Formula

A formula column computes its value from other columns in the same row. It calls no provider, so it costs no credits, and it recomputes when its inputs change.

This is where scoring, flagging, cleaning and combining happen.

What it is for

Add a formula column

1. Click Add Action and choose the formula column
2. Reference the columns you need as dynamic variables — type / and pick the column
3. Run one row and read the output before running the column
4. Run the column with ▶️ in the header

Because it is free, run it on the whole table — there is no reason to sample.

Formula Generator

You do not have to write the logic yourself. Formula Generator takes a plain-English description of the transformation and builds the formula column for you.

Describe the output, reference the source column, review what it generated, then run it.

Nulls are the thing that breaks formulas

A formula reading a column that has not populated yet does not error — it produces a value from an empty input, and that value looks like a real answer.

• Test every formula on a row where the input is empty, not just on rows where it worked
• A score built on missing inputs will score everything the same and nobody will notice
• If a flag has never come out false, it has not been tested

This matters most when the formula feeds a conditional formula — a gate reading a broken score passes or fails everything.

Formulas are free, so put them everywhere

The pattern that keeps a table cheap:

cheap columns  →  formula score  →  conditional gate  →  expensive columns

Everything left of the gate costs 0 or near 0. The formula is what makes the gate possible. See Credits for the arithmetic.

Common use cases

• Compute a fit score from columns you already paid for, then gate the expensive columns on it
• Flag rows that are missing an input so a provider never runs on them
• Normalize a messy input key — strip www., paths and parameters down to a clean domain
• Build a personalization string from a name, a company and a signal
• Bucket a raw number into a band — headcount into SMB / Mid-market / Enterprise
• Pull a piece out of an existing value — the domain from an email, the year from a date

Best practices

• Test every formula on a row where the input is empty, not only on rows where it worked
• One formula, one job — chain simple columns instead of writing one long expression
• Name the column after what it outputs (fit_score, qualified), not after the formula
• Run it on the whole table — it is free, so there is no reason to sample
• Read what Formula Generator produced before you trust it on a full column
• If a flag has never come out false, it has not been tested

Where to next

• Conditional Formula — use a formula’s output to decide which rows an action runs on
• Formula Generator — write formulas in plain English
• Built-in Enrichments — free actions that split names, merge columns and parse JSON
• Score and gate the spend — a full scoring model built from free columns