A plan for a difficult day: rehearse a service before it stops working
How to rehearse decisions, information and handovers during a fictional service disruption.
Article language: English

In this article
On paper the plan looks calm: contact the account owner, use a backup and inform the client. Then someone asks which copy to use and who can share it. You do not have to wait for a real interruption to discover that gap. Walk through a selected scenario together using the arrangements you actually have.
This exercise checks decisions and coordination within a limited part of a service. It does not require staging an attack or switching off a production system. Participants receive a situation and explain what they would do, on what basis and with whom. The facilitator distinguishes a declared possibility from an action that has actually been checked.
Which part of the service do you want to examine?
Start with something the company has promised a client. In a fictional scenario, a small team must deliver an approved set of materials before the client's meeting. In the morning the main repository is unavailable. We are not assuming a breach or data loss: the people preparing delivery cannot retrieve the required version and the meeting is still scheduled.
A useful objective is to determine who confirms a substitute set, who decides how it may be delivered and what the client will hear before the cause is known. Trying to examine all technology, cover arrangements, finances and reputation at once may distract the group before its first decision.
FEMA distinguishes discussion-based from operations-based exercises. In a tabletop, participants discuss a hypothetical scenario which can develop through prepared messages. Talking through an action is not the same as performing it.
Collect the current plan, contacts and delivery arrangements beforehand. If no working agreement exists, prepare one first. Participants should not have to guess what procedure was supposed to exist. Record a missing plan as a gap, recognising that it describes readiness to exercise rather than performance under an established procedure.
A premortem guide in Polish (in Polish) looks for possible reasons a project might fail. Here the team receives a particular disruption and works through successive decisions. A website outage response guide in Polish (in Polish) can inform a similar discussion, but this scenario concerns delivering a service result, not repairing a site.
Who should be at the table?
Include people who know the selected workflow and can explain decisions within their authority. In the example they are the client contact, the person preparing the materials and someone authorised to approve an alternative delivery. Managers who know the formal chart may not know where the team actually finds current files.
The NIST glossary defines a tabletop in the context of people with roles in a particular IT plan discussing a scenario and facilitator questions. We use a similar conversational form for this operational exercise; it is not a technical system recovery test.
Appoint a facilitator and a note-taker. The facilitator presents information and asks about decisions instead of immediately supplying the right answer. The note-taker records the basis for a choice and missing agreements, not every sentence. A small team can combine roles if notes do not disappear under the work of facilitating.
Mark materials as exercise material. Agree that participants will not contact clients, change permissions or perform actions in production. If an actual check is needed, record a separate task with appropriate scope and authorisation. A call to a real supplier should not become an unexpected part of the simulation.
Agree how to stop if a real event occurs during the meeting. Ordinary client needs take priority according to the organisation's arrangements. Stopping is not a failure. No one should mistake a fictional outage message for a real incident requiring a response.
Develop the scenario without adding a catastrophe every minute
The first message gives only facts available at that point: the repository does not open for the people preparing delivery, the meeting time is known and the cause is unknown. Ask who gathers the picture and what they tell others. Do not immediately suggest that the file can be found in someone's inbox.
Next, one person has a local copy but does not know whether it contains the last correction. Ask how the team will confirm the version and who may decide to use it. “Someone will remember” is a dependency on memory, not verification of the file.
A third message can introduce a change of contact: the account owner must hand the case to a substitute before the next update. The guide to handovers and ongoing cases in Polish (in Polish) addresses the information a substitute needs: the state, decisions and unknowns. Adding a person to a long thread does not prove they understand what has been agreed.

Do not add a lost phone, another outage and changing requirements just because the team handled the first question. Each new element should serve the objective. If delivery is the focus, keep attention on version, authority and communication. A growing list of surprises can turn the meeting into a test of the facilitator's imagination.
The client eventually asks whether they can prepare their meeting using a previously received set. The team must separate what it knows about that version from assumptions. It can promise to check and state when it will provide the next update. It need not invent a cause or repair deadline. The response should help the client decide what to do.
If a partner performs part of the work, consider the Polish guide to a white-label service workflow (in Polish). Who can confirm the version prepared by an external studio, and can the client contact obtain that confirmation? A partnership alone does not guarantee that the right person is available during a disruption.
Record more than “we would manage”
Separate confirmed arrangements, declarations requiring a check and gaps. If someone identifies an existing instruction, note its location and version. If someone says they could restore a file, record that this was not done in the discussion. If no one knows who approves a substitute, do not quietly fill in the most likely name.
The NYC Emergency Management guide, in the preserved draft version 1.0 dated 2019, describes reviewing performance against plans and identifying follow-up without assigning individual blame. Our discussion is neither certification nor proof of compliance with industry requirements.
Connect each observation to a decision. “No current copy” goes too far if availability was not checked. A more precise note is: “the team did not identify how to confirm whether the local set includes the approved correction.” That leads to a check without pretending a technical test was performed.
The meeting might end with an owner confirming where the approved set is held, the client contact refining an update, and the substitute checking whether they can find the case status. Each action needs a deadline and a way to confirm completion. Writing down a name does not by itself create a working alternative.
Do not reward speed alone. A person who immediately proposes sending an arbitrary copy may speak faster than someone who notices missing approval. Look for clear decisions and appropriate limits, not the shortest response time.
Check whether an answer depends on an absent person. “The administrator would share it” is a hypothesis about another person's ability and availability. Record a question for them and the limit of the current agreement. Do not act out their consent when the scenario does not provide it.
Discuss the return to normal work as well. Suppose materials were sent by an agreed alternative route and the client requested a correction there. Where will that decision go after access is restored? Who will recognise whether it concerns a new version? Handling the disruption includes information created outside the main tool.
A short observation card can record:
- The decision and information available at that moment.
- The chosen action and the person authorised to take it.
- The basis: an identified record, a declaration or missing information.
- The effect on the client and the proposed message.
- The open question, owner and evidence needed to close it.
Do not rewrite the notes afterwards as though better answers had been given during the exercise. Add improvements separately and preserve the original record. Store it according to its contents: operational contacts, access gaps and dependencies need not be distributed widely. A broader summary may only need agreed actions and owners.
Return to work after the exercise
Pass decisions to the people making changes and check that they understand the scope. A revised version label must also work in the next ordinary handover. A named substitute needs appropriate access and knowledge; a new document does not automatically provide either.
The Polish article on handling one enquiry in shared context (in Polish) considers communication among people. Distinguish tools that store materials from those that record decisions or merely send notifications. Two channels relying on the same unavailable resource may not form independent routes.
Repeat the selected part after a meaningful change or a change in conditions. Keep the earlier observation so the team can see what was actually improved. The aim is not a more dramatic catastrophe: it may be enough that the team can identify the right version, person and next decision without relying on memory.








