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 the quality score also powers the optional auto-approve rule in Collect → Auto-approval.

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.

ValueMeaningBadge tone
positiveCustomer expresses satisfaction, enthusiasm, or recommendationGreen
neutralFactual or short content without clear positive/negative leanDefault
negativeCustomer expresses frustration, disappointment, or criticismRed
mixedBoth clearly positive and clearly negative pointsYellow

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.

RangeMeaningBadge tone
7–10High quality — detailed, useful to other shoppersGreen
4–6Medium quality — has some substance but not marketing-gradeDefault
1–3Low quality — too short or generic to add shopper contextRed

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-approve

The “Auto-approve high quality” rule in Collect → Auto-approval uses the quality score with a merchant-configurable threshold (default 7). Reviews scoring at or above the threshold AND with positive sentiment auto-publish without needing your manual approval. See what happens after your customer submits for the full auto-approve rules.