Incentive Design Entry #0389 Classified Declassified

The reason a rating system inflates over time

Nobody in the system gains from a low score. When every participant benefits from inflation, the scale compresses upward until it stops distinguishing anything.

No visual record attached The written record below is complete.
Plate 270 — The scale that ran out of range

Intuition test — answer before you read on

A five-star scale drifts above 4.7 and functions as two points. What drives the compression?

A marketplace launches a five-star scale. Within three years the average listing sits above 4.7, and buyers have learned to treat anything below 4.5 as a warning. The five-point scale now operates as a two-point scale, and the compression happened without any rule change.

What everyone sees

This is often described as reviewers being too generous, which locates the cause in individual softness. But the pattern appears across marketplaces with different populations and different review mechanics, which suggests something structural rather than a coincidence of temperament.

What is actually happening

Every actor faces an incentive pointing upward. Sellers lose sales below the threshold, so they solicit high ratings and dispute low ones. Buyers who leave a low rating face possible retaliation in two-sided systems, which Bolton, Greiner and Ockenfels showed suppresses negative feedback. Platforms benefit from high averages because they signal a healthy market. Dellarocas and Wood documented that dissatisfied users disproportionately decline to rate at all, biasing the sample upward before any generosity is involved. With no participant advantaged by accuracy at the bottom of the scale, the distribution drifts up until the informative range is exhausted.

Why it stays hidden

The inflation hides because each rating is individually defensible. A seller who provided acceptable service has not earned three stars in any ordinary sense, and a buyer who withholds a complaint has done nothing wrong. The scale degrades through the accumulation of reasonable acts, which is why no participant can identify the moment it stopped working or a party responsible for it.

A rating scale does not collapse because reviewers are kind. It collapses because nobody in it profits from a low number.

A rating scale does not collapse because reviewers are kind. It collapses because nobody in it profits from a low number.

The hidden part — entry #0389

Collect this card

A rating scale does not collapse because reviewers are kind. It collapses because nobody in it profits from a low number.

0 / 10,000 collected

Sources & further reading 2
  1. Bolton, Greiner & Ockenfels — engineering trust: reciprocity in the production of reputation information
  2. Dellarocas & Wood — the sound of silence in online feedback

Circulate this file

Annotations are reserved for archive members.

Sign in to annotate