How the rankings work, and why a 4.5 can beat a 5.0
Chain A has one review. Five stars. A flawless 5.0.
Chain B has fifty reviews. Most are 4s and 5s, a handful of 3s. Average: 4.5.
Chain B ranks higher. Every time.
Here’s why: that single 5-star review on Chain A could be the owner’s mum. It could be a fluke. Statistically, with one data point, we genuinely cannot tell whether the “true” rating is a 5 or a 1. There’s just not enough information.
Chain B’s fifty reviews paint a picture. Even with a few 3s dragging the average down, we can say with 95% confidence that the true rating is somewhere between 4.3 and 4.7. That’s reliable. That’s useful.
So: Chain B at 4.5 outranks Chain A at 5.0. Not because we don’t trust the 5 — but because we don’t know yet. Get Chain A ten more reviews and it might shoot to the top. But it has to earn it.
The “Top Rated” list is sorted by the lower bound of a 95% Wilson confidence interval. More reviews = narrower interval = higher lower bound. Fewer reviews = wider interval = lower lower bound, even if the average is high.
This method was popularised by Evan Miller. His article explains the statistics in detail →
For a chain with average rating r (1–5) and n reviews:
p = (r − 1) / 4
score = lower bound of Wilson( p, n ) at 95% confidence
Each review scores nuggets across five dimensions:
Flavour — how good does it taste?
Mouthfeel — texture, tenderness, chew
Coating — crispness, adhesion, breading quality
Sauces — quality and variety of accompanying sauces
Overall — the holistic nugget experience
The Overall score is used for the Wilson ranking and the Top Rated list. On each chain’s page you can view the rating distribution for any dimension — toggle between Overall, Flavour, Mouthfeel, Coating, and Sauces to see how reviewers scored across each category. This lets you find a chain that excels at what you care about, whether that’s a crunchy coating or a superior sauce selection.
σ (standard deviation) measures how much reviewers disagree. A low σ means consensus: most reviewers gave similar scores. A high σ means the chain is divisive: some love it, some hate it. We compute σ from the overall ratings using the sample standard deviation formula.
The NCI is the mean overall rating across all reviews on the site, updated on each page load. The ± value is the 95% confidence interval. “Consumer sentiment” is the percentage of thumbs-up votes across all reviews.