Dataset

Before you build anything in Proma, you need to answer one question: what are the things I'm tracking? Customers. Orders. Tasks. Each one becomes its own dataset. A dataset is simply a structured list of one type of record — every row is one item, every column is one detail about it.

Getting this structure right early makes everything else easier. A well-named column with the right type can power a time-based automation, filter a view, or populate a form. The same field stored as plain text can't do any of that.

Create a dataset

Datasets live inside systems in your workspace. To add one, click + Add New in the left sidebar and select Dataset.

You'll be asked to choose how you want to create it:

• Create from scratch — build your dataset manually, with AI-assisted column suggestions

• Create from imported dataset — upload an existing file (PDF, CSV, or XLS) and map its data into Proma

Option 1: Create from scratch

Step 1 — Name your dataset
Give your dataset a name and description. The description is used by Proma's AI to suggest relevant columns, so be specific. "Track client projects with deadlines, assignees, and approval status" will get you better suggestions than just "Projects."

Step 2 — Select columns Proma's AI suggests columns based on your description. For each suggestion, click ✓ to include it or ✗ to remove it. You can also:

• Change the column type using the dropdown next to each suggestion

• Add your own columns using + Add Column

• Click Get more AI suggestions if you want additional recommendations

Step 3 — Finalize columns Review each column one more time. Select a column from the list on the left to edit its name, type, and any type-specific settings (such as min/max values for Number columns). Click Finish when you're done.

Option 2: Create from imported dataset

Step 1 — Upload your file and name the dataset
Upload a PDF, CSV, or XLS file. Proma auto-fills the dataset name from the filename — edit it if needed and add an optional description. Click Next.

Step 2 — Select the header row
Choose which sheet to import (for multi-sheet files) and select the row that contains your column headers. Click Next.

Step 3 — Match columns to data types
Map each column from your file to a Proma data type. If a column type doesn't match your data, you'll see a validation warning (e.g., "100% of data is invalid for this data type") —resolve these before moving on.

Step 4 — Validate and import
Review a preview of your records and deselect any rows you don't want to import. Click Import Dataset to finish.

Dataset structure

Best practices

Define your intent before you build. Ask yourself: what is this dataset tracking, who uses it, and what decisions will it inform? The clearer your intent, the better your structure.

Write a specific description, not a vague one. For example, "Manage our support process, track issues by type and priority, assign to agents, measure resolution time" will get you further than "track customer issues."

One dataset, one thing. Each dataset should represent one type of entity — customers, tasks, invoices. If you find yourself adding columns that belong to a different subject, that's a signal you need a second dataset.

Name things clearly. "Customer Contacts" is easier to work with than "List 2." Consistent naming speeds up building automations and logic later.

Start simple. Build for your core use case first, then expand. Adding columns later is easy — untangling a structure that grew too fast is not.