Every rating in England, Wales and Northern Ireland is published, and when an officer comes back the new one is published beside it. That leaves a record of what happens to a business after a bad day, which is the question an owner actually has and the one nobody answers.

We asked it of every re-rating in the register: where a business landed, and how long it waited for the chance.

Almost everyone goes up

Taking every business that held a 0, 1, 2, 3 or 4 and was later rated again:

Where a re-rated business landed. From a 4 the only way up is a 5, which is why its two middle figures match. Up and down sum to 100 on every row: a published change always moves the rating. Re-ratings to 16 October 2026.
Was rated	Re-ratings	Went up	Went straight to 5	Went down
0	5,536	100%	17.2%	0%
1	26,864	95.6%	31.3%	4.4%
2	22,481	86.1%	38.7%	13.9%
3	49,099	72.8%	38.4%	27.2%
4	85,315	64.8%	64.8%	35.2%

A business rated 1 that was rated again went up 95.6% of the time, and 31.3% of them went all the way to 5 in a single step. From a 0 there is nowhere to go but up, and 17.2% still managed 5 at the next inspection.

Some of that is simply a bad day being corrected. An officer who scores a kitchen 1 comes back sooner, and the business that was having a terrible week is not having one when they do. But the direction is not ambiguous, and it is the opposite of how a 1 feels at the time.

A 3 is where the officer stops coming back

+322days is what a 3 adds to the wait for the next inspection, against a 2. Every other step in the table below is between 9 and 59 days.

A 3 is “generally satisfactory”, and it is the rating most owners are quietly relieved to get. It is also where the register stops paying attention.

Days from one published rating to the next for the same premises. The 90th percentile is the long tail: one pair in ten waited longer than that. Pairs to 16 October 2026.
Rating you landed on	Pairs	Median days	Which is about	90th percentile
0	4,895	113	4 months	376
1	23,198	122	4 months	528
2	18,268	181	6 months	637
3	32,717	503	17 months	756
4	50,027	525	17 months	869
5	31,927	559	18 months	1,038

A 0 or a 1 brings the officer back in about 4 months. A 3 waits about 17, a 4 about the same and a 5 barely longer again. The whole gap opens between a 2 and a 3 and then flattens out. Inspection is risk-based, so that is the system working as designed: the kitchens that scared somebody get seen first.

The long tail runs the same way. One wait in ten that started from a 2 ran longer than 637 days; from a 3, longer than 756.

Falling is not what makes a 3 unusual. 27.2% of them came out lower, but so did 35.2% of businesses rated 4. The “Went down” column in the first table climbs the whole way up the scale, for the obvious reason that the higher you sit the more room there is below you. What is unusual about a 3 is how long you hold it before anyone looks again.

What this does not say

It does not say a bad rating will improve on its own, and nothing here measures why any particular business went up. A re-inspection interval is how an authority schedules its work, not a promise to anybody, and a median is not your date.

It does not say records are what moved the needle either. We sell an app for keeping them and we are not going to pretend the register proves that. What it shows is where businesses in each position ended up.

Worth saying plainly: 17 months is a long time to hold on to paperwork you will be asked about, and whatever you keep it in, keep it as you go. The one thing an officer can see immediately is a diary written up the night before.

Where these numbers come from

Food Standards Agency open data for England, Wales and Northern Ireland. Scotland runs a different scheme and is not included.

The two tables do not filter alike. The first counts rating changes where both the old and the new rating are a published 0 to 5, because a move to or from awaiting-inspection is not a business going up or down. The second counts every change that LANDED on a published 0 to 5, whatever it came from: a premises going from awaiting-inspection to a 3 has landed on a 3 and starts waiting like any other.

The wait is the number of days from one published rating to the next for the same premises. A spell as awaiting-inspection does not end a wait, because it is not a rating an owner can be judged on. Counting the next change of any kind instead, including those, reads 499 days for a 3 rather than 503, and 534 for a 5 rather than 559.

A wait is counted only where a next rating exists, so a business still waiting is not in it. That biases the figures DOWN, because the longest waits are the ones that have not ended yet, and it bites hardest on the 90th percentile: the tail is the part most likely to still be running.

Last re-run against the register on 16 October 2026, by the query in marketing/scripts/blog-figures.sql. Nothing on this page updates on its own.