You have one Tuesday. Three sites want it.
Kingsway's rating has drifted down a tenth. Northgate has a review mentioning a dirty toilet. Old Mill has had four cold-food complaints in a fortnight and you haven't been there since July.
The honest problem is not that you can't tell something is wrong. It's that all three look wrong, and the way most of us choose between them is whichever one we heard about most recently, or whichever manager is most likely to complain if we don't turn up. That's not a bad instinct. It's just not a repeatable one, and it means the quiet site gets visited last every single time.
What follows is a scoring sheet. It is deliberately crude — five factors, four levels each, weights that are whole numbers. The point is not precision. The point is that you write the weights down once, when you're calm, so that on Monday morning you're only filling in boxes rather than re-deciding your priorities under pressure.
The five factors, and why each one is weighted the way it is
Score each site 0–3 on each factor, then multiply by the weight.
| Factor | 0 | 1 | 2 | 3 | Weight |
|---|---|---|---|---|---|
| Rate vs group | below group | 1.0–1.5× | 1.5–2× | over 2× | ×3 |
| Direction | improving | flat | up this month | up three months running | ×3 |
| Severity | none | staff conduct | hygiene | alleged harm | ×4 |
| Concentration | diffuse | two or three themes | one theme, a third | one theme, over half | ×2 |
| Time since visit | under 2 weeks | 2–4 weeks | 5–8 weeks | over 8 weeks | ×1 |
Maximum possible score is 39.
Rate vs group has to be a rate, not a count. Your busiest site will always have the most complaints, and if you rank on raw numbers you will spend your whole year at the flagship. Complaints per hundred covers is the version that survives contact with reality — there's a longer argument for that here, including why you shouldn't bother with a site running under about 400 covers a month.
Direction carries the same weight as rate because a site at 1.2× and climbing is a worse bet for your Tuesday than a site sitting at 1.8× for the last year. The steady one is a known cost. The climbing one is a cost that hasn't finished growing.
Severity is weighted highest because it's the only factor where being wrong is expensive in a way you can't fix later. A cold main is a bad night. An allergy incident is a different category of event.
Concentration is the odd one out: it isn't a measure of how bad things are, it's a measure of whether a visit will accomplish anything. A site with one theme at 60 per cent has something specific you can go and look at. A site with eleven complaints about eleven different things is either genuinely averagely mediocre or has a management problem, and neither of those is a Tuesday-afternoon fix.
Time since visit is weighted lowest on purpose. It is there as a tiebreaker and a corrective — without it, the site nobody complains about never sees you.
What it looks like filled in
Three sites, same week.
| Rate | Direction | Severity | Concentration | Since visit | Score | |
|---|---|---|---|---|---|---|
| Old Mill | 1.9× → 2 | 3 months up → 3 | none → 0 | cold food 54% → 3 | 9 weeks → 3 | 24 |
| Northgate | 1.3× → 1 | up this month → 2 | hygiene → 2 | diffuse → 0 | 3 weeks → 1 | 18 |
| Kingsway | 1.1× → 1 | flat → 1 | none → 0 | diffuse → 0 | 6 weeks → 2 | 8 |
Old Mill: 6 + 9 + 0 + 6 + 3 = 24. Northgate: 3 + 6 + 8 + 0 + 1 = 18. Kingsway: 3 + 3 + 0 + 0 + 2 = 8.
Old Mill wins the Tuesday, and notice that it wins without a single severe review. It wins on a rate that's nearly double the group, three months of drift, a theme concentrated enough to investigate, and two months of you not being there. Northgate scores well on the one dramatic review and badly on everything else, which is exactly right — a single hygiene mention with no pattern behind it is a phone call to the manager and a check on the cleaning rota, not a drive.
Kingsway at 8 is the useful part of the exercise. It was the site that made you open the laptop, because a rating dropping from 4.4 to 4.3 is visible on a dashboard in a way that "0.6 versus 0.54 per hundred covers" is not. On the numbers it's fine.
The override that beats the score
Any review alleging physical harm — an allergic reaction, a foreign object, illness after a meal — goes first, regardless of what the sheet says. Don't score it, don't compare it, don't finish the table. That's a same-day response with a documented timeline, and the sequence for the first hour is here.
The score is for the ordinary case, which is most weeks. It is not for the week where something actually happened.
Where this gets you into trouble
Small sites swing wildly. A 95-covers-a-week site can move a full band on the rate factor from three complaints. If your smallest site keeps topping the sheet, look at whether it's topping it on genuine volume or on arithmetic noise.
A local problem and a group problem score identically. If your new menu spec is producing cold food everywhere, Old Mill will score highest simply because it's the site whose staff are least able to absorb a bad spec — and you'll drive there and fix nothing that lasts. Before booking the visit, check whether the same theme is rising at every site. That check takes about ten minutes and has saved us several pointless afternoons.
The weights are ours, not yours. If you run late-night sites, conduct complaints probably deserve more than ×4. If you're mid-refurbishment somewhere, time since visit is meaningless. Change the numbers — just change them on a quiet Thursday and then leave them alone for a quarter, because a scoring sheet you adjust every week is a gut feeling wearing a spreadsheet.
A high score is not a diagnosis. It tells you where to go, not what you'll find. Arriving without a specific thing to observe is how a site visit becomes a coffee with the GM. The version that isn't is the next thing to read.
The sheet takes about four minutes a week once you have the rates to hand. Getting the rates to hand is the actual work, and it's the bit worth automating — OMMU pulls complaints per hundred covers and theme concentration per site so the boxes fill themselves, but a Monday-morning spreadsheet does the same job if you're willing to keep it up.