Datasets and write rules
A recipe decides what to read. A write rule decides which dataset and columns receive it, and which record to update. A dataset is a table of captured results: each row is a record and each column is a field.
Create your first write rule
-
Capture one page first. Run a suitable Extract recipe in the extension and review the results. Choose an existing dataset or create a new one.
-
Review field mappings. Create a write rule and map recipe fields to existing or new dataset columns. Check meanings and types: a follower count, for example, belongs in a number column.
-
Choose a match field. To keep updating the same object, use a stable identifier such as a creator or product ID. Avoid display names that can change or belong to more than one person.
-
Save and confirm the write. Saving a rule and writing data are separate actions. Follow the page prompts to return and select Write with this rule, or use Save and write. After a successful write, check the records in the dataset.
Append or update
| Write mode | What happens on the next capture |
|---|---|
| Append records | Each new execution adds records, preserving separate capture results. Retrying the same submission does not add duplicates. |
| Update matches or add records | No match adds a record; one match updates it. Multiple matches prevent the write until you resolve the conflict. |
For example, creator-01 has 10,000 followers on the first capture and 10,500 on the next. Matching by creator ID updates the same row to 10,500. Appending keeps two rows.
Reuse rules and enrich records
Saved rules work with single-page capture, Bulk add and API submissions. Each source recipe needs its own write rule, and several rules can write to the same dataset.
To enrich one record from another source, both rules must map to a stable identifier with the same meaning. IDs from different websites are not automatically equivalent. Review mappings after changing a rule or recipe version, and export a fresh API configuration.
If a field is empty, check whether the source provides it and whether capture succeeded. A partially successful capture does not refresh every field, and the record’s capture time does not mean every column was updated together.