<- all tokdocs

Can Amazon Review Tone Predict Stock Returns? The Study and a Claude Test

By , founder of Mavric and Apso

View on TikTok ->

A peer-reviewed study by Shunyao (Cynthia) Jin of the University of Michigan-Dearborn found that the tone of Amazon product reviews predicts a company's next-quarter sales, earnings surprises and stock returns on average, and that negative tone is more informative than positive tone. A TikTok from Angus the Nontechnical (@angusthenontechnical) cites it correctly, then overstates it as "almost always" true. The video's own Claude demo, on 94 reviews of one Logitech mouse, records a miss on screen.

The video itself says "I am not a financial advisor," and this is not financial advice either.

What does the Amazon review tone study find?

The paper is "More than words: Can tone of consumer product reviews help predict firms' fundamentals?" in the Journal of Business Finance & Accounting, volume 50, issue 9-10, pages 1910-1942, published online 28 December 2022. Its Crossref record lists the author at the College of Business, University of Michigan-Dearborn, a separate campus from Ann Arbor.

The abstract says, in part: "I find that aggregate review tone successfully predicts a firm's forthcoming quarterly sales. Moderating analyses show that this predictability is stronger for firms operating in a highly competitive environment. I further find that review tone predicts a firm's quarterly earnings surprises, abnormal stock returns and risk." It ends: "negative review tone is more informative and useful than positive tone in predicting a firm's fundamentals." The video circles that last sentence on screen, word for word.

Unpaywall lists the published version as open access under CC BY-NC.

Does negative review tone "almost always" predict a sales drop?

No. The video says a sudden negative tone "will almost always precede a drop in sales and stock returns." The abstract describes an average relationship across many firm-quarters, stronger in competitive industries, and reports no hit rate. Results that are stronger in some settings than others are not a near-universal rule.

The direction is right. Lower tone is associated with weaker next-quarter sales on average. The certainty is not supported.

A 2026 Stockholm School of Economics paper by Lindé and Nyström summarizes Jin's finding as showing that "review tone on Amazon.com predicts a firm's abnormal stock returns and risk," with the effect on stock prices running through earnings. It also notes that Jin measures abnormal tone against a firm's own trailing average, the same approach as the video's "prior four quarter mean" chart.

How the video's Claude workflow works

The proposed pipeline:

  1. Pick 10 to 20 companies that sell heavily on Amazon.
  2. Have an AI collect two years of their reviews.
  3. Have a smaller model score each review's tone and tag the reason.
  4. Average tone per quarter and compare it with quarterly SEC filings and stock prices.

The Claude conversation in the video, titled "Logitech M510 review analysis," proposes a basket of Logitech, SharkNinja, Helen of Troy, Hamilton Beach Brands, Newell Brands, Spectrum Brands, Sony, Procter & Gamble, Colgate-Palmolive, Church & Dwight, Clorox and YETI. It scores reviews on a -2 to +2 scale. One two-star review from September 22, 2026, headed "Modern Versions Are Inferior," scores "Tone -1."

The collection step is the hard part, and the video does not show it. The only sample is 94 reviews of one Logitech M510 mouse, and Claude's own output tells the user to "check each company's Amazon share of sales and confirm that dated reviews can be accessed."

Did the Claude test predict Logitech's results?

No, and the video says so on screen. Its chart, "The negative swing / This quarter minus the prior four quarter mean," shows three positive quarters (FY26 Q3, FY26 Q4, FY27 Q1) and one negative quarter (FY27 Q2). Next to it, the overlay text reads "...but next-quarter revenue rises" and "Record the miss."

The revenue figure checks out. Logitech's Q4 fiscal 2026 results, released May 5, 2026, state "Sales were $1.09 billion, up 7 percent in US dollars and 3 percent in constant currency compared to Q4 of the prior year."

The one negative swing, FY27 Q2, could not be tested when the video posted on October 8, 2026. Logitech scheduled those results for October 27, 2026. The chart's own caption reads "Sparse sample. Not a company forecast."

Did Claude predict Nike's stock crash?

The video opens with "Claude predicted Nikes downfall," but it shows no Nike analysis. The Claude basket contains no Nike. The Nike material is a stock quote and a Reuters headline from June 28, 2024, "Nike stock plunges as gloomy sales forecast fans growth concerns."

That crash was real. Al Jazeera reported the same day that "It was the worst day ever for the stock, which slumped 20 percent on Friday, with the losses wiping out $28.41bn from the company's market valuation." It happened more than two years before the video. Nike is also a weak fit for this method, since it stopped selling wholesale on Amazon in 2019 and resumed direct sales there only in May 2025.

Is there other research on reviews and stock returns?

Yes. Jiekun Huang of the University of Illinois at Urbana-Champaign published "The customer knows best: The investment value of consumer opinions" in the Journal of Financial Economics in April 2018. It uses Amazon star ratings rather than tone and reports a long-short portfolio earning roughly 55.7 to 73.0 basis points per month.

Bottom line

The research is real: Amazon review tone carries information about upcoming sales and returns, especially negative tone in competitive industries. It is an average effect, not a rule that holds for each company. The video's own one-product Claude test logged a miss and had not yet been checked on the quarter that mattered.

Sources

Published October 8, 2026. Based on a TikTok by @angusthenontechnical. Full video notes (markdown)