An incident plan can look wonderfully complete until someone asks a small question: who can authorize the next action when the usual owner is unavailable?
You do not need to break production to discover that gap. Put the right people in a room, give them a plausible situation, and ask them to explain what they would do with the information available. Then change the information and see whether the response still makes sense.
That is the useful promise of an incident-response tabletop. NIST describes a tabletop exercise as a facilitated discussion of participants' responsibilities and responses to a scenario. It is a good way to inspect coordination, assumptions, and decision authority. It cannot prove that a recovery command works or that a service survives a failed dependency.
Keep that distinction visible. A meeting that produces better questions can be valuable. Calling it a successful failover test would turn useful learning into misleading evidence.
Choose a decision worth rehearsing
Start with a response problem, rather than an impressive disaster. “The entire region is gone and everyone is on vacation” offers plenty of drama but often produces vague answers. A smaller scenario can expose a specific operational weakness.
Useful questions include:
- Who coordinates when two teams see different symptoms of the same problem?
- What evidence is needed before a risky mitigation is proposed?
- How does support receive an update when impact remains uncertain?
- Who accepts an incident handoff, and what must be transferred?
Pick one primary question and perhaps one supporting question. If you try to cover detection, security containment, customer messaging, recovery, and executive coordination in forty minutes, the result will probably be a tour of headings.
Choose participants who can explain the relevant work. A service responder, incident coordinator, and support representative may be enough. If an approval is central to the exercise, include someone who understands that approval path or explicitly simulate the absent role. Do not silently let the facilitator invent organizational authority.
Use existing response documents as inputs. The incident-handoff guide can help define what a receiving responder needs. The exercise asks whether participants can apply that guidance under uncertainty; it should not become a read-through of every paragraph.
Define observable learning goals
“Improve readiness” is a reasonable aspiration and a poor observation criterion. Decide what someone can actually notice during the session.
For a coordination exercise, goals might be:
- Participants name a coordinator and distinguish that role from technical investigation.
- Each proposed mitigation includes its expected effect, required authority, and stopping condition.
- The group separates confirmed impact from an unverified hypothesis in a simulated update.
- A handoff names the receiving owner, unresolved questions, and the next update time.
These are proposed goals for this exercise, not industry performance thresholds. You can observe whether the group did them and what made them difficult. Avoid ranking people by how quickly they sound confident. The person who asks for missing evidence may be doing the most useful work.
Write the exclusions alongside the goals. This session does not test paging delivery, production credentials, database recovery, or actual customer communications. Those require different verification, authorization, and safeguards. Recording the exclusions prevents a later status report from claiming more than the exercise demonstrated.
Make the discussion-only boundary operational
Put “EXERCISE ONLY” on the invitation, scenario, notes, and simulated messages. Use a dedicated meeting and document. Do not create a fake incident in the real paging system or post realistic customer-impact messages to operational channels.
Participants describe actions; nobody executes them. No production commands, configuration changes, real pages, customer emails, or vendor tickets belong in this tabletop. A proposed dashboard check can be answered with a prepared screenshot or a written observation from the facilitator. Use synthetic data, with no customer identifiers or secrets.
For the first session, keep terminals closed. It is surprisingly easy for “let me just check something” to turn a coordination rehearsal into an uncontrolled investigation. Capture requests that cannot be answered from the packet as evidence gaps. Schedule any live verification separately through the normal process.
Agree on a stop phrase, such as “pause exercise.” Anyone can use it when simulated and real activity become confused. A real operational issue takes priority: end or pause the exercise, identify that the discussion has stopped, and follow the actual incident process. Never fold the real event into the fictional story.
Assign facilitation separately from response
The facilitator presents information, manages time, and asks clarifying questions. The facilitator should not also be the incident coordinator whose choices the exercise is intended to examine.
Assign these responsibilities explicitly:
- Facilitator: holds the scenario and information-release schedule, maintains boundaries, and avoids supplying the answer.
- Responders: explain investigations, hypotheses, proposed actions, and uncertainty from their normal roles.
- Coordinator: organizes the simulated response and makes ownership and next decisions explicit.
- Observer or scribe: records decisions, requested evidence, ambiguous ownership, and useful quotes without grading personalities.
A small group can combine observer and scribe. Combining facilitator and coordinator is much less useful because it hides whether participants can establish coordination themselves.
CISA's exercise-planner handbook treats tabletop exercises as discussion-based work on roles, plans, and coordination. Its materials provide a starting structure; the smaller software-service exercise below is an editorial example, not a claim that this meeting satisfies a formal exercise program.
Walk through one fictional service incident
Suppose a fictional subscription service uses an external billing provider. Some account changes time out, while ordinary page views remain healthy. A provider status page says service is operational. The exercise concerns coordination and communication, not implementation of retries or billing recovery.
Give everyone a short initial packet: a synthetic service diagram, role directory, response-plan excerpt, and the opening symptom. Keep the underlying cause in the facilitator's notes. Participants should not need to guess the facilitator's favorite diagnosis to make sensible decisions.
Use elapsed exercise time for the following releases. These timings are an example agenda, not response targets:
| Release | New information | Decision to observe |
|---|---|---|
| Opening | Account-change timeouts have increased; impact is incomplete | Who coordinates, and what evidence is requested? |
| After 8 minutes | Support reports customers retrying; payment outcomes are unknown | How does the group communicate uncertainty and bound risky proposals? |
| After 16 minutes | The provider acknowledges a delay; the usual service owner is unavailable | Who can approve a mitigation or find the authorized substitute? |
| After 24 minutes | A replacement responder joins; symptoms appear to improve | What transfers, and what evidence would justify ending the response? |
For the second release, a participant might propose asking customers to retry. The facilitator can ask, “What do you know about the previous requests' outcomes?” That question exposes an assumption without prescribing an answer. A sensible response may be to request a prepared reconciliation summary and draft a holding update.
If participants request information outside the packet, record the request. Either provide a prepared answer or say that the evidence is unavailable. Do not improvise convenient facts that rescue every decision. The uncertainty is part of the exercise.
Also avoid a hidden victory condition such as identifying one exact component. A group can choose a defensible response while holding an incomplete diagnosis. Observe whether it names the uncertainty, preserves ownership, and requests information that could change its decision.
Use a facilitator sheet you can reuse
Keep the working sheet short enough to consult while listening. A blank template is more useful than a long deck nobody updates:
EXERCISE ONLY — discussion; no live actions
Scenario / version:
Date / duration / participants:
Facilitator / observer:
Primary learning question:
Observable goals:
Explicit exclusions:
Allowed reference materials:
Stop phrase / real-incident interruption rule:
Opening information:
Release 1: elapsed time / new fact / question
Release 2: elapsed time / new fact / question
Release 3: elapsed time / new fact / question
For each consequential decision:
Available evidence:
Unknowns / requested evidence:
Proposed action / intended effect:
Authority / named owner:
Stop or reassessment condition:
Simulated update / next update time:
Debrief:
What helped:
What was ambiguous:
Follow-up / owner / due date:
How the follow-up will be verified:
What this exercise did not establish:
Keep fictional scenario information separate from observations about the real organization. “The scenario says billing is delayed” and “we could not find the approval path” are different kinds of statements. That distinction matters when someone reads the notes months later.
Use the incident-timeline guide for the habit of preserving decision evidence, while labeling every entry as exercise material. You are rehearsing the recording practice, not creating a historical record of an outage that never happened.
Debrief while the friction is still specific
Reserve the final fifteen minutes of a one-hour session for discussion. Stop releasing new facts in time to protect that space. Ask participants what surprised them, what helped, and where they had to invent an answer.
Review behavior against the learning goals. “Nobody knew who could approve the proposal after the service owner became unavailable” is actionable. “Communication was bad” is an opinion without a repair path. A useful facilitator asks which missing contact, unclear document, or conflicting expectation created the problem.
Choose a small number of follow-ups. Give each a named owner, a date, and a verification method. Updating a role directory might be verified by having a different responder locate the substitute approval path in a later exercise. A proposed technical mitigation needs its own separately authorized test; a good conversation cannot validate its effects.
The postmortem-action verification guide applies the same ownership discipline after real incidents. Here, keep the origin clear: these are improvements prompted by rehearsal.
Repeat the uncertain part
The next tabletop should revisit the repaired weakness with a slightly different situation. If the first exercise exposed unclear approval authority, changing the fictional provider name accomplishes little. Change which owner is unavailable or which teams need to coordinate, then see whether the revised guidance still helps.
Report the result narrowly: the group rehearsed these decisions, found these gaps, and verified these document or coordination changes. Leave resilience and recovery-performance claims to evidence that actually tests those properties.
A good tabletop does not need a spectacular failure. It needs a plausible decision, enough uncertainty to make assumptions visible, and someone who will repair what the discussion exposes. More practical operations guidance is available on Slaptijack.