What kind of structured data input does Bulkbase support?

CSV is the supported input format

Bulkbase accepts structured product data as CSV files. These are comma-separated, tabular files with a header row that names the columns and one row for each product or variant.

So what kind of structured data input does Bulkbase support? CSV files, and only CSV files. The list of bulkbase supported input formats does not include standalone .json files, although Bulkbase can process JSON data stored inside an individual CSV cell.

This CSV-based approach is system-agnostic. You can prepare data exported from shop systems such as Magento, Shopify, or WooCommerce, as well as from a PIM, ERP, CMS, or spreadsheet. The source system only needs to support CSV export and import.

How columns become prompt variables

After upload, each CSV column header becomes a variable that prompts can reference. Variables use double curly braces, such as {{ product_name }} or {{ brand }}.

Bulkbase processes each row with the values found in that row. If two products have different names, brands, or materials, the same reusable prompt receives the correct context for each product.

A prompt can combine several source variables into one generated field. For example, it might use {{ material }}, {{ color }}, and {{ category }} to create a product description that reflects the available attributes.

The same source columns can also feed multiple output fields. A product name, brand, and category could provide context for a description, an SEO meta title, and a translation without requiring separate source files.

What a Bulkbase-ready CSV looks like

A suitable file starts with a stable identifier such as SKU or product_id. This identifier keeps generated results tied to the correct product when the enriched file is reviewed, merged, or imported into another system.

Use one row per product or variant and keep column headers consistent. Include the attributes required by your prompts, such as name, brand, material, category, dimensions, and features. Unused columns can be removed before upload to keep the working dataset focused.

Follow common CSV conventions when formatting bulkbase structured data. Under RFC 4180 conventions, fields containing commas, line breaks, or double quotes should be enclosed in double quotes. Double quotes inside a quoted field are escaped by doubling them.

Generated results are written into new output columns while the rows remain aligned with their source data. Structured outputs enforce a strict CSV schema, and write modes determine whether results fill empty cells, overwrite existing values, or append content. This makes the enriched CSV suitable for controlled import back into a shop, PIM, or ERP.

Preparing your input data

Start by exporting only the rows and fields needed for the task. A smaller, focused file is easier to inspect and reduces the chance that irrelevant source data enters a prompt.

1. Clean the file: Open it in a spreadsheet and check for typos, missing values, inconsistent headers, and mixed attribute formats.
2. Confirm row identity: Make sure every product or variant has a stable SKU or product ID.
3. Test a small batch: Run about 5 to 10 rows before processing the full catalog. Check whether variables resolve correctly and whether output columns contain the expected content.
4. Review exceptions: Inspect flagged rows and correct missing or inconsistent source values before continuing.
5. Complete the round trip: Export the enriched CSV and import it into the target shop, PIM, or ERP.

This preparation turns the bulkbase csv import into a controlled data workflow. The original identifiers and row structure remain available for checking generated fields before they enter a production catalog.