Setup Considerations: rating scales

What this achieves

A rating scale is the set of levels a manager picks from when rating someone’s performance. It is the vocabulary every review, calibration session, and rating history is expressed in, so it is worth settling before a cycle is built on it.

Decision 1: how many levels

The number of levels decides how much distinction you are asking managers to make.

Option	Description
Fewer levels	Easier to apply consistently. Less able to distinguish between people, which matters if ratings feed pay decisions.
More levels	Finer distinctions, and more disagreement about where the boundaries fall.

Whatever you choose, at least two levels are required. Beyond that the constraint is not technical but human: managers have to be able to tell your levels apart and describe the difference to an employee.

Example: a five-level scale at HC Corp UK Ltd gives Olivia Bennett room to distinguish a strong performer from an outstanding one. A three-level scale would collapse both into the same answer.

Decision 2: whether there is a middle

An odd number of levels has a middle, and a middle attracts ratings. An even number forces a judgement on either side of centre.

Neither is right. An odd scale is honest where most people genuinely are performing as expected; an even scale is useful where you suspect managers are avoiding a decision. Decide which problem you have.

Decision 3: what each level means

Levels carry a label, a value, and an optional description. The description is where a scale earns its consistency, because “Exceeds expectations” means different things to different managers until somebody writes down what it means here.

Write descriptions in terms of observable behaviour, not adjectives. A description that says what someone at that level actually did is usable; one that restates the label is not.

Example: describing a level at HC Corp UK Ltd in terms of the scope someone handled and the outcomes they delivered gives James Whitfield’s manager something to judge against. “Very good performance” does not.

Decision 4: where the acceptable line sits

Every level carries an Acceptable switch, on by default. Turning it off marks that level as below acceptable, and an employee rated there becomes eligible for a performance improvement plan. This is the one decision on a scale that reaches outside performance, so it deserves more thought than its position in the editor suggests.

Option	Description
Leave every level acceptable	The scale never marks anyone. Whether someone goes on a plan is decided entirely off the back of the rating, by people, with no signal from the system.
Mark only the lowest level	The conventional line. It surfaces the clearest cases and leaves everything above them to judgement.
Mark the lowest two	A wider net. It surfaces more people, including some whose rating means “below where we want” rather than “failing”.

Eligibility is not enrolment. Nobody is placed on a plan by being rated, and no rating removes the decision from the people who make it — but the line you draw here decides whose name comes up.

Draw it where you would be comfortable defending it in every case it catches, because it will catch every case equally. A line drawn one level too high converts an ordinary weak year into a formal process, and that is the version people remember.

Example: HC Corp UK Ltd marks the lowest level of its five-level scale as below acceptable and leaves the rest acceptable. An employee at the second level is a conversation for their manager, not an eligibility flag.

Decision 5: which scale is the default

One scale can be the default, and it is auto-selected when creating a new cycle. Where most of your cycles use one scale, making it the default removes a decision from every cycle setup and a mistake with it.

Decision 6: retiring a scale

A scale cannot be deleted while it is in use by an active or draft cycle. Deactivating is the route for a scale you have stopped using: it is hidden from cycle setup but stays attached to existing cycles.

That is the behaviour you want. A scale detached from a completed cycle would leave historic ratings meaningless — a “4” with no scale behind it is a number, not a rating.

Decision 7: changing a scale that has history

Changing labels or level meanings on a scale that has already been used changes how past ratings read, because the rating records point at the scale.

Where the meaning genuinely needs to change, create a new scale rather than editing the old one. Comparing ratings across a scale change is a comparison you should make deliberately, not one the system should hide from you.

What is expensive to change later

Decision	Cost of changing later
Number of levels on a scale in use	Highest. Ratings already given sit on the old levels, and cross-cycle comparison breaks.
Level labels or meanings with history	High. It rewrites what past ratings appear to say.
Changing scale mid-cycle	High. Reviews already submitted used the previous scale.
Where the acceptable line sits, once cycles have run	High. It changes who is eligible for a plan, and past ratings do not re-read themselves.
Which scale is the default	Low. It affects the next cycle you create.
Deactivating an unused scale	Low.

What happens next

Create the scale, then build a cycle on it. The cycle names the scale it uses, and that choice flows through self reviews, manager ratings, calibration adjustments, and the permanent rating records written when the cycle is finalised.