The Oversight Framework, in real time

Where you stand right now on your own numbers, which of the framework’s 86 metrics you can evidence, and the published sums behind the segment you were given. We don’t compute a segment, and the page tells you why.

1 What the framework does to you

Every quarter NHS England puts each trust and each ICB into one of four segments. Segment 1 means you’re doing well nearly everywhere. Segment 4 means you’re a long way behind across several areas and can expect a co-ordinated response, possibly intervention. Once a year you also get a capability rating — green, amber-green, amber-red or red — decided by the regional team after your own self-assessment across six board domains.

Between those two dates you want to know something else: where do we stand right now, on our own numbers, and what can’t we evidence at all. That’s what this is for.

2 The thing that shapes everything else

Three of the four scoring models are relative. Target-and-spread, floor-and-spread and even-spread all rank you by percentile against every other organisation of your type. To work that out you need the national distribution, and only NHS England has it — your system holds your data. Only banded scores are absolute, and even those need thresholds we haven’t got.

So we don’t show you a segment. The page says so before it says anything else.

3 A pinned copy of the real document

The framework and its methodology manual are published documents, so we serve a pinned copy of each with a checksum on it instead of typing the contents into code. Eighty-six metrics, eight domains, four segments, four scoring models, the published thresholds. If the copy stops matching its own stated metric count it refuses to load, so you can never end up with a blank framework quietly reading as a framework with no metrics.

Each metric says whether it’s scoring or contextual. One row shows why that can’t sit on the metric on its own: the staff survey engagement theme score is scoring for trusts and contextual for ICBs. Get it wrong and you’d score an ICB on a metric the framework calls contextual for ICBs, and it would look perfectly fine on the screen.

4 Which metrics apply to you

The framework publishes how many metrics apply to each type of organisation: 43 for an ICB, 35 for an acute trust, 24 for mental health, 19 for community, 18 for ambulance. We derive the set from the published list and show you both numbers side by side.

For an ICB we get 43, which matches exactly. For any trust we get 50, because the published list tells you which metrics reach trusts but not which ones your type is let off — those sit in the technical annex. So the page says the derived figure is a superset and why. We don’t pick 24 of the 50 and hand them to you as the mental health set.

5 Checking the segment you were given

Steps 2 to 4 of the methodology are written out in full, and the manual says people should be able to recreate the results from the raw data. So we do that, using the scores NHS England published for you. Your board can see how the segment it was handed was arrived at instead of taking it on faith.

Feed it your own internal figures and it stops you: “only PUBLISHED metric scores may be used. Three of the four scoring models are relative and need the national distribution, which a single deployment does not hold, so an internally computed value is not a score and cannot be segmented.”

6 Where you stand right now

Between quarters your own current number is more use than a segment built on data from three months ago, and the framework accepts that a provider’s live figure may be ahead of the official one. So we work out what we can from your record and label it honestly.

These are our counts on our definitions. We haven’t read the annex, so we don’t know how NHS England defines a DNA rate or an 18-week compliant pathway, and we don’t claim ours match theirs. Where there’s no denominator you get “not reported” and the denominator, never 0.0 — a nought per cent DNA rate reads like a service nobody missed.

You can then record a figure against a framework metric. Every value you record says which of three things it is: published by NHS England, worked out internally, or not held at all. Try to hang a 0–1 score off an internal figure and you get: “a score may only be recorded against a PUBLISHED value. An internally computed figure is a value, not a score.”

7 What you can’t evidence

Your segment is worked out over the metrics you were scored on. The set that applies to you is bigger, because anything you didn’t submit scores zero and drops out of the average. So a segment can rest on part of the picture, and most organisations can’t tell you which part. NHS England raises the exception at its end. Nothing at yours has been keeping track.

Coverage answers it from your own systems. Ask for an acute set and you get 35 applicable as published, 50 derived, and every metric sorted into three buckets: evidenced here, held somewhere else, or not scored. Record one internal figure — a 4.8% sickness absence rate for 2026-Q2, out of the rostering record — and that metric moves out of not scored. The count went from 50 to 49.

8 It’s NHS England’s framework

The framework belongs to NHS England. Putting it in front of an organisation in Jersey, Gibraltar, Kenya or Montserrat as something they’re subject to would be wrong, so outside England it stays off until somebody adopts it.

Adopting it’s a decision someone signs for. It needs a named person, a reason and a date, and it won’t save without them: “an adoption decision requires adopted_by, basis. A decision with no named person and no stated basis is not a decision.” Superseded decisions are kept, so you can see who adopted what, when, and on what grounds.

Two more refusals sit alongside it. Ask for coverage without saying what type of organisation you’re and nothing gets guessed: “commvita does not default to acute, which would score a community trust against metrics it does not report.” Ask for a type the framework doesn’t define and it lists the five it does.

9 What we won’t do

The limits aren’t buried in a document you have to ask for. They’re on the page and in the API, next to the thing they limit.

One gap worth being straight about: coverage and the adoption record are API-only at the moment. The tab gives you the framework, the arithmetic, your own figures and the ceiling. Coverage is there today for anyone building a board pack or an integration, and it belongs on the page next.

Where it lives in commvita