How scoring works

In short

• Each council department gets a score from 0 to 100, for example a housing score or an adult social care score.
• A department score combines public feedback (50%), oversight reports (30%) and service performance data (20%), where they are available.
• Each council also gets a council score: the average of its department scores, once at least 3 departments have a full score.
• Recent ratings count more than older ones.
• Scores based on few ratings are pulled towards the average for that department in the same nation, so a handful of ratings cannot swing a score.
• We only publish and rank a score when there are enough ratings, and we always show how certain it is.
• Scores are updated monthly. Every setting is published on this page and in our open data.

The three parts of a department score

Each department score is made up of as many as three parts. Each part is a number from 0 to 100.

If a part is not available for a department, we leave it out and share its weight between the other parts in proportion. For example, if there is no performance data, public feedback counts for 62.5% and oversight reports count for 37.5%. Many departments have no regular inspections or national performance data, so their scores are based on public feedback alone.

A department must have a public feedback score to get a score.

Public feedback

From ratings to points

People choose one department and rate six areas of their experience from 1 to 5. The six areas are the same for every department, with wording and examples to suit each one. Read about the six areas. We turn each rating into points:

"Not applicable / don't know" answers are not counted. Each rating must include at least 3 rated areas.

Which ratings count

We use accepted ratings of the department from the last 24 months. A rating is accepted after a 48-hour cooling-off period and our integrity checks.

Recent ratings count more

A rating's weight halves every 12 months. A new rating has a weight of 1, and a rating from 12 months ago has a weight of 0.5.

weight = 0.5 ^ (age of rating in months ÷ 12)

We also take account of how long ago the person's contact with the service was, when they gave the rating:

For example, a rating given 6 months ago about contact in the previous 12 months has a weight of 0.71. A rating given 18 months ago about contact 1 to 2 years before that has a weight of 0.27.

Area scores

For each of the six areas, the score is the weighted average of the points.

Effective sample size

Because older ratings count for less, we measure how much information we have using an effective sample size. If every rating had the same weight, it would equal the number of ratings. When weights differ, it is smaller.

effective sample size = (sum of weights)² ÷ (sum of squared weights)

For example, 4 ratings with weights of 1, 1, 0.5 and 0.5 have an effective sample size of 3² ÷ 2.5 = 3.6.

Protection for small numbers of ratings

When only a few people have rated an area, one or two ratings could move its score a lot. To prevent this, we blend each area score with the national average for that area, as if we had added 10 extra ratings at the national average. The national average is worked out for the same department in the same nation, so housing ratings in Wales are compared with housing ratings in Wales. When there are many ratings, this makes very little difference. This is sometimes called Bayesian shrinkage.

adjusted score = (sum of weights × weighted average + 10 × national average) ÷ (sum of weights + 10)

The public feedback score

The public feedback score is the average of the six area scores, with each area counting equally. An area is only included if its effective sample size is at least 10, and at least 4 of the 6 areas must be included.

People also give an overall rating of their experience. That separate rating does not change the score.

Oversight reports

Oversight reports are published reports from inspectorates and regulators, such as Care Inspectorate Wales. Each report is recorded against the department it covers. A report about social care as a whole counts for both adult and children’s social care. We use verified reports from the last 48 months. A report's weight halves every 24 months.

Reports use different words and grades, so we convert each one to a score from 0 to 100 using a published rubric. Two different people must enter and check each conversion. The oversight score is the weighted average of the converted report scores.

Care Inspectorate Wales local authority inspections

Care Inspectorate Wales does not give local authority inspections an overall grade. We code its findings for each of four principles: People – voice and choice, Prevention, Well-being and Partnerships and integration. The report score is the average of the four. If priority action is required, the report score is capped at 25.

Conversion tables for other reports

We only use the reports that apply to each nation and department. Tables for other nations are included so they are ready when those nations are added.

Joint inspection of child protection arrangements (Wales)

Grade or rating	Score
Strengths clearly outweigh areas for improvement	85
Strengths and areas for improvement are broadly balanced	60
Areas for improvement outweigh strengths	35
Significant concerns or priority action required	15

Care Inspectorate Wales — regulated service rating

Grade or rating	Score
Excellent	90
Good	70
Requires improvement	40
Requires significant improvement	15

Care Quality Commission — local authority assessment

Grade or rating	Score
Outstanding	90
Good	70
Requires improvement	40
Inadequate	15

Ofsted — inspection of local authority children's services (sub-judgements)

Grade or rating	Score
Outstanding	90
Good	70
Requires improvement to be good	40
Inadequate	15

Care Inspectorate (Scotland) — six-point grade

Grade or rating	Score
6 – Excellent	100
5 – Very good	80
4 – Good	60
3 – Adequate	40
2 – Weak	20
1 – Unsatisfactory	0

Service performance data

Service performance data are official published measures, such as social services performance data published by the Welsh Government. Each measure is recorded against a department, or against social care as a whole, which counts for both social care departments. For each measure, we compare a council with the other councils in the same nation.

1. We work out how far the council is from the average, measured in standard deviations. This is called a z-score. For measures where a lower figure is better, we reverse the sign, so a positive z-score always means better than average.
2. We limit the z-score to between −2.5 and +2.5, so one unusual figure cannot dominate.
3. We convert it to a score: 50 + 20 × z-score. A score of 50 means the national average.

A measure is only used if at least 5 councils have comparable data, and data more than 24 months old is not used. A department needs at least 3 measures for a performance score. The performance score is the average of its measure scores.

Because performance scores compare councils with each other, they show relative position, not whether a particular standard has been met.

Departments and the council score

People can rate 12 council departments: adult social care, children’s social care, education and additional needs, housing and homelessness, council tax and benefits, bins, recycling and street cleaning, planning and building control, roads, transport and parking, environmental health, licensing and trading standards, libraries, leisure and culture, parks, green spaces and community services and customer contact and complaints handling. Departments are opened nation by nation, so not every department is open everywhere yet.

Department scores

Each department a council runs gets its own score, calculated as described above using only ratings and evidence about that department. There is also a social care score, which combines the ratings and evidence for adult and children’s social care.

Not every council runs every department. In parts of England, the county council and the district council run different services. In Northern Ireland, some services are run by other organisations, such as Health and Social Care Trusts. We only score a council for the departments it runs, and we say who runs the others.

The council score

• The council score is the average of the council’s department scores that have the status "Scored". Each department counts equally, however many ratings it has.
• A council needs at least 3 departments with a full score to get a council score.
• If fewer than 3 departments have a full score, but at least 3 have a full or provisional score, the council gets a provisional council score, using those departments.
• Only departments that are open in the council’s nation count towards its council score.
• The likely range of the council score comes from the likely ranges of the departments it includes: standard error = √(sum of each department’s standard error²) ÷ number of departments, where a department’s standard error is the width of its likely range ÷ (2 × 1.96).
• Council scores for each of the six areas are the average of the department area scores, when at least 3 departments have enough ratings for that area.
• The public feedback, oversight and performance figures shown with a council score are averages of its departments’ figures, for context. They are not weighted again.

A worked example

Hypothetical example

This example uses invented numbers for an imaginary "Council X" to show how the method works. It does not describe any real council. Steps 1 to 5 work out Council X’s adult social care score. Step 6 combines it with other department scores to make a council score.

Step 1: area scores and public feedback

Council X's Communication ratings have a weighted average of 60 points and a sum of weights of 30. The national average for Communication in adult social care is 55. The adjusted score is (30 × 60 + 10 × 55) ÷ (30 + 10) = 58.8.

Compare an area with only a few ratings: a weighted average of 90 with a sum of weights of 5 becomes (5 × 90 + 10 × 55) ÷ (5 + 10) = 66.7. The small sample is pulled strongly towards the national average.

Each area has an effective sample size of at least 10, so all six are included.

Step 2: oversight score

Council X had a Care Inspectorate Wales inspection 6 months ago. Its four principles were coded like this:

The report score is the average: 60.0. An earlier report, 30 months old, rated a council-run care service "Requires improvement", which converts to 40.

The weights are 0.5 ^ (6 ÷ 24) = 0.84 and 0.5 ^ (30 ÷ 24) = 0.42. The oversight score is (60.0 × 0.84 + 40 × 0.42) ÷ (0.84 + 0.42) = 53.3.

Step 3: performance score

The performance score is the average of the three measure scores: 48.3.

Step 4: the department score

0.5 × 60.1 + 0.3 × 53.3 + 0.2 × 48.3 = 55.7

Scores are shown as whole numbers, so Council X's adult social care score is 56, in the Mixed band.

If Council X had no performance data, its score would be 0.625 × 60.1 + 0.375 × 53.3 = 57.6.

Step 5: how certain the public feedback score is

The ratings for Council X have a standard deviation of 30 points and an effective sample size of 45. The 95% confidence interval for the public feedback score is 60.1 ± 1.96 × 30 ÷ √45 = 60.1 ± 8.8, or 51.4 to 68.9.

With an effective sample size of 45, Council X's adult social care score has the status "Scored", so it is published and ranked.

Step 6: the council score

Council X runs other departments too. These are its (invented) department results:

3 departments have a full score, which meets the minimum of 3, so the council score uses those departments and leaves out the provisional one. The council score is the average of 55.7, 71.0, 49.0, which is 58.6.

Its likely range, from the departments’ likely ranges, is 54.5 to 62.6.

Shown as a whole number, Council X’s council score is 59, in the Mixed band.

How certain is a score?

Every score is an estimate. We show a 95% confidence interval: the range of values the score could reasonably take, given how many ratings there are and how much they vary.

score ± 1.96 × standard deviation ÷ √(effective sample size)

A wide interval means the score is less certain. The interval gets narrower as more people rate a department.

The interval only reflects random variation. It does not tell you whether the people who took part are representative of everyone who uses services.

When we show a score

A council score is "Scored" when at least 3 of its departments are scored, and provisional when at least 3 departments are scored or provisional. Otherwise no council score is shown.

Councils that are not accepting ratings yet are shown as "Not yet accepting ratings".

Score bands

Each score falls into one of five bands. Colour is never used on its own: the band name and score are always shown.

Trends and changes

• We show how a score has changed since the previous month, and since 12 months earlier. Changes are shown with a plus or minus sign, for example +2.3 or −1.5.
• Changes of less than 1.0 point are described as "no meaningful change".
• The trend uses up to 12 monthly scores, and needs at least 4.
• A score is described as improving or declining if it is changing by at least 0.25 points a month (3 points a year), and the change is at least 2 times its standard error. The standard error measures how much of the trend could be down to chance. Otherwise, the trend is described as stable.

Rankings

• Only scores with the status "Scored" are ranked, and only once at least 5 comparable councils have one. Until then, full scores are shown without a rank.
• Councils are only ranked against other councils in the same nation, for the same score: council scores against council scores, and housing scores against housing scores. We do not rank councils across nations, because laws, inspections and data are different.
• In England, county councils and district councils run different parts of some services. Council scores, and scores for services shared between the two, are ranked separately for county councils, district councils and all-purpose councils (unitary, metropolitan borough and London borough councils).
• Councils with the same score share a rank.
• Each rank comes with a plausible rank range, based on how much the council's confidence interval overlaps with other councils' intervals. Where ranges overlap, small differences in rank are not meaningful.

Indicators to watch

Indicators to watch are automatic, rule-based checks. They show where it may be worth looking more closely. They are not accusations or findings of wrongdoing.

For a department score, the overall level is high if R3 applies or 3 or more indicators apply, elevated if 1 or 2 indicators apply, and none if no indicators apply.

For a council score, the indicators are R1, D1 and D2. The level is high if 2 or more departments have a high level, or 1 does and R1 applies; elevated if any indicator applies; and none otherwise.

How much weight to put on a score

Trust indicators help you judge how much weight to put on a score. They are about the evidence, not about the council.

Anything that does not meet the moderate level is low. We combine the indicators into an overall level of high, moderate or low by averaging them. If the number of ratings indicator is low, the overall level is low. Department scores use T1 to T6. Council scores use T1 (the ratings for all their departments together), T2, T5 and T7.

When scores are updated

• Every hour: ratings that have passed the 48-hour cooling-off period are accepted, and integrity checks run.
• Every day: data that has reached the end of its retention period is deleted.
• Every month: on the 1st, preview scores are calculated for the previous month. An administrator reviews them before they are published.
• Every year: an annual score is calculated for the financial year ending 31 March, using a 12-month window, without extra weight for more recent ratings.
• When published: new oversight reports and performance data are added once they have been entered and checked by two different people.

Limitations

• People choose whether to rate a council, so ratings are not a random sample. People with very good or very poor experiences may be more likely to take part.
• We cannot check individual experiences, because we do not know who submitted them.
• Scores do not adjust for differences between councils, such as population, deprivation, funding or demand.
• Councils run different mixes of services. A council score summarises the departments a council runs, so it is only compared with councils that run similar services.
• Each department counts equally in the council score, even though departments differ in size, budget and how many people use them.
• Departments that fewer people contact, such as planning, may take longer to get enough ratings for a score.
• Oversight reports are published infrequently and may describe services as they were some time ago.
• Performance data can be published months after the period it covers, and how measures are defined can change.
• Performance scores are relative. They show how a council compares with others, not whether a standard has been met.
• Pulling small samples towards the national average makes scores more stable, but it can hide real differences at councils with few ratings.
• The categories, weights and rubrics reflect our judgement. We publish them so that others can examine and challenge them.
• A score cannot tell you what service any one person will receive.

Versions and changes

The current method is version 1.1, effective from 01/10/2026. Any change to a setting or rule creates a new version, and we publish what changed and why.

Changelog

• Version 1.1 (effective from 01/10/2026): scores for each council department, a combined social care score and a council score. National averages and rankings are worked out for each department within each nation. English county, district and all-purpose councils are ranked in separate groups where they run different services. New indicators D1, D2 and T7 for council scores.
• Version 1.0: first published version of the method, for council social care.

Get the settings as data

Every setting on this page is also available in machine-readable form from our methodology API at /api/v1/methodology. See open data for details.

Last reviewed: 15/09/2026