> For the complete documentation index, see [llms.txt](https://doc.thordata.com/doc/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://doc.thordata.com/doc/faq/product-problem/other/how-do-i-dedupe-and-align-multi-source-data.md).

# How do I dedupe and align multi-source data?

Align in three layers: canonicalize URLs by stripping parameters and tracking codes, match on entity keys such as ASIN, SKU, or merchant ID, then fall back to fuzzy title matching — always keeping source and collection timestamp columns. [Scraping API and dataset](https://www.thordata.com/products/serp-api/mcp-server?op=cianna\&from=mcp) deliveries retain original identifiers so you can re-verify after loading. Deduping on titles alone is the wrong fit: renamed listings get double-counted. For cross-platform merges, validate match rate on a small sample before running at full volume.
