What changes when the scale has five steps instead of ten

Switching from ten points to five stars rarely changes a rating's verdict - the ordering survives - but it changes how precise the result looks. A 7.4 out of ten and three and a half stars say nearly the same thing; only the decimal invites you to debate the second digit.

What a five-point scale can distinguish

Five stars, with half-star increments, gives ten effective positions - the same count as a ten-point scale reporting whole numbers only, and fewer than a ten-point scale reporting one decimal. What a five-star display cannot do is separate a 7.2 from a 7.6. Both round to the same three-and-a-half stars, and the tool has quietly declined to tell you they were ever different, which is a loss of information compared to a decimal scale - but only if that decimal was carrying real information in the first place, which the previous posts in this series should make you doubt by default.

What that costs, and what it buys

The direct cost is resolution: a coarser scale genuinely cannot express a small difference that a finer one can. Survey research puts rough limits on how much resolution helps. Preston and Colman (2000) had 149 people rate the same services on scales of 2 to 11 points and found reliability and validity improved with more categories "up to about 7," while test-retest reliability tended to fall beyond 10 - and respondents liked the 10-point scale best. Two submissions that a ten-point tool would separate by two tenths land on the identical star rating, and if that gap was real, the coarser scale has erased it.

A three-and-a-half-star result also invites a different reflex from the reader: nobody sits and debates whether it should really have been three and three quarters, because the scale visibly does not offer that option. A 7.4, wearing a decimal, invites exactly that debate even when the underlying process supports it no better.

The benefit is that a coarse scale cannot claim more precision than it has. A five-star display with half-star steps is visibly a ten-position scale and reads like one - nobody mistakes three and a half stars for a laboratory measurement. A 7.4 out of ten, by contrast, looks exactly like a number with two significant figures, even when the underlying process was never that stable between two identical submissions. The star scale is honest about its own coarseness by design; the decimal scale has to be made honest by the reader, who has to know to distrust the second digit.

Does the verdict actually change?

Rarely, in the sense that matters. A submission that scores well on a ten-point tool will almost always score well as stars too - the ordering a reader cares about, roughly where a result sits relative to the rest of a population, survives the conversion in almost every case. What changes is not the verdict but the confidence a reader assigns to it. A 7.4 invites you to treat the difference between it and a 7.6 as meaningful. Three and a half stars invites no such comparison, because the scale simply does not offer the resolution to make it.

Converting between them

A rough conversion - halve the ten-point score to get a five-point one - preserves ordering reasonably well but not much else, because the two scales are rarely calibrated against the same underlying distribution. Comparing two tools properly means converting to rank within each tool's own results, not converting the raw number, and that holds whether the two tools use ten points, five stars, or anything else - the scale is a display choice, and display choices do not make two different calibrations comparable just because the arithmetic is easy.

Where the choice tends to land

Tools built around a single quick impression - the shape closer to a human reviewer's response than to a detailed breakdown - tend toward stars, because stars fit the size of the claim being made. Tools that want to look analytical tend toward the decimal, whether or not the underlying process supports the extra digit. Neither choice says anything about the model doing the actual estimating underneath either display, and neither is a substitute for an actual unit - a length still needs a tape and a stated method regardless of how many stars or points sit next to it. Rate Cock uses the decimal convention like most of the category, but pairs it with a visible six-axis breakdown, which does more to earn the extra digit than the digit does on its own.