What is an anchor example?
An anchor example is a worked illustration of what a particular score looks like on a particular criterion. It accompanies the rubric, giving reviewers something concrete to compare a submission against instead of an abstract scale label. Anchors are drawn from real submissions, annotated to show why that score and no other was correct.
Also called: Score anchor, Anchor
Why a scale without anchors drifts
Anchor examples are routinely confused with guidance notes, and the difference matters. A guidance note describes a criterion in the abstract, where an anchor shows a reviewer an actual submission that earned a three and explains why it did not earn a four. Without anchors, each reviewer silently invents their own scale, and the resulting scores look comparable while measuring different things. The problem is invisible in aggregate reporting, which is why it survives for years.
- Anchors convert a numeric scale into a shared standard reviewers can check themselves against.
- A new reviewer reaches an acceptable standard faster with anchors than with training alone.
- Disagreement about a score becomes a disagreement about which anchor applies.
- Anchors give an appeal panel a reference point outside the reviewer's memory.
How an anchor example works
- Real submissions are collected from a previous cycle and anonymised.
- A calibrated panel scores them and records the reasoning for each score.
- One submission per point on the scale is selected as the anchor for that point.
- The annotation names the features that place it at that point rather than the adjacent one.
- Anchors are presented alongside the criterion at the moment of scoring.
- Each anchor is reviewed at the start of a cycle and replaced when the applicant pool shifts.
How an anchor example is evaluated
An anchor earns its place by moving reviewers towards each other. Measure the following before and after anchors are introduced:
- agreement between reviewers on submissions near each anchor
- frequency with which each point on the scale is actually used
- time a reviewer takes to settle on a score
- proportion of scores a calibration panel subsequently revises
- drift in scoring between the opening and closing weeks of a cycle
Limitations, risks and common failure modes
- Anchors taken from an unrepresentative year teach reviewers the wrong standard.
- A single anchor per point leaves the space between points undefined.
- Reviewers match on surface features of the anchor rather than on the criterion.
- Anchors are written once and then left unchanged as the applicant pool changes.