Ranking subjective things is easy until two items are close. Then the criteria start moving. The café you like gets credit for atmosphere even though atmosphere was not on the original list; the bench with the better view suddenly loses points for a hard seat; an old job receives a higher score because you remember the people fondly.
A rubric does not make taste objective. It makes the judgment inspectable.
Decide what the rating is for
“Best” is not a criterion. Best for what? A park bench can be best for reading, people-watching, resting after a run, accessibility, shade, conversation, or a view. A job can be best for pay, learning, stability, autonomy or personal change. Different purposes produce different rankings.
Write the purpose at the top of the sheet before choosing criteria.
Use criteria that can disagree with each other
If every criterion rewards the same underlying thing, the rubric only repeats one preference. A useful rating system creates trade-offs. For a bench, comfort, view, privacy and shade can pull in different directions. For a bookshop, selection, browsing comfort, staff knowledge and price may not move together.
Define the point before you award it
“Good view” invites score inflation. “One point if the seated view is not dominated by a road, car park or blank wall” is clumsy but repeatable. The definition should make a later score easier, not prettier.
In the 50-bench project, the value is not that four points magically identify the world’s best bench. It is that the same four questions can be asked fifty times.
Keep the scale smaller than your confidence
A 100-point scale suggests distinctions you probably cannot defend. If you cannot explain why something deserves 83 instead of 81, do not use that precision. Binary points, five-point scales or a small set of labels often produce more honest rankings.
Write a note for ties instead of forcing decimals
Ties are information. Two items can reach the same score for different reasons. One bench may have perfect shade and a poor view; another may have the opposite. The short comparison note is often more useful than inventing 3.7 versus 3.8.
Test the rubric before rating the full set
Score five to ten examples. If one criterion never changes, it may not be useful. If one criterion dominates every result, decide whether that dominance is intentional. If you change the rubric after the test, rescore the early examples so the final table uses one method.
Record context when conditions change
A rating can depend on time, weather, crowding, price, version, location or company. Record the conditions that would matter to a later reader. A quiet bench at 7 a.m. may be a noisy bench at 6 p.m. A software tool reviewed in one version may change entirely after an update.
Separate the score from the recommendation
A high score can still be wrong for a specific person. A steep walk to a scenic bench may make it a poor choice for someone with limited mobility. A high-change job may have been emotionally exhausting. State who the ranking is for and what the rubric leaves out.
A simple reusable template
| Item | Criterion 1 | Criterion 2 | Criterion 3 | Criterion 4 | Total | Context / trade-off |
|---|---|---|---|---|---|---|
| Example A | 1 | 0 | 1 | 1 | 3/4 | Strong overall, but fails one condition that matters to some users |
The point of a fair subjective rating is not to eliminate opinion. It is to stop opinion from pretending it had rules only after the winner was chosen.
For a different kind of retrospective score, see Ranking Every Job I’ve Ever Had by How Much It Changed Me.
Leave a Reply