- A category's complaints only show up when you read every brand at once.
- The 1–2 star share from the histogram beats any hand-picked sample.
- A recurring complaint across five brands is a gap, not a bad batch.
- Conclusions are tied to review text a person can open and check.
Why one product's reviews mislead
Read one listing's reviews and every complaint looks like that brand's problem. Read the whole category and some of them turn out to be everyone's problem. The second kind is the one worth building a product around.
So product review analysis here starts with a category scan. Up to 50 listings come back with their review text. The rollup names each recurring aspect and counts how many brands it appears under.
- Example: "handle gets hot" under 6 of 10 brands is a category gap.
- Example: "arrived dented" under 1 brand is that brand's packaging.
The star histogram, not the average
A 4.3 average hides very different products. One has a steady spread of fours and fives. Another is loved by most buyers and returned by a loud fifth of them.
Each listing carries its full star histogram, and the scan reports the share of ratings sitting at one and two stars. That share comes from all the ratings, not from the handful of reviews a page displays. It is the fastest way to spot an incumbent that sells well and disappoints often.
Tip:Sort a scan by 1–2 star share, then read only those reviews. That is where the positioning angle usually is.
Large-scale review analysis you can audit
Large-scale review analysis usually ends in a sentiment score. The score is fast to read and impossible to argue with, because the reviews behind it are gone.
Here the reading is done by your model, and the reviews stay attached. The insights card draws that read beside the listings. A person can click a product and check any claim against the text it came from. If the read cites an Amazon product the scan does not contain, the card flags it.
Amazon's aspects, and ten more sites
On Amazon the scan also returns Amazon's own review-aspect counts — the aspect tags buyers see, such as quality or value. On Temu, AliExpress, Mercado Libre and the rest it returns reviews and histograms per product.
Turning review text off makes a scan faster and cheaper, but leaves only the aggregates. Leave it on when the question is why, not how many.
Frequently Asked Questions
How many reviews can Nooticr analyze at once?
A category scan collects up to 50 listings with their review text, so a single scan typically covers hundreds of reviews. Each listing also carries its full star histogram, which covers every rating, not only the reviews collected.
Does it give a sentiment score?
No. It groups reviews by recurring aspect and counts brands, then your model reads the text. A score would replace the reviews you came to read.
Can I analyze reviews for one specific product?
Yes. get_amazon_product and get_marketplace_product return one listing's price, rating, full star histogram and reviews for 3 credits.