Every travel operation plans for peak volume. Very few break because of volume. They break because of the four assumptions underneath the plan, all of which fail in the same fortnight.
Ask an operations director what went wrong in peak and you will usually get a volume answer. Calls were up, the team was stretched, we did what we could. It is an honest answer and it is almost never the real one, because volume is the one variable everybody plans for. If the season broke, it broke somewhere the plan was not looking.
In travel and tourism the same four failures recur, and they compound — which is why peak feels less like a gradual strain and more like something snapping in a particular fortnight.
Failure one: your forecast is monthly and your problem is hourly
Most seasonal forecasting in this industry works at a granularity that hides the actual problem. A month is forecast, a headcount is derived, and the plan is signed off. Inside that month sit a handful of hours where demand is three or four times the daily mean — the Sunday evening after a bank holiday, the morning after a cancellation, the hour an attraction's ticket office opens on the first Saturday in August.
Averaged across the month, the resourcing looks adequate. It is adequate for most hours and catastrophically short for the ones that determine the customer's experience and your abandonment rate. The number that matters is not contacts per month; it is the ratio of your busiest hour to your median hour.
Operators who fixed this did not necessarily add people. They changed what happens in the peak hour — overflow, self-service, callback rather than queue — which is a different intervention entirely from the one a monthly forecast implies.
Failure two: shrinkage moves against you exactly when it cannot
Shrinkage is planned as an annual percentage and then behaves seasonally. In summer it goes the wrong way for several reasons at once: annual leave concentrates precisely in the months you need cover, sickness rises with sustained pressure, and the training you deferred to get through the season stops happening.
The compounding is the dangerous part. Higher shrinkage means longer queues, longer queues mean more repeat contacts from people chasing, more repeat contacts mean higher volume, and higher volume means more pressure and more sickness. By the time it shows up in a weekly report you are a fortnight into a loop that does not stop on its own.
The fix is not a better annual shrinkage assumption. It is a seasonal one, modelled separately for the eight to ten weeks that matter, and a hard rule about which activities are protected rather than sacrificed. There is no published seasonal shrinkage benchmark for this sector in any of our markets, so the only valid figure is your own.
Failure three: it is the skill mix, not the headcount
The third failure surprises people who have already solved the first two. You have the people. You do not have the right ones in the right hour.
In a multi-market operation this is nearly always a language problem first. Peak volume does not arrive evenly across your language groups; it arrives disproportionately in the markets whose holidays fall that week. An operation staffed to an annual language mix will be simultaneously over-resourced in one language and uncovered in another, in the same hour.
It is a complexity problem second. The interactions that spike hardest in peak are amendments, cancellations and problem resolution — the ones that need experienced people. Seasonal hires, however good, cannot take them for the first several weeks.
Failure four: the escalation path silently inverts
This one applies to anywhere with a physical point of service, and it is the failure operators recognise fastest when it is described to them.
The designed model is that the central team handles inbound and the site handles what is in front of it. Under peak load that inverts. When central cannot answer, the caller rings the site directly. The desk — already managing a queue of people — becomes the overflow contact centre, with no training for it and no reporting on it. The same thing happens at a tourist information counter and at an attraction's ticket window.
Two things follow. Your contact centre metrics improve during the worst weeks, because the demand went somewhere you cannot see. And the person standing at the desk waits behind a phone call, which is where your review scores go in September.
If your central abandonment rate looks better in peak than in shoulder season, this is almost certainly what is happening — and the number you need to confirm it, inbound contact volume at site level, is one most groups do not collect.
Failure five, and the one that hides the rest: measurement lags the season
Everything above is knowable in advance, and most operations discover it in the post-season review, by which point the fix is twelve months away. Weekly reporting is too slow when the failure window is a fortnight, and monthly reporting simply describes what already happened.
What to fix before next peak
1. Re-forecast at hourly granularity for the eight weeks that matter, and calculate your peak-hour to median-hour ratio.
2. Model shrinkage seasonally rather than annually, and name in advance which activities are protected.
3. Map skill and language coverage by hour rather than by headcount, and identify the specific hours where a language is uncovered.
4. Instrument inbound contact volume at site level so you can see the escalation path invert.
5. Agree three interventions you can trigger within twenty-four hours, and the threshold for each, before the season starts rather than during it.
None of these is a technology decision, which is the point. They determine whether a technology decision would help, and what it would need to do.