Review signal — GET /v1/products/{id}/reviews

Review signal

Aggregate sentiment across sellers, decomposed into the attributes people actually complain about — not a single star average that hides everything.

GET /v1/products/{id}/reviews

Request

No parameters beyond the path and your bearer token.

Request & response

GET /v1/products/prd_8Fk2xQ/reviews

→ 200
{ "aggregate": { "score": 4.4, "count": 28417, "sellers": 4 },
  "by_attribute": [
    { "attribute": "noise_cancelling", "sentiment": 0.91, "mentions": 6210 },
    { "attribute": "comfort",          "sentiment": 0.78, "mentions": 4102 },
    { "attribute": "call_quality",     "sentiment": 0.41, "mentions": 1877,
      "note": "Consistently the weakest attribute across all sellers" } ],
  "divergence": { "note": "One seller's average is 0.6 higher than the others — treat with caution." } }

Notes

Attribute decomposition is what an agent needs. A 4.4 average tells it nothing; 'noise cancelling excellent, call quality poor' lets it match the product to the stated intent.

`divergence` flags when one seller's reviews disagree sharply with the rest, which is usually a sign of incentivised reviews rather than a better product.