Most small businesses discover they need an IncidentOps routine on the worst possible afternoon: the checkout page is down, three people are typing in three different chats, and a customer has just emailed to ask whether you are still in business. What follows is a self-check that fits into one hour, a coffee and a shared document. Tick what you already do, jot down what you do not, and pick two gaps to close this month. Nothing here needs a big budget, and you do not need a dedicated operations department.
A short definition first, so the rest makes sense. IncidentOps is the practice of responding to an unplanned disruption in a coordinated way: deciding that it is an incident, putting someone in charge, keeping people informed, fixing or softening the problem, and learning afterwards. The Ops Atlas notes that the label is informal and mostly overlaps with what many people call incident management, so do not worry if your team uses a different name.
Part one: before anything breaks
Do you have a written definition of “incident”?
Sit down and write three or four plain sentences that say when a problem becomes an incident. For example: customers can see it, more than one team is needed, or nobody has made headway after an hour. Put it somewhere everyone can find it. Written criteria take the personal risk out of raising the alarm, because nobody is stretching their own judgement, they are just following the page.
Is there a simple severity scale?
Two or three levels are plenty. Level one might mean customers cannot buy or log in, level two might mean something important is degraded, level three might mean an annoyance that can wait for morning. Add one line for each level that says who gets woken up and how quickly updates must go out.
Does a real person own the pager each week?
An on-call rota that exists only in someone’s head will fail the first holiday weekend. Write the schedule down, include a backup, and agree how handovers happen. If you only have a handful of people, do a quick sum of how often each person would be on call. The Atlas offers a sobering example in which six engineers covering two slots every week of the year end up carrying roughly a third of the load each, which is more than a commonly cited 25 percent ceiling. If your arithmetic looks similar, you need more people in the rota or fewer slots to cover.
Have you printed role cards?
Make a half-page card for each hat: the commander who coordinates but does not touch the system, the person who actually changes things, and the person who writes updates. Keep them short enough to read while stressed. Remember that these are hats, not job titles, so anyone trained can wear one.
Part two: while it is happening
Is there a place where customers can see what is going on?
A hosted status page that does not depend on your own website costs very little compared with what it saves you. The point is simple: if your main site is down, the page telling people it is down must still load. Choose a service on separate infrastructure and test it once from a phone on mobile data. Many tools in this space use four states: investigating, identified, monitoring and resolved. Borrow them as they are.
Who writes updates, and how often?
Pick one communicator per incident. A workable rhythm drawn from public playbooks is to post within about five minutes of declaring, follow up with scope shortly after, and then post again at least every twenty minutes for the first couple of hours. When it is over, say what happened and whether any customer data was lost, in plain language. If you are looking at how this feeds into the rest of your online presence, our overview of how web, cloud and automation pieces fit together for growing businesses covers the foundations that make quick communication easier.
Do you fix first and explain later?
Ask your team what they would do if the latest release looked suspicious. If the honest answer is “investigate until we find the root cause”, add a note to your playbook: roll back, switch off the new feature, or fail over first. These generic moves need only a rough idea of where the problem sits. You can dig for the deeper cause once customers are happy again.
Is the channel tidy?
Open one working channel for the responders and a separate one for read-only updates. Without that split, curious colleagues pile into the working space, and the core team ends up taking the real conversation into private messages, which defeats the purpose.
Do you say the handover out loud?
If someone has been in charge for hours and needs to rest, the next person should hear a plain spoken line, something like “you hold command from this moment, agreed?”, and reply with a yes. It sounds theatrical until the first time it prevents two people from issuing contradictory instructions.
Part three: after the dust settles
Is a review owner named before the call ends?
Someone should leave the incident with the job of running the review, and a date attached. Some published practices set a limit of three calendar days for the most serious incidents and five business days for the next tier. Choose figures you can honestly meet. A review that happens three weeks later mostly captures regrets rather than facts.
Is the action list short and owned?
Pick a few actions, each with one named owner and a due date. Avoid vague items such as “improve monitoring” and avoid actions that are really just mitigation in disguise. One well-known company tracks its priority fixes against a target of four or eight weeks depending on the service. Whatever number you choose, review the open list at a regular meeting, otherwise it becomes a graveyard.
Do you stop at conditions instead of people?
“Human error” is rarely the end of the story. If a person made a slip, ask what let the slip reach customers. Maybe a checklist was missing, maybe an approval step was too easy to skip. Your goal is to improve the setup, and your team will report problems sooner if they know declaring an incident will not be held against them.
Do you measure carefully?
If you track “average time to resolve”, be wary. The Atlas points out that this single phrase carries several meanings and that a lone dramatic outage can distort an average. Look at the spread, keep notes on how long it took to notice and to mitigate, and watch how often your on-call people are paged, since that is a good proxy for burnout. Our companion read on spotting small leaks before they drain your runway makes a similar point about money: one big number hides the trend, so watch the pattern.
What your tick count says
Add up the questions you could answer yes to without squirming. Under six means you are in the heroics stage, where outcomes depend on whoever happens to be awake. Between six and eleven you have the bones of a defined process. Above eleven you are probably coordinated and just need practice. The most useful single step is a tabletop exercise: invent a plausible failure, gather the people, and talk through who does what. It takes an hour and reveals gaps no document can.
For a wider background on the terms and the sources behind these suggestions, you can read a detailed IncidentOps reference compiled in the Ops Atlas. Think of it as a menu, not a mandate: adopt the bits that fit a team of your size, and revisit this checklist every quarter.

