Parse the table before judging the products
Delimiter
Detect comma, semicolon, tab or pipe delimiters without assuming every table is comma-separated.
Headers
Require two to 200 non-empty, unique headers before rows are mapped to product fields.
Quotes and rows
Use a real streaming CSV parser with bounded record size so embedded delimiters do not silently shift columns.
Identifiers
Flag missing and duplicate product IDs after the rows are normalized.
Product values
Check supported content, URL, image, price, availability, brand, identifier and category fields.
Capacity
Stream up to 100,000 products within independent 200 MiB transfer and decompression limits.
Example CSV catalog problems
| id | title | price | brand |
|---|---|---|---|
| 1001 | Running Shoe | 99.00 EUR | Acme |
| 1002 | Running Shoe | ||
| 1002 | Running Shoe Blue | 109.00 EUR | Acme |
The first three classifications correspond to current rules. The repeated-title observation is educational and is not currently part of the score.
CSV parsing problem vs catalog data quality problem
| Layer | Examples | What to fix |
|---|---|---|
| CSV parsing | Delimiter, quotes, duplicate headers, oversized record | Export format or serialization |
| Catalog data | Duplicate ID, blank brand, malformed price, short title | Source fields and mapping rules |
| Order of work | Parsing must succeed first | Then prioritize repeated product failures |
One streaming path for structural and product checks
Deterministic checks, documented limits
Product Feed Scan parses public XML and CSV feeds incrementally, applies versioned rules to supported fields and groups failures by impact. It does not predict platform policy decisions, modify feeds or guarantee channel approval.
CSV catalog audit questions
Questions about this guide
Which delimiters are supported?+
The production parser detects comma, semicolon, tab and pipe delimiters from the header row.
What makes a CSV fail before the product audit?+
Missing or duplicate headers, unsupported delimiters, malformed quoting, oversized records and rows rejected by the CSV parser can stop processing.
Does Product Feed Scan audit Excel files?+
No. The input must be a public XML or CSV URL. Native XLSX files and direct uploads are outside the current product scope.
Are duplicate titles detected?+
Not by the current scoring rules. Duplicate IDs are detected. Repeated title patterns are shown here as an educational catalog review, not an automated Product Feed Scan finding.