Amazon Review Tone Predicted Sales, and Negative Reviews Did the Work

A 2023 study of 18.8 million Amazon reviews found that the tone of review headlines predicted company sales, earnings and stock returns, and that negative wording carried the signal.
data-science
entrepreneurship
Author

Kabui, Charles

Published

2026-10-24

Keywords

consumer-reviews, sentiment-analysis, text-as-data, stock-returns, earnings-surprises

Download as PDF

The tone of Amazon review headlines predicted a company’s sales for that quarter, in a 2023 study of 18.8 million reviews covering 292 public companies between 1996 and 2014. Shunyao (Cynthia) Jin scored each headline, the short summary line above the review text, with a simple text classifier that marked it positive, negative or neutral, then averaged the scores per company per quarter. Amazon reviews run positive, so she measured the change from the previous four quarters instead of the level. A move from the 25th to the 75th percentile of that change came with 9.6% higher sales in the same quarter and earnings 13.9% further above analyst forecasts. Star rating, review count and how widely the ratings varied were all controlled for.

Less negative headlines went with higher sales; positive wording showed no relationship at all. For companies in the more competitive half of the sample, the same tone move lined up with 15.2% higher sales, and in the less competitive half the estimate was indistinguishable from zero. A portfolio that bought the top quarter of companies by tone change and sold the bottom quarter earned 0.60% a quarter. Improving tone also lowered measured stock risk.

Jiekun Huang’s 2018 study found the star rating pays by itself: a portfolio buying high-rated and selling low-rated stocks earned 0.56% to 0.73% a month. Jin kept the rating in every model, so the 9.6% sales difference is what the text adds on top of the number.

Read More: Vertus Says Its Brain-Like AI Returned 51% Trading. What Can You Verify?, on what to check before trusting a trading return.

Sources:

Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service

Reuse

GNU GENERAL PUBLIC LICENSE v3.0(View License)

Citation

BibTeX citation:
@misc{kabui2026,
  author = {{Kabui, Charles}},
  title = {Amazon {Review} {Tone} {Predicted} {Sales,} and {Negative}
    {Reviews} {Did} the {Work}},
  date = {2026-10-24},
  url = {https://toknow.ai/posts/review-tone-predicts-firm-fundamentals/},
  langid = {en-GB}
}
For attribution, please cite this work as:
Kabui, Charles. 2026. “Amazon Review Tone Predicted Sales, and Negative Reviews Did the Work.” https://toknow.ai/posts/review-tone-predicts-firm-fundamentals/.