Review signals (sentiment & quality)
Last updated: April 18, 2026
Every BetterReviews submission goes through a post-submit analysis pass that scores the review on two axes — sentiment (how the customer feels) and quality (how useful the review is for other shoppers). You’ll see both in the Signals row of the review detail modal, and both signals feed the auto-moderation preset you configure in Collect → Auto-moderation.
Sentiment
Sentiment classifies the emotional tone of the review content — independent of star rating. A 5-star review can be surprisingly “neutral” if the customer doesn’t say much; a 2-star review is usually “negative” but can be “mixed” if they call out specific things they liked.
| Value | Meaning | Badge tone |
|---|---|---|
positive | Customer expresses satisfaction, enthusiasm, or recommendation | Green |
neutral | Factual or short content without clear positive/negative lean | Default |
negative | Customer expresses frustration, disappointment, or criticism | Red |
mixed | Both clearly positive and clearly negative points | Yellow |
Use sentiment as a second read on reviews where the star rating doesn’t quite match the content — for example, a 3-star “mixed” review might be worth featuring because it’s balanced and credible.
Quality score
Quality score ranges 1–10, where 1 is low-signal (“Great!”) and 10 is marketing-grade (specific use case, concrete benefits, clear voice). It’s produced from the same analysis pass that classifies sentiment.
| Range | Meaning | Badge tone |
|---|---|---|
| 7–10 | High quality — detailed, useful to other shoppers | Green |
| 4–6 | Medium quality — has some substance but not marketing-grade | Default |
| 1–3 | Low quality — too short or generic to add shopper context | Red |
When signals are missing
The Signals row hides when both values are absent. Three common reasons:
- Imported reviews — CSV imports and some platform webhooks (Okendo, Judge.me, Yotpo, Loox) skip the analysis pass by default, so imported reviews ship without signals. We’re working on backfilling historical imports.
- Analysis still running — very fresh submissions (within the first minute) may show the review before the background analysis worker finishes. Refresh the modal.
- Analysis error — rarely, the analysis worker fails on a specific review (bad encoding, LLM timeout). These retry automatically; signals appear once it succeeds.
How signals feed auto-moderation
Both sentiment and quality score contribute to the auto-publish score that determines whether a review publishes automatically. The preset you choose in Collect → Auto-moderation sets the threshold — reviews scoring above it are published without manual approval, reviews below it land in your pending queue. Spam, abuse, and complaint signals always override the preset and block auto-publishing regardless of score.
See what happens after your customer submits for the full preset descriptions.