Fig Helpers

What Fig Helpers Are

fig is a small, curated Python library built into every Go Fig Code node. It gives you five purpose-built functions for the data tasks that are painful to write from scratch in pandas: fuzzy matching, fiscal year conversion, statistical scoring, safe rate calculations, and near-duplicate detection. You don’t install or import fig. It’s always available inside the Code node’s Python mode, right alongside pd (pandas) and np (numpy).

When to Use Fig Helpers

Use fig.* when:

You need to match records across two sources where names or IDs aren’t exact (CRM ↔ accounting, messy leads, product SKUs)
Your organization uses a non-January fiscal year and you need to group dates correctly
You want a z-score, rolling z-score, or safe percentage change without writing the edge cases yourself
You’re cleaning a list with near-duplicate entries and DISTINCT won’t catch them

Don’t use them when:

A built-in node can do the same job. The XLOOKUP node handles exact joins without code
The task is simple enough for plain pandas (adding columns, basic filtering, .groupby)

Available Functions

The examples below reference B1 as a second parent table. In Go Fig, each parent connected to a code node is exposed as a DataFrame named after its step label (A1, B1, C1, …), the same label you see on the node in the canvas. The position-0 parent is also bound to data. See Joining Multiple Parents in the Code node docs.

`fig.fuzzy_join`

Match rows between two tables by approximate string similarity. For each row in the left table, finds the best match in the right table that scores above your similarity threshold. Signature:

fig.fuzzy_join(
    left,                # pandas DataFrame (your base table)
    right,               # pandas DataFrame (the lookup table)
    left_on='col',       # string column in left
    right_on='col',      # string column in right
    threshold=0.85,      # similarity cutoff, 0.0 to 1.0
    how='left',          # 'left' keeps all left rows; 'inner' drops unmatched
)

Returns: a DataFrame combining columns from both tables, plus a _match_score column (0.0 to 1.0). Unmatched rows get NaN on the right-side columns and on _match_score. Example: reconcile HubSpot deals with QuickBooks customers:

# A1 is HubSpot deals, B1 is QuickBooks customers
matched = fig.fuzzy_join(
    A1,
    B1,
    left_on='company_name',
    right_on='customer_name',
    threshold=0.85,
    how='left',
)
# Rows with _match_score == NaN are deals with no matching QBO customer
return matched

`fig.to_fiscal`

Convert a date column into a fiscal-year integer, named by the calendar year the fiscal year starts in. Supports any fiscal-year start month. Signature:

fig.to_fiscal(
    dates,                          # pandas Series or column of dates
    fiscal_year_start_month=1,      # 1 = calendar year, 7 = July-start, etc.
)

Returns: a NumPy array of fiscal year integers. Example: July fiscal year start:

# Your org's FY runs July through June
# Jul 2025 → FY 2025, Jun 2026 → FY 2025, Jul 2026 → FY 2026
data['fiscal_year'] = fig.to_fiscal(
    data['order_date'],
    fiscal_year_start_month=7,
)
return data

`fig.zscore`

Calculate z-scores, or how many standard deviations a value is from the mean. Useful for anomaly detection and identifying outliers. Signature:

fig.zscore(
    series,              # pandas Series
    window=None,         # None for global z-score, int for rolling window
)

Returns: a pandas Series of z-scores. Zero-variance inputs return all zeros (no NaN, no division error). Example: global z-score for outlier detection:

data['revenue_zscore'] = fig.zscore(data['revenue'])
# Filter to outliers (more than 2 standard deviations from mean)
outliers = data[data['revenue_zscore'].abs() > 2]
return outliers

Example: 30-day rolling z-score:

data['rolling_zscore'] = fig.zscore(data['daily_revenue'], window=30)
return data

`fig.pct_change_safe`

Percentage change between consecutive rows, with division-by-zero and missing-first-row handled safely. Signature:

fig.pct_change_safe(
    series,              # pandas Series
    fill=0.0,            # value to use where pct_change is inf or NaN
)

Returns: a pandas Series. First row and any divide-by-zero rows get the fill value instead of NaN or inf. Example: month-over-month revenue change:

data = data.sort_values('month')
data['mom_change'] = fig.pct_change_safe(data['revenue'], fill=0.0)
return data

`fig.dedupe`

Drop duplicates by exact match on a column subset, or cluster near-duplicates by fuzzy string similarity. Signature:

fig.dedupe(
    df,                  # pandas DataFrame
    subset=['col'],      # column(s) to check for duplicates
    keep='first',        # 'first' or 'last' row per group
    fuzzy=False,         # True clusters near-matches; False is exact
    threshold=0.9,       # similarity cutoff when fuzzy=True
)

Returns: a deduplicated DataFrame. Example: collapse messy company name variations:

# "IBM", "I.B.M.", "International Business Machines" all become one row
data = fig.dedupe(
    data,
    subset=['company'],
    fuzzy=True,
    threshold=0.85,
)
return data

Recipe Picker

Every fig function has a matching snippet in the Insert recipe dropdown at the top of the Python Code editor. Pick a recipe, edit the column names to match your data, and you’re done. No typing required.

Start with a recipe when you’re new to fig. The dropdown inserts a working example at the cursor so you just edit the column names and threshold values to match your data.

Usage Notes

fig is only available in Python mode of the Code node. SQL mode has no equivalent.

Fuzzy matching is computationally expensive. fig.fuzzy_join and fig.dedupe with fuzzy=True scale with the product of row counts so a 10,000 × 10,000 fuzzy join will be slow. For large tables, filter or sample down first.

Thresholds are 0.0 to 1.0, not 0 to 100. If you’re coming from fuzzywuzzy or rapidfuzz, remember that Fig helpers normalize scores to a 0-to-1 scale. 0.85 means 85% similarity.

Code Node

Write custom SQL or Python directly in your flow

XLOOKUP Node

Exact-match joins without code

Getting started

Flows

Dashboards

Collaboration

Embedded analytics

What Fig Helpers Are

When to Use Fig Helpers

Available Functions

`fig.fuzzy_join`

`fig.to_fiscal`

`fig.zscore`

`fig.pct_change_safe`

`fig.dedupe`

Recipe Picker

Usage Notes

Code Node

XLOOKUP Node

​What Fig Helpers Are

​When to Use Fig Helpers

​Available Functions

​fig.fuzzy_join

​fig.to_fiscal

​fig.zscore

​fig.pct_change_safe

​fig.dedupe

​Recipe Picker

​Usage Notes

​Related Topics

Code Node

XLOOKUP Node

What Fig Helpers Are

When to Use Fig Helpers

Available Functions

`fig.fuzzy_join`

`fig.to_fiscal`

`fig.zscore`

`fig.pct_change_safe`

`fig.dedupe`

Recipe Picker

Usage Notes

Related Topics