Methodology

The Top Rated list is sorted by the lower bound of the Wilson score confidence interval at 95% confidence (z = 1.96). Given n ratings with mean r on a 1–5 scale the lower bound penalises small sample sizes. A single 5-star review produces a lower bound near 0; fifty reviews averaging 4.5 produce a lower bound around 4.3. The ranking rewards consensus among many reviewers over isolated perfect scores. (z = 1.96; n ratings, mean r on a 1–5 scale):

p = (r − 1) / 4

score = (p + z²/(2n) − z√(p(1−p)/n + z²/(4n²))) / (1 + z²/n)

See Evan Miller, "Ranking Items With Star Ratings" for the derivation.

Reviews score nuggets on five dimensions, each on a 1–5 integer scale. Only Overall feeds into the Wilson ranking. Dimension scores are aggregated independently and displayed on each chain page as toggleable distribution charts.

Per-chain standard deviation (σ) is the sample standard deviation of overall ratings. Low σ indicates reviewer consensus; high σ indicates a divisive product.

σ = √(Σ(x_i − x̄)² / (n − 1))

The NCI is the arithmetic mean of overall ratings across all reviews on the site, recalculated per request. The ± value is the 95% confidence interval of that mean. Consumer sentiment is the proportion of thumbs-up votes site-wide.