Restaurant hygiene ratings in Cardiff

Limits of these figures

Known results: 805 ratings; needing improvement: 31 ratings; waiting or unknown: 96 records; exempt: 0 premises. Waiting and exempt records are outside the known results.

Rating dates missing from 0 ratings. The median and age share use 805 ratings with known dates.

Low ratings with complete component scores: 31 ratings. Low ratings excluded because a score is missing: 0 ratings. Published ratings are not reconstructed from their components.

Unknown-result scenarios: 3.4% to 14.1% if every waiting or unknown result became known. These are scenario bounds, not confidence intervals. Exempt premises remain excluded.

Comparison covers 805 ratings; excludes 0 ratings without an adequate same-type base. The base uses other premises in Wales under the same scheme.

No 30-day change is published: two accepted files exactly 30 days apart are required. One snapshot cannot show how many premises improved or worsened.

Inspections are not a random sample. These intervals describe the known-result shares, not the chance that a particular premises is safe.

It does not tell you about one premises. A share describes the known ratings in an area on the day of the file, and a single rating can change the day after.

It does not count premises waiting for a rating, and it does not count exempt premises. Both stay outside the denominator, so a low share is not evidence that everything was inspected.

It does not say an inspection is overdue. The age of a rating is the gap since the last published inspection, and the publisher sets no deadline we could measure against.

It does not rank areas. Where two intervals overlap, the order of the two numbers carries no information.

It does not combine schemes. A Scottish pass result and a 0 to 5 score are different measurements and are never averaged.