Overview
When a customer sends you a document like a strata roll, compliance report, or any multi-page PDF, manually copying data into your expected format is slow and error-prone. With Vern’s document extraction, you:- Upload the document to a workbook
- Let Vern parse and understand the document structure
- Provide extraction instructions describing what data to pull
- Review a preview and import the extracted rows
Step 1: Upload your document
Open the workbook with the template you want to populate and click Import. Select the PDF or Excel file you want to extract from — you can upload it directly or choose a previously uploaded file. Click Continue and Vern will parse the document to understand its structure and content. This may take a couple of minutes depending on the size of the document.Step 2: Write extraction instructions
Once parsing is complete, Vern shows a text field with suggested extraction instructions. Edit these to describe exactly what data you want to pull from the document. For example, if you’re extracting contacts from a strata roll:Extract all contacts including all of the different types — managing agents, tenants, and owners — from all units, even if some of the values are missing. Each row should be an individual name with their relevant contact details and unit info.See Extraction instructions best practices below for tips on writing effective instructions.
Step 3: Review and import
Click Extract and Vern will process the document. In about a minute, you’ll see a preview of the extracted data. Review the rows and columns to make sure everything looks correct. If you need to adjust column mappings, you can do so before importing. When you’re happy with the preview, click Continue and then Import Data to add the rows to your sheet.Extraction instructions best practices
The extraction instructions tell Vern what data to look for and how to structure it. Better instructions lead to better results.Use a sample template as a reference
Vern already knows your template columns, so your instructions should focus on what to extract and how to structure it — not which columns to fill. For example, if your template has columns for Name, Email, Phone, and Type:Extract all people listed in the document. Each row should be one person with their name, email, phone number, and role type (e.g. owner, tenant, managing agent).
Be specific about what to include
Tell Vern exactly which types of records or sections to extract from. Vague instructions lead to incomplete results. Good:Extract all contacts including owners, tenants, and managing agents from every unit in the document, even if some fields are missing.Too vague:
Extract the contacts.
Give examples when the data is ambiguous
If the document uses inconsistent formatting or labels, provide examples so Vern knows what to look for.Extract all line items. “Levy Amount” may appear as “Quarterly Levy”, “Admin Fund”, or “Sinking Fund” — treat all of these as the levy amount.
Specify how rows should be structured
Clarify what each row in the output should represent, especially when the document groups data in a way that differs from your template.Each row should be an individual contact. If a unit has multiple contacts (e.g. an owner and a managing agent), create separate rows for each person.
Handle missing values
Documents often have incomplete data. Tell Vern whether to include partial records or skip them.Include all contacts even if some values like email or phone number are missing. Leave those fields blank rather than skipping the row.
Tips
- Large documents — Parsing time scales with document size. A 40-page PDF typically takes 1–2 minutes to parse.
- Multiple document types — You can extract from PDFs, scanned documents, and unstructured Excel files.
- Iterative extraction — If the first extraction isn’t quite right, adjust your instructions and re-extract. Small tweaks to wording can significantly improve results.
- Clean after import — Use Chat to fix formatting or fill in missing values after importing. See How to write effective prompts for guidance.
Next steps
- How to write effective prompts — Clean and transform data with natural language after extraction
- Turn unstructured Excel files into clean data — Alternative flow for Excel files with header mapping
- Templates — Set up validation rules for better extraction accuracy
- Workbooks — Learn more about workbook features