A food hygiene rating is not one number. It is three scores added together: how food is handled, the state of the building, and confidence in management, which is the officer’s read of whether the people running the kitchen have a grip on it. Each is scored in demerit points, where zero is fully compliant and higher is worse. What the three scores are sets them out.

Two of those three describe things an officer can see and measure. The third is a judgement. So we expected it to carry more signal: if an inspector forms a view about whether a kitchen is being run properly, that view ought to predict whether the business is still trading in a year better than a note about a cracked tile does.

We checked before writing it up, and it is false. The useful part turned out to be in a column we were not looking at.

All three scores predict about equally

Taking 114,516 food businesses that opened, were inspected within six months, and were on the register long enough to be followed for a year, and asking how many were gone twelve months after opening:

Share gone within a year of opening, by band. Cohort 114,516.
Score	Scored 0-5	Scored 10-15	Scored 20+	Businesses scored 20+
Confidence in management	4.9%	8.7%	21.0%	3,770
Structure and cleanliness	5.1%	10.9%	25.6%	497
Food hygiene and safety	5.1%	11.9%	27.9%	438

Read along the first three columns and the three lines are the same line. A business marked badly on management is 21.0% likely to be gone within a year; on structure, 25.6%; on hygiene, 27.9%. If anything the judgement score is the weakest of the three at the bad end, which is the opposite of what we went looking for.

The finding is the last column

8.6×Confidence in management was marked 20+ on 3,770 businesses. Structure on 497, hygiene on 438, scored by the same officers on the same visits.

Inspectors marked management badly 3,770 times. They marked structure 497 times and hygiene 438 times. A severe management mark was 8.6 times commoner than a severe hygiene one, on the same businesses and by the same officers.

Confidence in management is not the better predictor, then. It is the one new businesses actually fall down on. When one is marked down hard, it is far more often for how the place is run than for the fridge temperature or the state of the building.

Three equally predictive scores sounded like a dull result until we looked at where the severe marks fall. Scores that predict alike do not land alike: of every severe mark handed out across the three scores in this cohort, 80% are on management.

It is not just takeaways

Takeaways are both scored harder and shorter-lived than most food businesses, so a gap like this can be an artefact of which businesses sit at which end. Comparing within a business type instead, the gap holds:

Business type	Gone in a year, scored 20+	Gone in a year, scored 0-5
Takeaways	25.9%	8.3%
Restaurants and cafes	20.3%	5.4%
Retailers	12.3%	4.4%

What this does not say

It would be easy, and wrong, to read this as “keep better records and your business will survive”. The register does not support that and neither do we. What it holds is what officers scored and what happened afterwards. Businesses close for rent, for staff, for a landlord changing the locks, for reasons a hygiene inspection never sees, and a poor management score may be a symptom of the same trouble rather than a cause of anything.

Every inspection scores all three, so none of this says an officer reaches for one score more than another. It says where the severe marks landed.

What the numbers do support is narrower. A 20+ mark fell on management far more often than on hygiene or the building. If you are preparing for an inspection and deciding where to put your attention, that is where new businesses have been losing marks, and it is the half of the job that is written down rather than scrubbed.

Where these numbers come from

Food Standards Agency open data for England, Wales and Northern Ireland. The cohort is 114,516 businesses whose first published inspection fell within 180 days of opening and which were on the register long enough to be followed for a full year. “Gone within a year” means the premises no longer appears in the published register twelve months after it opened, which is not the same as having closed: businesses also leave the register on a change of ownership or a reclassification.

Two tables behind it, both rebuilt nightly from that data: establishment_versions, which keeps every published version of a premises record with its inspection date and its three sub-scores, and business_survival, which derives how long each business stayed on the register. For each business we take its first published inspection, band each of its three scores, and ask whether the premises was still listed 365 days after it opened.

All three scores are banded the same way here, 0-5, 10-15, 20 or more, because confidence in management has no 15 band while the other two do, and banding them differently would not be a comparison. Cells under 25 businesses are not given a percentage.

Last re-run against the register on 16 October 2026. Scotland runs a different scheme and is not included.