renaming old services, or should we live with the history?
It depends on how often the old name causes confusion. A good rule is to rename anything that appears in an incident discussion more than once a quarter. Rare names can keep their quirks. For services still in use, rename carefully: add the new name, keep the old as an alias in documentation, and retire the alias when nobody mentions it any more.
One operations lead described the first rename on their team as feeling like moving house, yet admitted that three months later nobody could remember why it had been hard.
How do we name scheduled jobs?
Name them for what they do and what they touch: “nightly-invoice-export” or “weekly-log-archive-storage”. Avoid vague verbs like “cleanup” or “sync” on their own, because they hide the scope. When a job misbehaves at 3 a.m., a name that states the target tells the responder what is at stake before the first log line is read.
What about chat channels and mailing lists?
Channels collect names faster than anything else. Use a short prefix pattern that tells people the purpose: “inc-” for live incidents, “team-” for standing groups, “proj-” for temporary work. Archive channels when the work ends and keep their names unchanged so searching still finds them. A channel called “general-2” signals that a decision was skipped somewhere.
Do automated tools change the picture?
They raise the stakes without anyone noticing. Tools that group, route and summarise events work from the labels you give them, so poor names are amplified rather than fixed. Reading about the way modern tooling supports responders, such as a guide to incident handling with AIOps techniques, makes the point plain: automation helps most where names and tags are consistent. The step-by-step thinking in the hands-on AIOps route map written for DevOps, reliability and operations groups assumes that systems can be identified reliably, which is a naming job before it is a tooling job.
How do we get people to follow a convention?
Make the easy path the correct one. Put the pattern in the template that creates the ticket or the channel, so people fill in blanks instead of inventing. Keep the convention short enough to remember. Praise good examples publicly, and fix poor names quietly and without blame, because embarrassment teaches people to hide, not to improve.
Another team shared a trick that worked: each month they chose one confusing name from the previous incidents and renamed it as a group, five minutes at the end of a meeting. It turned naming into a small ritual rather than a chore.
Where does the outside world come into this?
A fair number of operations people also care about the community around them, such as local gatherings and talks. A regional listing site like GujaratOrbit, a Gujarat events website whose pages cover listings, venues, a forum and a blog, is a handy bookmark for anyone in Gujarat, and the way it separates its sections is a small example of naming that helps a first-time visitor.
What does a good before-and-after look like?
A small support team shared theirs. Before: a shared mailbox rule called “sort”, a dashboard folder called “misc”, and a pager schedule called “rota-new”. After: “route-billing-emails-to-billing-queue”, “payments-service-views” and “weekday-primary-oncall”. Nothing about the underlying tooling changed. The difference showed up the first time a stand-in covered the schedule and did not need to phone anyone to learn what the three things were. The team’s only regret was waiting a year to do it.
Can we name things in two languages?
Many teams work across languages, and that is fine as long as one name leads. Choose a primary label that everyone can type and search, then add the local-language term as a secondary note in the description. Mixing both into one name tends to fail in search tools and sorting. A clear primary name plus a documented alias keeps colleagues in different offices reading the same thing.
Who should own the naming rules?
Give the convention a home with a named role, not a named person, such as the platform team lead, and let anybody propose changes through the same review process as code. A rule without an owner drifts, and a rule owned by a committee freezes. Keep the page short, revisit it twice a year, and delete rules nobody follows rather than pretending they apply.
What is the smallest first step?
Pick the five names that appear most often in your incident channel and ask whether a new joiner could explain each one. Rename the worst. Write the pattern on a single page. Add it to the incident template. That is an afternoon of work, and the next night shift will feel the difference. It also costs nothing beyond time, and it needs no approval from anyone outside the team.
Is there a risk of over-engineering this?
Yes. A naming scheme with a dozen rules and a committee to approve exceptions will be ignored. The aim is not perfect taxonomy; the aim is that someone worn out after a long shift can read a label once and guess right. If your convention achieves that, it is enough.

