Review CSV workflow

1. Prepare and map your export

Start with a UTF-8 CSV you are authorized to use, with unique headers and one record per review. Review text is required; rating, date, review ID and source are optional. Use Download CSV template in the analyzer, or Try an example for fictional reviews.

Import the file and check each column choice. Familiar headers may be suggested; ambiguous names require your choice. Map each column only once. Ratings must be whole numbers from 1 to 5 and dates use YYYY-MM-DD. Blank optional values stay unknown. Check the input limits first.

2. Count, then read the original context

Choose Analyze reviews. Repeated IDs are compared within each case-sensitive source. Identical normalized records retain the first occurrence; conflicting versions are all excluded. Records without IDs remain separate. The ID-check panel identifies kept and excluded record numbers; record 1 is the first row after the header.

The average uses rated reviews only. Unrated reviews still contribute to topics. Monthly trends use dated reviews; missing months do not establish zero reviews. Topics match keywords, including negations: “Not helpful” still mentions “helpful.” These are counts, not sentiment judgments.

3. Filter with the denominator in view

Date endpoints are inclusive. A date range excludes undated records. Source/date settings apply after analyzing again; ID checks happen before those filters. Reconcile imported records, ID exclusions, filter exclusions and included reviews.

Selecting a word under Word counts shows the whole-word matches behind its count, using the same Unicode letter/number boundaries, normalization and case folding as the count. Repeating a word within one review counts that review once. For example, “easy” does not include “uneasy” or “easygoing.” A visible word filter can be cleared; typing in Search replaces it with a partial-text search. Topic and rating choices narrow the current list. These controls affect browsing only. They do not change overview totals or export scope. Read the matching reviews before interpreting a count.

4. Choose the right export

• Export HTML report: a standalone HTML brief with counts, rating and monthly tables, topic rules and up to three source excerpts per topic in original order. This bounded excerpt is not a representative sample. Read or print it offline; it is not a restorable session.
• Export JSON report: JSON counts, rules, included/excluded evidence, source records and the original CSV's SHA-256 hash.
• Export reviews CSV: all analyzed reviews, regardless of browsing filters. Formula-like cells receive a leading apostrophe.
• Save session: original CSV and settings for restoration. Reports and review CSVs are not session files.

Exports contain review content and are editable, not authenticated records. Keep them private and read save and recovery instructions before leaving.