Pull two years of reviews, plot the complaint rate by week, and the chart is not noise. The same weeks are bad both years.
Not vaguely similar — the same weeks, with the same complaints in them. Which is a slightly deflating discovery, because it means a meaningful share of what gets treated as a bad fortnight requiring investigation was visible on a calendar in January.
It is also useful, because a predictable problem can be staffed. Here are the recurring ones, what each actually produces, and the move that is only available in advance.
The first warm Saturday
The single most reliable complaint spike in the year, and it catches people every time because it is not on any calendar — it is weather.
The first Saturday above about 20°C produces a demand shock of a kind no other day does. Everyone decides at eleven in the morning, simultaneously. Covers can run 40–60% above a normal Saturday with no bookings to warn you, outdoor seating opens for the first time since September, and the staff working it have had six months off from volume like that.
What the reviews say: waits to be seated, waits to order outside, warm drinks, service that "fell apart after about seven".
The complaint behind the complaint: outdoor sections are nearly always under-covered. A terrace that adds 30% to your capacity is frequently served by the same number of staff, with a longer walk and no station.
Only available in advance: decide in February what your outdoor service model is — who covers it, where their station is, whether there is a dedicated runner — and put two people on call from the first week of April. A forecast two days out is enough notice to use them.
The week schools break up, and the week they go back
Two distinct events with opposite failure modes.
Breaking up brings a guest mix shift: more children, more large tables, earlier sittings, longer dwell times, lower spend per head. The complaints are about noise, waits for high chairs, and children's food arriving with the adults' rather than before.
Going back is the reverse — the Tuesday your 5pm covers halve and your lunch doubles — and its complaint is about being under-staffed at a daypart you were not watching.
Only available in advance: check your high-chair count and your kids' menu stock in the last week of term, and move the kids' courses to fire first as a standing instruction for six weeks. The second one is free and removes most of the complaints.
Bank holiday Sundays, and the Tuesday after
Bank holiday weekends generate two separate problems and the second one is the expensive one.
The Sunday is a volume problem and people usually plan for it. The Tuesday is a staffing problem nobody plans for: your best people have worked three days straight, the rota is covered on paper by whoever had the weekend off, and the kitchen is working through produce ordered for Sunday volume.
What the reviews say: on the Sunday, waits and roast availability. On the Tuesday, flat service and inconsistent food, described in a puzzled tone because the guest cannot see why a quiet Tuesday was like that.
And the hours problem, which is in a category of its own: bank holidays are where wrong opening hours do their worst damage, because a guest who drives over to a closed door leaves the angriest review you will get all year. The only review written by someone who never ate your food is worth preventing with a half-hour in advance.
Only available in advance: staff the Tuesday properly, and enter a year of special hours in one sitting rather than remembering each one.
December, and specifically the party weeks
The second and third weeks of December are a different business operating in your building: set menus, pre-orders, large groups, deposits, dietary lists, and a guest who booked in October and has been looking forward to it since.
What the reviews say: three themes, in order — courses arriving unevenly across a long table, pre-orders coming out wrong, and anything involving dietary requirements going astray. Almost never about the food itself.
The underlying cause: a party of twenty-four with pre-orders is an administrative task pretending to be a service task, and it fails at the point where a paper pre-order sheet meets a till designed for tables of four.
Only available in advance: decide in October how pre-orders get transcribed and who owns the dietary list for each booking. In December there is no capacity to design a process.
The Tuesday after any menu change
Not seasonal, but calendar-driven in the same way and worth including because it is the most controllable item on this list.
A new menu produces a complaint bump for ten to fourteen days, reliably. Dishes are cooked by people who have made them four times, servers are describing them from a tasting rather than from experience, and the till has new buttons in new places.
What the reviews say: inconsistency, longer waits, and servers who could not answer a question.
Only available in advance: launch on a Tuesday rather than a Friday, keep three familiar dishes on, and accept the fortnight. Judging a new menu from its first ten days of reviews is reading the learning curve rather than the menu.
Building your own calendar
The general version is worth more than my list, because your estate has its own events — a stadium, a university term, a festival, a ferry timetable.
- Two years of reviews, complaint rate by week. Rate, not count, or you will simply rediscover when you are busy.
- Mark the weeks that are bad in both years. Those are your structural weeks. A week that was bad once was a bad week.
- For each one, read the reviews and name the dominant theme. This is the step that makes it actionable — "week 31 is bad" is useless, "week 31 is waits on the terrace" is a rota decision.
- Write the pre-emptive action and the date it has to be taken, which is nearly always weeks earlier than the event.
- Put those dates in a shared calendar, with the action, not the event.
Half an hour, once, and it mostly does not need redoing — these patterns move very little year to year.
The traps in reading seasonality
Two years is the minimum and it is still thin. A single Easter that was bad because the kitchen flooded will look structural in a two-year series. Where you can, look at three.
Seasonality is a reason, not an excuse. The point of knowing week 31 is hard is to staff it, not to discount the reviews from it. A guest who waited forty minutes on the first warm Saturday is not interested in the fact that you expected to be busy, and the review counts the same as any other.
It will distort your month-on-month reading. If a quarter contains December and a quarter does not, comparing them tells you about the calendar rather than the operation. Compare like weeks year on year — which is also the reason published industry benchmarks are nearly useless at this level, and why your own trailing baseline is the only valid comparison.
Weather-driven events cannot be diarised. The first warm Saturday is a standing readiness, not a date. The best available version is a two-day forecast check from April and two people on call.
Do not pre-empt everything. Five calendar events with real actions will get done. Twenty becomes a document nobody opens, and the January version of you is substantially more optimistic about the August version of you than is warranted.
OMMU holds complaint rates by week with last year's series underneath, which makes "is this a bad fortnight or is it week 31 again" a glance rather than a debate. Two years of reviews and an afternoon with a spreadsheet produces the same calendar — and it is the kind of work that only ever gets done in advance, by somebody who has decided to.