When you create a template in ExtractGrid and apply it to a large batch of PDFs or images, every file in that batch needs to follow the same layout. If page size, margins, or the position of fields changes from one document to the next, the system may read text from the wrong place.
Why layout consistency matters
Templates work by mapping fixed regions on the page to specific fields. Each bounding box targets a defined area—such as an invoice number, date, or total amount. Those regions only produce reliable results when the underlying content appears in the same position on every document.
If one scan is cropped differently, rotated, or uses a different page format, a bounding box that worked on your sample file may overlap unrelated text on another file. The extraction will still run, but the output can be incorrect or incomplete.
What to check before batch processing
Review your batch against the document you used to build the template:
- Page dimensions — All files should use the same page size and orientation.
- Text placement — Labels and values should appear in the same locations across documents.
- Scan margins and cropping — Avoid uneven borders that shift content up, down, or sideways.
- Document version — Use one consistent form or template version, not a mix of old and new layouts.
Best practices for reliable batch results
- Build your template from a representative sample — Choose a document that reflects the typical layout of the full batch.
- Standardize inputs before upload — Re-scan or re-export files so they share the same format and alignment.
- Run a small test batch first — Process a handful of files and confirm the extracted data matches expectations.
- Create separate templates for different layouts — If you receive multiple document formats, use a dedicated template for each one rather than forcing one layout onto all files.
Summary
Batch processing saves time only when every document matches the layout your template expects. Keeping size, placement, and formatting consistent across all PDFs and images is the most important step to prevent accidental extraction from unintended regions.