The demand-surge playbook — when the world sends everyone to your queue

Operations · ~7 minute read

A regulator makes an announcement at 09:30. A product recall hits the evening news. Something about your brand goes viral overnight. By morning your forecast is fiction and your queue is a multiple of anything the plan imagined. External demand surges are not forecastable individually — but they are entirely playbook-able as a class.

Detection — before the queue tells you

The queue is the slowest possible way to discover a surge. By the time service level collapses, the event is an hour old and your options have already narrowed. Faster signals exist and the playbook names them: a media-monitoring alert on the brand and category, the regulator’s publication schedule actually watched, website traffic and search spikes that precede calls by thirty to sixty minutes, social mentions accelerating, and the corporate comms team plugged in — because nothing is more absurd than the press office announcing something the contact centre learns about from its own customers.

The diagnosis check matters as much as the signal: is this a genuine external surge or an ordinary blip wearing dramatic clothes? The tells of a real surge are a single dominant contact reason, cumulative divergence that keeps widening rather than self-correcting, and an identifiable external trigger. Confirm the driver before acting — the response to a recall is not the response to a Tuesday that started badly, and misdiagnosis wastes the only cheap hour you will get.

Surge triage — reforecast and declare

Once confirmed, the first hour is triage. Reforecast the day against the new reality, using whatever early read you have on volume and the surge contact type’s AHT — surge contacts often run long, because they are anxious, novel, and not in the script. Declare the stance: which queues you defend, which you consciously let slip, and at what point this stops being a hard day and becomes a declared incident with a command structure. The thresholds for that declaration were written in peacetime; the surge morning just reads them.

The characteristic failure is incremental denial: treat hour one as a blip, hour two as probably peaking, hour three as surely peaking soon — and by hour four the backlog guarantees a bad week instead of a bad day. Surges from external events do not obey your intraday curve; they run until the underlying cause is resolved or the worried population is served. Plan pessimistically and be pleasantly wrong. The cost of over-reacting to a surge is some overtime; the cost of under-reacting compounds daily.

Deflection that doesn’t damage

The strongest surge lever is answering the question before it becomes a contact. Most surge demand is a single question asked thousands of times — am I affected, what do I do, when will I know? A clear page on the website, an IVR message that genuinely answers rather than merely delays, an outbound message to the affected population — each can remove a quarter of the queue at almost no cost. The discipline is to publish the answer where customers are already looking, within the first hour, in plain language, and updated as the situation moves.

But deflection has a failure mode the playbook must name: deflection that blocks rather than answers. An IVR that loops worried customers back to a website that does not answer their case does not reduce demand — it stores it, angrier, for tomorrow, and it abandons exactly the people who most need a human. The rule is simple: deflect the question, never the customer; keep a clean route to an agent for anyone whose case is not covered; and protect the vulnerable-customer pathway absolutely, because a surge is precisely when distressed and vulnerable callers are most likely to be in the queue.

The staffing levers, in order

With demand reduced as far as honesty allows, work the supply levers from lightest to heaviest. Hold discretionary shrinkage — re-time (never cancel) training, coaching and non-urgent offline work out of the peak. Flex breaks and lunches around the worst intervals, through the team leaders, with the reason attached. Pull cross-skilled cover from queues you have consciously deprioritised. Then the heavier levers: overtime and extended hours, back-office and support colleagues with current skills brought onto simple surge contacts, and — for multi-day events — a one-page surge brief plus FAQ that lets borrowed help handle the dominant contact type credibly within hours rather than weeks.

Each lever has a cost, and the playbook is honest about it: held shrinkage is borrowed, not saved, and must be repaid within days or it becomes a quality and engagement debt; overtime fatigues; borrowed colleagues dilute quality on complex work, which is why they take the simple dominant contact and nothing else. The lever ladder exists so the operation reaches for the cheapest lever that works — not the most dramatic one — and logs what was spent, so recovery is planned rather than discovered.

After the peak — the tail and the loop

Surges have tails. Callback promises made during the peak come due; customers who could not get through retry; the deferred work returns; the held shrinkage must be repaid. The playbook treats the tail as part of the event — a recovery forecast for the following days, a deliberate plan for clearing the backlog in priority order, and the training and coaching that were re-timed actually re-delivered, visibly, so the floor learns that re-timed was true.

Then the review, within the fortnight: how long from external trigger to detection, from detection to reforecast, from reforecast to first lever and first published answer? Which deflection genuinely answered and which merely blocked? What did the surge teach about the contact types you will see next time — because for any operation in a regulated or consumer-facing industry, there is always a next time. The surge you cannot forecast individually is still an event class you can get measurably better at.

The closing principle

You cannot forecast the individual surge, but you can playbook the class: detect before the queue does, reforecast and declare honestly, answer the question before it becomes a contact, and climb the lever ladder from the bottom. The operations that look composed on surge day decided all of this months earlier.