Data types · Reviews
What customers actually say. Every rating and review, on every site, structured.
Reviews are the most honest product research anyone publishes: unprompted, specific and dated. They are also scattered across dozens of retailers and platforms, syndicated, duplicated and full of personal data. Import.io collects them into one clean dataset — rating, text, verified status, helpfulness — with personal information removed and topics tagged.
| Retailer | Rating | Reviews | New 7d | Top topic | Sentiment |
|---|---|---|---|---|---|
| Amazon | 4.6 | 12,408 | 214 | bowl capacity | +0.62 |
| Walmart | 4.5 | 3,902 | 88 | motor noise | −0.18 |
| Target | 4.7 | 1,811 | 41 | ease of use | +0.71 |
| Best Buy | 4.4 | 968 | 19 | attachments | +0.44 |
| Home retailer A | 4.6 | 740 | 12 | colour | +0.58 |
| Syndicated duplicates | — | 2,106 removed | — | — | — |
The best product research is already written.
Every week, customers explain in their own words what they love and what is wrong with your product and your competitors’. Product teams find defects in them, marketers find language, and analysts find demand signals — if the reviews can be read at scale.
The obstacles are practical: reviews live on dozens of sites, the same review is syndicated to many retailers, and review text contains names and personal details. We collect, deduplicate, strip personal data and tag topics, so what’s left is signal.
- Product and quality teams
- Brand and consumer insights
- Ecommerce and digital shelf
- Customer experience
- Market research and data companies
What reviews data answers.
Six questions teams use it for most.
Product quality signals
Defects and failure modes surfaced from review text before they reach returns data.
qualityCompetitive benchmarking
Ratings, volume and topic sentiment for your products against competitors’.
benchmarkVoice of customer
The words customers use, by topic, for positioning and content.
insightReview velocity and health
Volume, recency and rating trends by retailer, with alerts on sudden changes.
velocityLaunch tracking
How a new product’s reviews develop in its first weeks on each retailer.
launchModel and analytics inputs
Clean, labelled review text for sentiment models, search and recommendation.
AIThe data, field by field.
| field | type | example |
|---|---|---|
| review_id | str | amz-R2X… |
| retailer | str | amazon |
| product_id | str | B0… |
| rating | int | 4 |
| title | str | Great mixer, loud motor |
| body | str | review text, PII removed |
| reviewed_at | date | 2026-09-22 |
| verified | bool | true |
| helpful_votes | int | 12 |
| syndicated_from | str | brand site |
| topics | list | motor noise · capacity |
| sentiment | num | −0.18 |
- Retailer product pages
- Marketplaces
- Review platforms
- Brand sites with syndicated reviews
- App stores and software review sites
The hard parts, handled.
What breaks when this is done with scripts, and how Import.io handles it.
Syndication, once
Reviews syndicated across retailers are detected and counted once, so volume and averages aren’t inflated.
Personal data removed
Names, handles, emails and phone numbers are detected and stripped before delivery.
Deep pagination
Thousands of reviews per product are paged through completely, including sort orders that hide older reviews.
Topics and sentiment
Each review is tagged with the product attributes it mentions and a sentiment score per topic.
Two ways to get it.
Same capture engine underneath each one.
Managed feeds
Insights, product and data teams who want clean review data delivered daily, under an SLA.
learn more →Self-service platform
Analysts who want to build their own review extractors.
learn more →Review text is collected from public pages with reviewer names, handles and contact details removed. Programs deliver aggregated or de-identified review data by default.
Reviews data questions.
Straight answers.
Do you remove personal data from reviews?
Yes. Reviewer names, handles and contact details are detected and removed before delivery.
How do you handle syndicated reviews?
Reviews that appear on several retailers are detected by content and metadata and counted once, with every location recorded.
Can you get all reviews for a product, not just recent ones?
Yes. Review pages are paged through completely, including older reviews behind alternative sort orders.
Do you provide sentiment?
Each review is tagged with the topics it mentions and a sentiment score per topic; raw text is available where your use allows it.
How fresh is the data?
New reviews are collected daily; priority products can be refreshed more often.