Amazon Review Export Format: What You Can Upload

Where uploads work today

Two upload surfaces exist, with one honest distinction between them:

• The free review pain-point word cloud — the user-facing upload. Drop a review export or paste raw review text; the file is parsed entirely in your browser and never uploaded anywhere. Output is a complaint word-frequency cloud with a shareable card.
• The R2 upload subsystem — review-export CSV storage (up to 5 MB per file) that backs the platform's file handling. It is infrastructure, not a report generator: today's AI reports are generated from cloud-scraped samples (how that works).

So: to eyeball complaint patterns in a file you already own, upload it to the word cloud. To get the full report — root-cause clustering, severity ranking, verbatim-quote anchoring — submit the ASIN and let the cloud scrape a star-balanced sample (quick start).

Supported file formats and the size limit

The word-cloud upload accepts spreadsheet files up to 5 MB per file:

• SellerSprite (卖家精灵) review exports in xlsx — both known export generations: the older "B000…-Reviews" sheet and the newer AI-comment-analysis sheet with its renamed columns. Both parse identically.
• Any generic CSV spreadsheet with a header row and one column of review text — Amazon's own review exports, other tools' exports, or a sheet you assembled yourself.
• No file at all: pasting raw review text straight into the tool works too, and there is no size pressure when you paste.

Over 5 MB? Split the export into smaller files, or paste the reviews instead. The limit exists because parsing happens in your browser — keeping it local is the privacy promise, and 5 MB covers exports of thousands of reviews.

Column requirements: how the review text is found

You do not need to rename your headers. The parser locates the review-text column in two passes: first it looks for a known column name, and if none matches, it falls back to auto-detection.

• Pass 1 — name match: a header named 内容, content, review, comment, body, text, or 评语 is used directly.
• Pass 2 — auto-detect: if no header matches, the parser samples the first 20 data rows and picks the column with the most text in it. Review exports almost always have the review body as their longest text column, so this catches custom and translated headers.

One safety rule worth knowing: any cell whose content starts with =, +, -, or @ is neutralized before parsing — the standard defense against CSV-injection formulas. Your export's formulas cannot execute.

The full review-file field spec

For reference — and for anyone preparing review data destined for the analysis pipeline — here is the complete Amazon review export format spec that Amzanalyst's review-file adapter reads. Star rating and review text are the required minimum; everything else enriches the analysis when present.

Whatever the source, review data that enters analysis passes a canonical adapter first, so downstream engines always receive one identical shape — a SellerSprite export and a platform scrape are indistinguishable by the time they reach the AI.

Preparing a clean export: three tips

• Keep the header row first. The parser expects row one to be headers; merged cells and title banners above the header can defeat name matching (auto-detect will usually still recover).
• Don't translate the 内容 column header if you can avoid it — or do, auto-detection has you either way. Only a file with no long text column at all will fail.
• For the word cloud, star ratings are optional — it counts complaint words regardless. For star-band statistics, keep the star column numeric (1–5).

Once your export looks right, the fastest validation is the tool itself: drop the file, and if a cloud renders, the parse succeeded. From there, the quick-start guide picks up where the word cloud leaves off.

Upload format FAQ

Continue reading