The records you import into a matching model are the reference data it matches incoming messages against - your product catalog, account list, or any other set of entries. This article explains how to import, update, and re-import those records, and how the unique identifier works.
If you are setting up a matching model for the first time, start with Set up your first matching model.
Importing records
You import records by uploading a file from the records area of the matching model. Tekst accepts:
- CSV files
- JSON files
Each row or object becomes one record, and each column or field becomes an attribute the model can use when deciding on a match. Large reference sets are supported, so you can import a substantial catalog in a single file.
The unique identifier column
One column must be designated as the unique identifier: the value that identifies each record. This matters because every match, test item, and correction refers to a record by that value. Choose a column whose values are unique and stable, such as a product code or account ID, rather than a name that might be duplicated.
You select it from the model's settings, under Unique identifier. The list only becomes available once you have uploaded records, since the options are the columns found in your file. You must set the unique identifier before you can add test items to the model.
Updating your records
When your master data changes - new products, retired items, corrected details - upload the file again to bring the model up to date. Use Replace in the model's settings to swap in the new file. Tekst processes the new data and switches the model over to it so that matching reflects your latest records.
What happens to test items when records change
If you re-import and a record that a test item pointed to no longer exists, that test item can no longer be scored against a valid record. Tekst flags these items so you can review them and either point them at the correct record or remove them, keeping your accuracy figure trustworthy.
Tips for good records
- Keep the unique identifier unique and stable across imports.
- Include the descriptive columns that genuinely help identify a record; extra context helps the model choose correctly.
- Re-import whenever your source data changes so the model never matches against stale records.
Related articles
- To guide matching beyond the records themselves, see Using domain knowledge to guide matching.
- To understand how matches are scored, see How matching accuracy is measured.
0 comments
Please sign in to leave a comment.