Data Profiling

A collection of nine Tabular Editor C# macros for exploring and profiling data in Power BI semantic models. Select columns or tables, run a macro, and get a structured data grid in TE3's output pane — one row per column with stats at a glance.

Prerequisites

• Tabular Editor 3 with a live connection to a Power BI or Analysis Services model
• The macros use C# features that require the Roslyn compiler — standard in TE3, but TE2 users need to enable it first

Macros

Macro	Context	Purpose
VizColumnProfiles	Column	Quick scan: cardinality and blank stats per column
VizColumnProfilesTopN	Column	Same, limited to first N rows (configurable)
VizColumnDistributions	Column	Deep scan: distinct/blank stats, min/max, Unicode histogram, mean/median/stdev
VizColumnDistributionsTopN	Column	Same, limited to first N rows (configurable)
VizSummaryData	Table / Column / Measure	High-level summary across selected tables, columns, and measures
VizRelationshipCoverage	Model	FK and PK coverage across all relationships, sorted by worst coverage first
VizStringQuality	Column	Whitespace issues (leading spaces, double spaces) and mixed types in text columns
VizOutliers	Column	IQR-based outlier detection for numeric columns; string length outliers for text columns
VizDateGaps	Column	Consecutive date spacing analysis; flags unusually large gaps

The first six macros handle broad profiling; the last three target specific data quality dimensions.

Installation

1. Copy the .csx files to your Tabular Editor 3 macros folder (%AppData%\Local\TabularEditor3\Macros), or use File → New C# Script → paste contents → Save as Macro
2. Set the macro context (Column, Table, or none) as noted in the table above and each script's header
3. Macros with Column context appear in the right-click menu when you select columns in TOM Explorer; optionally assign keyboard shortcuts in Preferences → Keyboard Bindings

Output reference

Each macro's header comment contains the full column reference. Below is a summary for the macros with non-obvious output.

VizColumnDistributions (and TopN variant)

Column	What it shows	How to read it
# Distinct	Distinct non-blank values	High relative to row count means many unique values; low on a key column may indicate data quality issues
% Distinct	Distinct as % of total rows	100% = every row is unique; 1% = highly repeated values
# Blank	Rows with blank/null values	Any blanks on a foreign key column mean fact rows that won't join to the dimension
Min / Max	Lowest and highest value	Alphabetical for text; True/False for boolean; useful for spotting impossible values
Distribution	12-bin Unicode histogram (▁▂▃▄▅▆▇█)	Shape of the value distribution; a single tall bar means most values cluster in one range
Mean / Median / StdDev	Numeric columns only	Large gap between mean and median suggests skew or outliers

VizRelationshipCoverage

One row per relationship in the model, sorted by worst coverage first. Covers both directions of each relationship.

Column	What it shows	How to read it
Relationship	FactTable[FK] → DimTable[PK]	Identifies which relationship the row describes
Active	Whether the relationship is active	Inactive relationships require USERELATIONSHIP() in DAX to filter
# FK Values	Distinct non-blank values in the FK column	—
# FK Unmatched	FK values with no match in the PK	These rows resolve to the blank member in the dimension; they show as BLANK in report groupings
% FK Coverage	(FK Values − FK Unmatched) / FK Values	100% = every fact row links to a dimension row; anything below deserves investigation
# PK Values	Distinct non-blank values in the PK column	—
# PK Unused	PK values not referenced by any FK row	Unused dimension members inflate slicer lists; may be legitimate (inactive customers) or a data gap
% PK Coverage	(PK Values − PK Unused) / PK Values	Low coverage is not necessarily a problem, but worth understanding

Note: blank FK values are not counted as unmatched — they map to the blank dimension row, which is expected behaviour.

VizStringQuality

Column	What it shows	How to read it
# Whitespace	Rows where the value differs from TRIM(value)	Catches leading spaces and multiple consecutive internal spaces; trailing spaces are stripped by Power BI at load time and cannot appear in the model
# Numeric Values	Non-blank values that parse as a number	Text columns storing numbers as strings; may cause implicit conversion issues in DAX
# Date Values	Non-blank values that parse as a date	Text columns storing dates as strings; relationships and time intelligence won't work on these

Note: casing inconsistencies (e.g. "Shipped" vs "SHIPPED") are not detected — VertiPaq's default case-insensitive collation collapses them into a single stored value.

VizOutliers

Uses the interquartile range (IQR) method: values outside Q1 − 1.5×IQR or Q3 + 1.5×IQR are flagged as outliers.

Column	What it shows	How to read it
# Low	Values below the lower IQR bound	Numeric columns only; negative values in a quantity column often appear here
# High	Values above the upper IQR bound	Numeric columns only; unusually large amounts or placeholder values
Normal Range	Lower bound – upper bound	The IQR-defined "expected" range for this column
# Outliers	Total outlier count (low + high for numeric; length outliers for text)	For text columns, flags rows whose string length is outside the IQR of string lengths
% Outliers	Outliers as % of non-blank rows	Helps distinguish a handful of edge cases from a systematic issue

Further reading

For a walkthrough showing these macros in action, see the Tabular Editor blog.

For sample AI Assistant conversations that build on macro output — drilling into whitespace issues, unused dimension members, and negative quantities — see EXAMPLES.md.