How to Analyze Amazon Review Exports in Excel or Google Sheets

A review spreadsheet becomes useful when every row keeps its product context and the analysis separates evidence from interpretation. Start with an AMZShark CSV or Excel export containing ASIN, marketplace, product title, review URL, rating, discovery date, reviewer, review title, and review text. The Amazon product review export guide shows how to create that file.

Preserve a raw sheet

Import the file into a tab named Raw Reviews and do not edit source values. Create a second tab for helper columns and a third for summaries. Record the export filters and run date at the top of a Notes tab.

Add useful helper columns

If the rating is in column E and the review text is in column J, examples include:

• Sentiment bucket: =IF(E2<=2,"Negative",IF(E2=3,"Mixed","Positive"))
• Mentions leak: =IF(ISNUMBER(SEARCH("leak",J2)),"Yes","No")
• Review length: =LEN(J2)
• Month discovered: =TEXT(F2,"yyyy-mm")

Adjust column letters to the actual file. Treat rating-derived sentiment as a coarse label, not a substitute for reading the text.

Build a controlled theme taxonomy

Create a Theme column using a small, decision-oriented vocabulary such as reliability, fit, setup, packaging, comfort, cleaning, compatibility, value, or support. Add a Subtheme only when it changes the action. Keep an “Other/Unclear” value so ambiguous rows are not forced into a misleading category.

Create pivot tables

Normalize before comparing

Raw counts favor products with more available review rows. Add a rate such as theme matches divided by available rows for that ASIN, and always display the denominator. Do not present a rate based on a tiny sample as equivalent to a large one.

Keep source evidence visible

Create a summary table with the theme, product count, row count, representative review URL, short paraphrase, and proposed action. Prefer paraphrases over copying long customer passages. Preserve the original row and URL for verification.

Spreadsheet limits

Google Sheets has cell and performance limits, and large Excel workbooks can become fragile. For hundreds of thousands of rows, use multiple files, Power Query, a database, or Parquet with an analytical engine. The goal is a reproducible workflow, not forcing every dataset into one workbook.