Every review routine that gets recommended is too big. Fifteen minutes, five signals, three platforms, a note, an action. It is all sensible and it all dies in about five weeks, usually during a fortnight when someone is off sick.
The failure is not laziness. It is that the routine was designed to produce a weekly answer, and a weekly answer is almost worthless. One week of reviews at a single restaurant is eight to thirty data points with no comparison attached. You cannot tell whether four complaints about pace is bad, because you do not know what a normal week looks like.
What is valuable is the series. And the series needs almost nothing from you each week — it needs one number, recorded reliably, for long enough.
The number
Pick one. Not five.
Complaints this week ÷ covers this week × 100
Complaints per hundred covers. Write it down with the date, in the same place, every Monday.
Why this one, if you only get one: it is normalised, so a busy week and a quiet week are comparable; it responds to operational reality rather than to your review volume; and it moves fast enough to be informative weekly. Average rating fails all three — it is cumulative and barely moves. Review count measures your asking habit rather than your restaurant.
Counting the complaints requires a decision made once and then never revisited: a complaint is a review of three stars or below, or any review containing a specific negative about the food, the service or the room. Write that definition at the front of the notebook. An inferior definition held constant beats a better one you change in March, because changing it breaks the series, and the series is the whole asset.
Three minutes a week. Two to count, one to write.
What a year of it gives you that a week cannot
A normal range. After about twelve weeks you know your band — say 0.4 to 0.9. From then on, every Monday's number answers a question that was previously unanswerable: is this a bad week, or is this a week? Most of the weeks that feel alarming are inside the band, and knowing that saves a genuinely surprising number of unnecessary Tuesday interventions.
Direction, which is the only early warning available. Three consecutive weeks above your band is a signal. A single week above it is not. There is no way to distinguish those two things without the series, which means that without it you are responding to every bad week identically — either by over-reacting or, more commonly, by doing nothing and hoping.
Before-and-after on your own changes. This is where it pays for itself. You move two start times, change a plate, add a runner on the terrace. Did it work? With a series and a marked date, the answer is visible in a fortnight. Without one, the answer is whatever the loudest person in the room believes, and the improvements that quietly did nothing stay in place for years.
Seasonality. By the second year, you know which weeks are structurally hard, which stops a predictable August from being investigated as a crisis. A good share of your variance is on the calendar.
Keeping it to three minutes
The routine survives because it is small, and every addition is a threat. Specifically:
Do not add metrics. Five numbers takes five times as long and is five times as likely to be skipped. One number for a year beats five numbers for six weeks by an enormous margin.
Do not read the reviews during the count. This is the main failure mode. You open the page to count, something infuriating is there, and forty minutes go on composing a reply. Count, write, close. Reading the text properly is a separate and valuable job that belongs in a different slot — and where that reading sits determines whether anything gets fixed.
Do not require a decision. No action item, no note, no follow-up. The number only has to exist. A routine that demands a judgement every week is a routine that gets postponed on the weeks you have no judgement available.
Paper is fine. Better than fine — a notebook in a drawer has no login, no dashboard, and nothing to be out of date. Any format works as long as it is always the same one.
Same slot, same person, named. Monday morning, before anything else, by whoever opens.
What to do when it moves
The one rule: three weeks above the band, then act. Not one.
When it triggers, that is the point at which you do the bigger exercise — read the text, find the dominant theme, work out whether it varies with load. Laying complaint times over the rota is the usual next step and it takes an afternoon. The series is what tells you the afternoon is warranted, which is most of its job: it is a trigger for a real investigation rather than a substitute for one.
And when you make a change, mark the date in the notebook. The marked dates are what turn a column of figures into a record of cause and effect.
What this habit cannot do for you
Low volume makes weekly numbers unreadable. Below roughly 25 reviews a month, one review moves the figure by a third. Record weekly anyway, but read it as a rolling four-week average — same habit, smoother line.
Covers have to be available on a Monday. If your cover count arrives mid-week, record it Tuesday. The day matters less than it always being the same day.
It tells you nothing about what to fix. Deliberately. It is a thermometer. Anyone expecting a weekly number to generate insight will be disappointed and will start adding things to it, at which point it stops being a three-minute habit.
Twelve weeks before it is useful. This is the honest cost and the reason most people never have a series. The first quarter produces a column of numbers with no interpretation available, which feels like pointless admin, which is exactly when it gets abandoned — about six weeks in, with the valuable part still ahead.
One year is the minimum for the seasonal read. Two is noticeably better.
The argument for starting this week rather than designing something better
The main competitor to a three-minute habit is not a bigger routine. It is a bigger routine that does not exist yet — the dashboard somebody is going to set up, the weekly meeting that will start after the refurbishment.
A year from now, the operator with fifty-two numbers in a notebook can tell you whether their restaurant got better, which specific change was responsible, and whether this month is unusual. The operator waiting for the proper system cannot answer any of those, and has been able to answer none of them for twelve months.
OMMU keeps the series automatically with your own changes marked on it, which removes the three minutes and, more usefully, removes the twelve-week gap before it is readable. A notebook does the same thing more slowly, and the notebook is available right now.