Nuage Solutions All articles
Cloud Strategy

When the Status Page Turns Red: A Practical Outage Rehearsal for UK SMEs

Nuage Solutions
When the Status Page Turns Red: A Practical Outage Rehearsal for UK SMEs

Photo: business team tabletop exercise meeting room whiteboard incident planning, via createwithlakshya.com

A cloud outage rarely arrives as lost data. It arrives as a room full of capable people asking each other what they are supposed to be doing. The files are still there, the tenant is intact, and the provider's status page carries an amber banner reading "investigating". What has disappeared, temporarily, is the ability of your business to make decisions in a fixed order.

That is a coordination problem, and coordination problems are the cheapest thing in IT to rehearse. You do not need a standby data centre or a consultant on retainer. You need a meeting room, six people, a clock and a scenario slightly worse than you hope for.

Why a tabletop, and why short

Full technical recovery tests have their place, but they test systems. A tabletop tests the people around the systems: who rings the supplier, who decides to send staff home, who speaks to customers, and who notices that the fallback authentication path was a mobile phone in somebody's desk drawer.

Ninety minutes is deliberate. Anything longer becomes a workshop; anything shorter never reaches the awkward parts. The aim is not to finish the scenario but to surface three or four assumptions that turn out to be untrue.

Setting it up

Choose one scenario and resist the urge to make it a composite. "Our Microsoft 365 tenant is unavailable for four hours from 09:30 on a Tuesday" is a far better exercise than "something goes wrong with the cloud". Specificity forces specific answers.

Invite six to eight people: a director with authority to commit money, whoever runs IT (internal staff or your managed service provider), the person who owns customer relationships, the office or operations manager, and a note-taker who is not part of the discussion. Ask attendees to bring only their phones and a printed contact list — the printed list is itself part of the exercise, because you will discover that nobody has one.

Appoint a facilitator to read the injects and keep time. The facilitator should not be your IT lead; the IT lead needs to be a player, not the referee.

The four seats

Write these on the whiteboard before you begin.

Incident lead. One named person with the authority to pause normal work. The most common failure in a real outage is diffused responsibility: everyone is vaguely in charge, so nobody is.

Technical lead. Owns diagnosis, supplier escalation and verification of recovery. Does not handle communications.

Communications lead. Owns what staff, customers and — where personal data may be implicated — the Information Commissioner's Office are told, and when. The 72-hour notification window is a live consideration in any incident that might become a personal data breach, and rehearsing who makes that judgement is valuable in itself.

Scribe. Records timestamps and decisions. Without timestamps, the debrief dissolves into anecdote.

Minutes 0 to 15: the alarm

Read the first inject aloud: staff cannot sign in, the service health dashboard shows a regional issue, and the supplier's status page has read "investigating" for forty minutes.

Ask the room three questions and write the answers up: who has just been told, by what means, and what are staff doing right now? The usual answer — that people are refreshing the page and messaging one another on the very platform that is down — is the first finding of the day.

Minutes 15 to 35: escalation

A support ticket is not an escalation. Walk the room through the actual ladder: the status feed, the support portal and the severity definitions that determine response times, the named account or technical contact, and the contractual service level with its credit mechanism. Then ask the harder question: who else in your supply chain is caught by the same event? If identity, email, file storage and telephony all sit behind one vendor, you have one system, not four.

Record every gap. "We would ring our account manager, but she left in March" is precisely the kind of finding this exercise exists to produce.

Minutes 35 to 55: staff and customers

Inject: the outage has passed ninety minutes and a customer has emailed to ask why an order has not been acknowledged.

Rehearse an out-of-band channel now, while it costs nothing. In most SMEs that means a staff text or WhatsApp broadcast group, a recorded message on the main number, and a holding statement on a website hosted somewhere other than the affected platform. Customer-facing teams need one sentence they are permitted to say, and a named person to approve updates.

Minutes 55 to 70: working by hand

The question here is not "can we work offline" but "which three processes must keep moving, and what do they require?" Typically: order intake, scheduling or dispatch, and anything touching money.

Identify the manual forms, the locally held spreadsheets, the card terminal's offline mode and its limits, and who may approve spend without the usual system checks. Then test the dull dependency: if printing requires a server login, or the card terminal sits on the same Wi-Fi as everything else, the workaround has a hole in it.

Minutes 70 to 85: recovery and reconciliation

Inject: service restored at 13:20.

Recovery is where disorder quietly returns. Ask what will be re-keyed, by whom, and how duplicates will be spotted. List the specific checks: queued email released in a burst, documents edited locally needing reconciliation, paper records still awaiting entry, and any commitment made verbally with no audit trail behind it.

Agree who signs off "we are back to normal" and what evidence they require before doing so. That signature is a control, not a formality.

Minutes 85 to 90: scorecard and commitments

Score each area from 1 (we improvised) to 5 (documented and rehearsed), and capture the single worst moment in each.

Area Score Worst moment observed Owner By when
Roles and authority
Supplier escalation
Staff communications
Customer communications
Manual workarounds
Data reconciliation
Decisions under pressure

Close by agreeing no more than five actions, each with a name and a date attached. A long action list is a short-lived action list.

From exercise to plan

Sort the actions into three buckets. Within thirty days: printed contact lists, a broadcast channel, a named incident lead and deputy, and a corrected escalation ladder with current names. Within sixty days: a one-page playbook and written manual procedures for the three critical processes. Within ninety days: run the tabletop again with a different scenario — a cyber incident, a telephony failure or the loss of a key supplier — and see whether the scores moved.

Repeat annually at minimum, and whenever the estate changes materially. The exercise costs ninety minutes and a plate of sandwiches. The alternative costs whatever a disorganised Tuesday costs your business, multiplied by however long the provider takes.

All Articles

Related Articles

Cheap Cloud Storage Has a Long Invoice: A UK SME Cost-Control Guide

Cheap Cloud Storage Has a Long Invoice: A UK SME Cost-Control Guide

Repatriation Without the Slogan: A 2026 Workload Placement Framework for UK Organisations

Repatriation Without the Slogan: A 2026 Workload Placement Framework for UK Organisations

The Second Line Is Not a Spare: Testing Internet Failover Before You Need It

The Second Line Is Not a Spare: Testing Internet Failover Before You Need It