Create Conditional Columns in Review Tables
Reference responses in existing columns as context for responses in new columns to power workflows like conditional follow-up logic, targeted summary of entities, and transactional extraction.
Last updated: Sep 23, 2026
Overview
A conditional column is a review table column that uses the answers in an existing column as context for its own answers.
For example, one column named "Non-Compete Inclusion" might prompt, "Does this contract include a non-compete clause?" and a second column might prompt, "If @Non-Compete Inclusion says yes, summarize the non-compete terms. Otherwise, leave blank."
Conditional columns are useful for:
- Conditional Follow-Ups: Check whether a clause, party, or term exists, then describe it in the next column if it exist.
- Targeted Entity Summaries: Identify a specific person or entity in one column, then summarize information about them in the next.
- Classifications and Extractions: Sort each document by type(s) first, then pull out the details relevant to that type(s).
Example Use Cases
Conditional Follow-Up
Use one column to detect a condition, and another to act on it only when the condition is met.
Column | Prompt |
|---|---|
Column 1 | Does an X clause exist? |
Column 2 | If an X clause exists per @Column 1, describe it. Otherwise, leave blank. |
Targeted Entity Summary
Focus a summary on a specific entity identified in a prior column.
Column | Prompt |
|---|---|
Column 1 | What is the name of the defendant? |
Column 2 | Summarize statements made by @Column 1. |
Classifications and Extractions
Classify the document type first, then extract details relevant to that type in a new column.
Column | Prompt |
|---|---|
Column 1 | Classify the type of agreement. |
Column 2 | Based on the agreement type in @Column 1 if this is a power agreement, extract energy pricing structure, as described as fixed, indexed, or market-based. If this is a sub-engineering agreement, extract fee structure, as described as lump sum, time & materials, or cost-plus. |
How to Create a Conditional Column
- Open a review table with at least one existing column.
- Select + Add column to add a new column.
- Write a question with conditional logic. Use the
@symbol to reference a prior column by name. - Select Continue.
The new conditional column is now dependent on the prior column.
Refreshing Conditional Cells
When you’re refreshing a cell, Harvey lets you choose whether the refresh applies to just that cell or also to the cells that depend on it.
- Select the cell you want to refresh and choose
Re-run. - When prompted, select one of two options:
- Refresh cell: Re-run only the selected cell.
- Refresh cell and dependent cells: Re-run the selected cell and every conditional cell that references it.
Note: If the cell you refresh is locked, it won't re-run, and if a cell that depends on it is locked, it won't update when you choose Refresh cell and dependent cells. Unlock a cell first if you need it to reflect a new answer.
Tips for Success
- Be explicit about when to leave a cell blank. If a condition is not met, say so clearly (e.g., "Otherwise, leave blank"). Without this, the column may generate a response even when there is nothing relevant to report.
- Break complex logic into multiple columns. If you need to classify, then filter, then extract — use three columns rather than one. Simpler prompts tend to produce more accurate and predictable results.
- Refresh dependent cells to keep conditional logic in sync. When you re-run a cell that other columns depend on, choose Refresh dependent cells so that downstream conditional cells update automatically. If you refresh only the single cell, make sure to review any conditional columns that reference it for accuracy.
Was this article helpful?