Svg Visuals

SVG Visuals via DAX Measures (PBIR)

Report modification requires tooling. Two paths exist:

1. `pbir` CLI (preferred) -- use the pbir command and the pbir-cli skill. Install with uv tool install pbir-cli or pip install pbir-cli. Check availability with pbir --version.

2. Direct JSON modification -- if pbir is not available, use the pbir-format skill (pbip plugin) for PBIR JSON structure and patterns. Validate every change with jq empty <file.json>.

>

If neither the pbir-cli skill nor the pbir-format skill is loaded, ask the user to install the appropriate plugin before proceeding with report modifications.

Generate inline SVG graphics using DAX measures that return SVG markup strings. These render as images in table, matrix, card, image, and slicer visuals. Store as extension measures in reportExtensions.json.

How It Works

1. A DAX measure returns an SVG string prefixed with data:image/svg+xml;utf8, 2. The measure's dataCategory is set to ImageUrl 3. Power BI renders the SVG as an image in supported visuals

Supported Visuals

• Table (tableEx): grid.imageHeight / grid.imageWidth -- references/svg-table-matrix.md
• Matrix (pivotTable): same as table -- references/svg-table-matrix.md
• Image (image): sourceType='imageData' + sourceField -- references/svg-image-visual.md
• Card/New (cardVisual): callout.imageFX -- references/svg-card-slicer.md
• Slicer/New (advancedSlicerVisual): header images -- references/svg-card-slicer.md

Workflow: Creating an SVG Measure

Step 0: Design and Preview

Before writing DAX, design the SVG visually:

1. Query the model first -- use DAX Studio or Tabular Editor CLI to get actual values with the intended filter context. Use real numbers, not placeholders. 2. Write static SVG to a temp file -- save to /tmp/mockup.svg and open it in a browser to preview layout, colors, and proportions. 3. Ask for feedback before converting to DAX -- iterating on static SVG is far easier than on DAX string concatenation. 4. Colors must be hex codes with `#` -- e.g., fill='#2B7A78'. Never use %23 URL encoding or named colors. Always hex.

Step 1: Create the Extension Measure

Create the extension measure in reportExtensions.json manually (see the pbir-format skill in the pbip plugin for JSON structure).

# Example using pbir_object_model (if available):
report.add_extension_measure(
    table="Orders",
    name="Sparkline SVG",
    expression='''
        VAR _Prefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 30'>"
        VAR _Bar = "<rect x='0' y='0' width='50' height='30' fill='#2196F3'/>"
        VAR _Suffix = "</svg>"
        RETURN _Prefix & _Bar & _Suffix
    ''',
    data_type="Text",
    data_category="ImageUrl",
    display_folder="SVG Charts",
)
report.save()

Step 1b: Review

Before presenting the measure to the user, dispatch the svg-reviewer agent to validate syntax and provide design feedback.

Step 2: Bind to a Visual

Extension measures use "Schema": "extension" in the SourceRef:

{
  "field": {
    "Measure": {
      "Expression": {
        "SourceRef": {"Schema": "extension", "Entity": "Orders"}
      },
      "Property": "Sparkline SVG"
    }
  }
}

For image visuals, set sourceType='imageData' with sourceField in the visual.json (see references/svg-image-visual.md).

Step 3: Validate

Validate JSON syntax with jq empty <reportExtensions.json> and inspect the file to confirm measure definitions and data categories.

Prefer UDF Libraries Over Custom DAX

Before writing a custom SVG measure from scratch, check whether an existing UDF library already provides the chart type:

• PowerofBI.IBCS (Andrzej Leszkiewicz) -- IBCS-compliant bar, column, waterfall, pin, small multiples, and P&L charts. Preferred for business reporting with AC/PY/BU/FC comparisons. Install from https://daxlib.org/package/PowerofBI.IBCS/
• DaxLib.SVG (Jake Duddy) -- general-purpose sparklines, bars, boxplots, heatmaps, jitter, violin, progress bars, pills. Install from https://daxlib.org/package/DaxLib.SVG/ -- source at https://github.com/daxlib/dev-daxlib-svg
• PowerBI MacGuyver Toolbox (Stepan Resl / Data Goblins) -- C# scripts that generate SVG measures via Tabular Editor

To check if a library is installed, look for functions/measures starting with PowerofBI.IBCS., Viz., Compound., or Element.. Only write custom SVG DAX when no library function covers the required visualization. See references/community-examples.md for full function listings and additional libraries.

Installing UDF libraries

UDF libraries are installed into the semantic model, not the report. Use one of these tools:

• Tabular Editor CLI (te command) -- use the te-docs skill for guidance
• Power BI MCP server -- if available, use it to modify the model directly
• `connect-pbid` skill -- connect to Power BI Desktop's local Analysis Services instance via TOM/PowerShell
• `tmdl` skill -- edit TMDL files directly in a PBIP project (last resort)

DAX SVG Conventions

Measure Structure (VAR Pattern)

Every SVG measure must follow a strict VAR-based structure. Organize code into clearly separated regions:

SVG Measure =
-- CONFIG: Input fields and visual parameters
VAR _Actual = [Sales Amount]
VAR _Target = [Sales Target]
VAR _Scope = ALLSELECTED ( 'Product'[Category] )

-- CONFIG: Colors
VAR _BarColor = "#5B8DBE"
VAR _TargetColor = "#333333"

-- NORMALIZATION: Scale values to SVG coordinate space
VAR _AxisMax = CALCULATE( MAXX( _Scope, [Sales Amount] ), REMOVEFILTERS( 'Product'[Category] ) ) * 1.1
VAR _AxisRange = 100
VAR _ActualNormalized = DIVIDE( _Actual, _AxisMax ) * _AxisRange

-- SVG ELEMENTS: One VAR per visual element
VAR _SvgPrefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 25'>"
VAR _Sort = "<desc>" & FORMAT( _Actual, "000000000000" ) & "</desc>"
VAR _Bar = "<rect x='0' y='5' width='" & _ActualNormalized & "' height='15' fill='" & _BarColor & "'/>"
VAR _TargetLine = "<rect x='" & DIVIDE( _Target, _AxisMax ) * _AxisRange & "' y='2' width='2' height='21' fill='" & _TargetColor & "'/>"
VAR _SvgSuffix = "</svg>"

-- ASSEMBLY: Combine in rendering order (back to front)
VAR _SVG = _SvgPrefix & _Sort & _Bar & _TargetLine & _SvgSuffix

RETURN _SVG

Key conventions:

• CONFIG section first -- input measures, scope column, colors, font settings. Users change only this section.
• NORMALIZATION section -- scale raw values to SVG coordinate space (see below)
• SVG ELEMENTS -- one VAR per <rect>, <circle>, <text>, <line>, etc.
• ASSEMBLY -- concatenate elements in document order (first = back layer, last = front)
• `<desc>` sort trick -- embed FORMAT(_Actual, "000000000000") in a <desc> tag so the table/matrix can sort by the SVG column

Axis Normalization (Critical)

SVG coordinates must be normalized to a fixed range. Raw measure values (e.g., 1,234,567) cannot be used directly as pixel coordinates. The standard pattern:

-- 1. Define the SVG coordinate range
VAR _BarMin = 0       -- leftmost position (or offset for labels)
VAR _BarMax = 100     -- rightmost position

-- 2. Find the maximum value across all rows in the visual's filter context
VAR _Scope = ALLSELECTED( 'Table'[GroupColumn] )
VAR _MaxInScope = CALCULATE( MAXX( _Scope, [Measure] ), REMOVEFILTERS( 'Table'[GroupColumn] ) )
VAR _AxisMax = _MaxInScope * 1.1   -- 10% padding

-- 3. Normalize each value to the SVG range
VAR _AxisRange = _BarMax - _BarMin
VAR _Normalized = DIVIDE( _Actual, _AxisMax ) * _AxisRange

Use ALLSELECTED for the scope when the chart should respond to slicer context. Use ALL for a fixed axis across all filter contexts. The * 1.1 padding prevents bars from touching the edge.

HASONEVALUE Guard

Table/matrix SVG measures must guard against subtotal/total rows where multiple categories are in scope:

IF( HASONEVALUE( 'Table'[GroupColumn] ),
    -- SVG code here
)

Without this guard, the measure evaluates on grand total rows with meaningless aggregated values.

Escaping and Color Rules

• Single quotes for SVG attributes -- avoids DAX double-quote escaping: fill='#2196F3'
• Double quotes in DAX: escape as "" (DAX convention)
• `viewBox` for responsive scaling: viewBox='0 0 100 25'
• `xmlns` required on <svg> element
• Hex colors with `#` only -- e.g., fill='#2196F3'. %23 URL encoding causes errors in image visuals. Never use named colors.
• No JavaScript -- SVG must be purely declarative

SVG Coordinate System

• Y=0 is at the top -- invert values for charts: _Height - _Value
• Use viewBox with a 0-100 range for normalized coordinates
• Elements render in document order (first = back, last = front)

CONCATENATEX for Series Data

For sparklines and multi-point charts, build coordinate strings with CONCATENATEX:

VAR _Points = CONCATENATEX(
    _SparklineTable,
    [X] & "," & (100 - [Y]),
    " ",
    [Date], ASC
)
-- Produces: "0,80 10,60 20,40 30,20"
-- Use in: <polyline points='...'/>

Best Practices

• Check UDF libraries first -- use DaxLib.SVG or MacGuyver Toolbox functions before writing custom DAX
• VAR pattern mandatory -- one VAR per config value, one VAR per SVG element, assembly at the end
• Normalize all values -- raw measure values must be scaled to SVG coordinate range
• HASONEVALUE guard -- always guard against total/subtotal rows in table/matrix context; use ISINSCOPE for nested hierarchy levels
• <desc> sort trick -- embed formatted value in <desc> for sortable SVG columns
• Use viewBox for responsive scaling instead of fixed width/height
• Round coordinates to integers for performance (shorter strings, cheaper FORMAT calls)
• Store as extension measures -- SVG measures don't belong in the semantic model
• Use display_folder to organize SVG measures (e.g., "SVG Charts")
• Preview first -- save static SVG to /tmp/, open in browser, iterate before writing DAX
• 32K character limit on the rendered SVG string per cell (not the DAX expression); see references/svg-table-matrix.md for diagnosis and mitigation
• Pre-aggregate in model measures -- let the storage engine cache aggregations; the SVG measure maps numbers to coordinates only
• Hex colors only -- # directly, never %23 URL encoding
• Image visuals need no query block -- only objects.image with sourceType='imageData' and sourceField
• Accessibility: every SVG encoding primary data needs adjacent readable columns and a dynamic alt-text measure; see references/svg-accessibility.md

reportExtensions.json Format

{
  "$schema": "https://developer.microsoft.com/json-schemas/fabric/item/report/definition/reportExtension/1.0.0/schema.json",
  "name": "extension",
  "entities": [{
    "name": "ExistingTable",
    "measures": [{
      "name": "Sparkline SVG",
      "dataType": "Text",
      "dataCategory": "ImageUrl",
      "expression": "...",
      "displayFolder": "SVG Charts"
    }]
  }]
}

Limitations

• No interactivity -- SVG images are static (no hover, click, tooltip)
• No JavaScript -- inline scripts are stripped
• 32K character limit per rendered cell string (not the DAX expression); CONCATENATEX over 30+ series points easily approaches this; prefer <polyline> over individual shapes, integer coordinates, and pre-aggregated series; see references/svg-table-matrix.md for full diagnosis
• Per-cell formula-engine cost -- each visible cell evaluates the string-building expression; push aggregations into model measures so SVG assembly is coordinate-mapping only
• Accessibility gap -- screen readers receive no per-cell data from an SVG URI; mitigate with adjacent readable columns and dynamic alt text; see references/svg-accessibility.md
• Classic card (card) does NOT support SVG -- use cardVisual instead

When to Use SVG Measures

SVG measures are the preferred choice for simple inline graphics embedded in tables, matrices, cards, and image visuals. Use SVG when you need:

• Sparklines, data bars, progress bars, or status indicators inside table/matrix cells
• KPI micro-charts in card visuals
• Lightweight visuals that don't require interactivity or complex data transforms
• No additional custom visual registration (works with native visuals)

Use Deneb instead for complex, interactive visualizations (cross-filtering, tooltips, hover states) or chart types that require extensive data transforms. Use Python/R instead for statistical analysis charts (distributions, regressions, correlations).

References

Community Examples and Libraries

• `references/community-examples.md` -- Community SVG templates organized by target visual type (Table/Matrix, Image, Card), including DaxLib.SVG functions, Kerry Kolosko templates, and PowerBI MacGuyver Toolbox patterns

By Visual Type

• `references/svg-table-matrix.md` -- Patterns for Table/Matrix: data bar, bullet chart, dumbbell, overlapping bars, lollipop, status pill, sparkline, bar sparkline, area sparkline, UDF patterns; axis normalization, sort trick, image size configuration, and per-cell performance guidance
• `references/svg-image-visual.md` -- Patterns for Image visuals: KPI header, sparkline with endpoint, dashboard tile; sourceType binding, dynamic/conditional layout, responsive width guidance
• `references/svg-card-slicer.md` -- Patterns for Card/Slicer: arrow indicator, mini gauge, mini donut, progress bar, narrative sentence; card binding via callout.imageFX

Accessibility

• `references/svg-accessibility.md` -- Accessibility for SVG measures: adjacent readable columns, dynamic alt text, color-only encoding, contrast requirements, and severity guidance for audit findings

General

• `references/svg-elements.md` -- SVG element reference (rect, circle, line, polyline, text, path, gradient, group)

Examples

Ready-to-use DAX measure expressions in examples/:

• `sparkline-measure.dax` -- Line sparkline (polyline + CONCATENATEX)
• `progress-bar-measure.dax` -- Conditional progress bar
• `dumbbell-chart-measure.dax` -- Actual vs target dumbbell
• `bullet-chart-measure.dax` -- Bullet chart with sentiment action dots
• `overlapping-bars-measure.dax` -- Overlapping bars with variance label
• `boxplot-measure.dax` -- Box-and-whisker plot (inspired by DaxLib.SVG)
• `ibcs-bar-measure.dax` -- IBCS-compliant horizontal bar (inspired by avatorl)
• `jitter-plot-measure.dax` -- Dot strip chart with jitter (inspired by DaxLib.SVG)
• `overlapping-bars-with-variance-measure.dax` -- Overlapping bars with variance bar + arrow icon + % label (Kurt Buhler / Data Goblins)
• `lollipop-conditional-measure.dax` -- Lollipop with scaled dot + auto-formatted label (Kurt Buhler / Data Goblins)
• `waterfall-measure.dax` -- Waterfall with cumulative OFFSET positioning + connector lines (Kurt Buhler / Data Goblins)
• `status-pill-measure.dax` -- Rounded pill badge with category color + text label (Kurt Buhler / Data Goblins)

Helper Libraries

Related Skills

• `pbi-report-design` -- Layout and design best practices
• `deneb-visuals` -- Vega/Vega-Lite for complex interactive visualizations
• `python-visuals` -- matplotlib/seaborn for statistical charts
• `r-visuals` -- ggplot2 for statistical charts
• `pbir-format` (pbip plugin) -- PBIR JSON format reference (extension measures, ImageUrl binding)

SVG Community Examples and Libraries

Organized by target visual type. Use these as reference when building SVG DAX measures.

Libraries (UDF-based)

See also references/svg-table-matrix.md for the UDF pattern and calling convention.

PowerofBI.IBCS Functions (Andrzej Leszkiewicz)

IBCS-compliant SVG chart functions for business reporting. Install via DaxLib: https://daxlib.org/package/PowerofBI.IBCS/ -- source: https://github.com/avatorl/dax-udf-svg-ibcs

All functions target Table/Matrix visuals. Uses DAX UDF syntax with typed parameters.

Bar Charts (horizontal)

Column Charts (vertical)

Waterfall / P&L

Extras

DaxLib.SVG Functions (Jake Duddy)

High-level Viz.* functions -- each produces a complete SVG data URI for Table/Matrix Image URL columns:

Mid-level Compound.* functions create positionable chart components you combine with SVG() wrapper. Low-level Element.* functions create raw SVG primitives (rect, circle, line, polyline, text, path, group).

Docs: https://evaluationcontext.github.io/daxlib.svg/

Examples by Target Visual: Table / Matrix

SVG measures rendered in table/matrix cells via Image URL column binding. DaxLib.SVG and PowerofBI.IBCS functions (listed above) all target Table/Matrix.

Kerry Kolosko (kerrykolosko.com/portfolio/)

PowerBI MacGuyver Toolbox (Kurt Buhler / Data Goblins, SVG measures by Stepan Resl)

MacGuyver Toolbox provides both native Power BI visual templates (bar/line/KPI patterns) and SVG DAX measures. The SVG measures are added via C# scripts in Tabular Editor. See examples/ for extracted DAX measures from this toolbox.

Examples by Target Visual: Image

SVG measures rendered in standalone Image visuals via sourceType='imageData'.

Kerry Kolosko (kerrykolosko.com/portfolio/)

Examples by Target Visual: Card (New)

SVG measures rendered in the new Card visual via callout.imageFX.

PowerBI MacGuyver Toolbox (Stepan Resl / Data Goblins)

KPI card templates using SVG measures for inline micro-charts in card visuals:

Repo: https://github.com/data-goblin/powerbi-macguyver-toolbox

SVG Measures: Accessibility

A DAX SVG measure renders as an image keyed off ImageUrl; the screen-reader path reads the visual's title, type, author alt text, and the "Show data" table. A cell whose value is an svg+xml URI contributes nothing meaningful to either path (Show-data lists the literal markup string), and there is no per-cell alt-text slot. Every inline SVG chart is an accessibility dead zone by default.

Mitigation at the host-visual level

Apply both of the following for any SVG measure encoding primary KPI data; the second alone is sufficient for decorative micro-charts.

1. Keep the numbers in adjacent readable columns

For an SVG sparkline column, also bind the numeric measures the SVG encodes (min, max, last value; actual, target, variance) as plain value columns in the same table or matrix. Show-data then carries real numbers rather than markup strings, satisfying WCAG 1.1.1 for the data-table fallback.

Decide per project which columns to expose vs hide at the visual level (they can be hidden from the visual but still present in the query so Show-data picks them up).

2. Set dynamic alt text on the host container

Author a sibling _Alt measure returning a spoken sentence (under ~250 characters) that narrates the SVG's content for the current filter context:

KPI Sparkline Alt =
VAR _Last = [Sales Amount]
VAR _Trend = IF([Sales Amount] > [Sales Amount PY], "up", "down")
RETURN
    "Sales: " & FORMAT(_Last, "$#,0,, M") & ", trending " & _Trend & " vs prior year"

Bind via pbir set:

pbir set "MyPage/MyTable.Visual" "general.altText" --measure "_Report.KPI Sparkline Alt"

Or set directly in visual.json under visualContainerObjects.general[0].properties.altText:

"general": [{
  "properties": {
    "altText": {
      "expr": {
        "Measure": {
          "Expression": { "SourceRef": { "Schema": "extension", "Entity": "Sales" } },
          "Property": "KPI Sparkline Alt"
        }
      }
    }
  }
}]

The measure must guard BLANK() (a blank alt text is no better than none):

IF(COUNTROWS(VALUES('Date'[Month])) > 0,
    "Sales: " & FORMAT(_Last, "$#,0,, M") & "...",
    "No data for the current selection."
)

For an image visual rendering a single SVG, the container has a general.altText slot; always set it. The static literal form is acceptable when the SVG is not filter-sensitive.

Decorative SVGs

For purely decorative SVGs (dividers, brand backgrounds), mark the host hidden in tab order so screen readers skip it entirely:

pbir set "MyPage/MyDecorativeSVG.Visual" "position.tabOrder" -1

Color-only encoding

An SVG that conveys status purely via fill/stroke color fails WCAG 1.4.1. Pair every semantic color with at least one of:

• A shape or glyph change (circle vs diamond, filled vs hollow)
• A text label or abbreviation inside the SVG
• A <title> element inside the SVG markup itself

The status pill pattern already does this (text label alongside fill color). The dumbbell and bullet patterns do not; add a symbol or label if they are primary status indicators.

Contrast

SVG elements are subject to the same WCAG contrast requirements as rasterized content:

• Text inside the SVG: >= 4.5:1 against the cell background (3:1 for large text >= 18pt)
• Graphical elements (bars, lines, dots used for data): >= 3:1 against adjacent colors

Theme tokens from the host report's theme do not automatically propagate into an SVG string; verify contrast manually, especially on dark themes or high-contrast accessibility themes.

What the Desktop Bridge screenshot does not confirm

The Bridge screenshot is useful for layout verification; it does not confirm:

• Alt text is populated and readable
• Adjacent readable columns are present in Show-data
• Tab order positions the visual correctly for keyboard users

Confirm accessibility by reading back the bound measure expression and auditing visual.json for general.altText and the presence of non-SVG bound columns.

Severity guidance for audit findings

When surfaced by the review-report skill:

• Primary KPI SVGs with no alt text and no adjacent numeric column: high severity
• SVG measures with color-only encoding and no paired shape or label: high severity
• Decorative micro-charts with no alt text: medium severity (low if clearly decorative and tab-order is -1)

SVG Patterns for Card and Slicer Visuals

Card (cardVisual) and Slicer (advancedSlicerVisual) visuals support SVG measures through specific binding patterns. The classic card does NOT support SVG -- only the new card visual works.

Card Visual (cardVisual)

Binding

Card visuals render SVG via callout.imageFX. Bind the SVG measure to the card's calloutValue field, then configure imageFX:

{
  "objects": {
    "callout": [{
      "properties": {
        "imageFX": {"expr": {"Literal": {"Value": "true"}}},
        "imageHeight": {"expr": {"Literal": {"Value": "40D"}}},
        "imageWidth": {"expr": {"Literal": {"Value": "100D"}}}
      }
    }]
  }
}

Pattern: Arrow Indicator

Compact directional indicator for KPI cards.

Arrow Indicator =
VAR _Growth = [Growth %]
VAR _Up = _Growth >= 0
VAR _Path = IF(_Up, "M 10,15 L 5,10 L 15,10 Z", "M 10,5 L 5,10 L 15,10 Z")
VAR _Color = IF(_Up, "#4CAF50", "#F44336")

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 20 20'>" &
"<path d='" & _Path & "' fill='" & _Color & "'/></svg>"

Pattern: Mini Gauge

Semi-circular gauge for progress or performance.

Mini Gauge =
VAR _Pct = DIVIDE([Value], [Target], 0)
VAR _Angle = (_Pct * 180) - 90
VAR _R = 40
VAR _CX = 50
VAR _CY = 50
VAR _Rad = _Angle * PI() / 180
VAR _NX = _CX + (_R * 0.8 * COS(_Rad))
VAR _NY = _CY + (_R * 0.8 * SIN(_Rad))

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 60'>" &
"<path d='M 10 50 A 40 40 0 0 1 90 50' fill='none' stroke='#E0E0E0' stroke-width='8'/>" &
"<line x1='" & _CX & "' y1='" & _CY & "' x2='" & _NX & "' y2='" & _NY & "' stroke='#333' stroke-width='2'/>" &
"<circle cx='" & _CX & "' cy='" & _CY & "' r='3' fill='#333'/></svg>"

Pattern: Mini Donut

Percentage completion as a donut ring.

Mini Donut =
VAR _Pct = [Percentage]
VAR _Angle = _Pct * 360
VAR _LargeArc = IF(_Angle > 180, 1, 0)
VAR _R = 40
VAR _CX = 50
VAR _CY = 50
VAR _Rad = (_Angle - 90) * PI() / 180
VAR _EndX = _CX + (_R * COS(_Rad))
VAR _EndY = _CY + (_R * SIN(_Rad))

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'>" &
"<circle cx='" & _CX & "' cy='" & _CY & "' r='" & _R & "' fill='none' stroke='#E0E0E0' stroke-width='8'/>" &
"<path d='M " & _CX & " " & (_CY - _R) & " A " & _R & " " & _R & " 0 " & _LargeArc & " 1 " & _EndX & " " & _EndY & "' fill='none' stroke='#2196F3' stroke-width='8'/></svg>"

Pattern: Progress Bar

Horizontal bar with label, sized for card visuals.

Progress Bar =
VAR _Pct = [Completion %]
VAR _W = 100
VAR _H = 20
VAR _FillW = _Pct * _W
VAR _Label = FORMAT(_Pct, "0%")
VAR _Color = SWITCH(TRUE(), _Pct < 0.5, "#F44336", _Pct < 0.8, "#FF9800", "#4CAF50")

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 " & _W & " " & _H & "'>" &
"<rect width='" & _W & "' height='" & _H & "' fill='#E0E0E0' rx='" & (_H / 2) & "'/>" &
"<rect width='" & _FillW & "' height='" & _H & "' fill='" & _Color & "' rx='" & (_H / 2) & "'/>" &
"<text x='" & (_W / 2) & "' y='" & (_H / 2 + 5) & "' font-size='11' text-anchor='middle' fill='white' font-weight='bold'>" & _Label & "</text></svg>"

Pattern: Narrative Sentence

When the story needs inline formatting that reacts to data (a clause that changes color on a miss, a bold figure, a verdict word keyed off performance), compose the whole sentence as a DAX measure returning an SVG with <text> and <tspan> runs. This is the only path to "X out of Y targets hit (~Z%)" where the number, color, and verdict word all key off data in a single declarative, version-controlled measure.

Decision guide for narrative elements:

• Plain "label: value" -> dynamic-value textbox run (simpler, inherits measure format string)
• Whole-string title with no per-clause styling -> expression-based DAX visual title
• Conditionally-styled clauses or sentence with inline micro-chart -> SVG narrative measure (below)
• Multi-paragraph AI summary -> Narrative visual (non-deterministic, license-gated, cannot be diffed; avoid for deterministic reporting)

Performance Narrative =
VAR _Actual   = [Sales Amount]
VAR _Target   = [Sales Target]
VAR _Variance = DIVIDE(_Actual - _Target, _Target)
VAR _Hit      = _Actual >= _Target
VAR _VerdictText  = IF(_Hit, "on track", "behind")
VAR _VerdictColor = IF(_Hit, "#2D6A2E", "#982F2F")
VAR _ActualFmt    = FORMAT(_Actual,   "$#,0,, M")
VAR _TargetFmt    = FORMAT(_Target,   "$#,0,, M")
VAR _VarFmt       = FORMAT(ABS(_Variance), "+0.0%;0.0%")

RETURN
"data:image/svg+xml;utf8," &
"<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 400 30'>" &
-- Plain prefix
"<text x='4' y='20' font-family='Segoe UI' font-size='12' fill='#333'>" &
"Sales " &
-- Bold actual value, brand color
"<tspan font-weight='700' fill='#1f4e79'>" & _ActualFmt & "</tspan>" &
" vs target " &
"<tspan fill='#555'>" & _TargetFmt & "</tspan>" &
" -- " &
-- Verdict: color changes with performance
"<tspan fill='" & _VerdictColor & "' font-weight='700'>" & _VerdictText & " (" & _VarFmt & ")</tspan>" &
"</text>" &
"</svg>"

Key notes:

• Each styled clause is a separate <tspan> carrying its own fill and font-weight; mix styled and unstyled runs in one <text> element
• Colors come from IF/SWITCH over data; use hex codes aligned with the report's theme tokens
• Format numbers inside the measure via FORMAT(); the SVG string ignores the measure's model format string
• The 32K limit and no-interactivity constraints apply; keep sentences concise
• The classic card (card) does not support ImageUrl callouts; use cardVisual, an image visual, or a table single-row image column

For the binding, wire the measure to callout.imageFX in a cardVisual (see the binding block above), or to sourceField in an image visual. To embed a tiny inline sparkline mid-sentence, add a <polyline> element after the text run within the same SVG.

---

Slicer Visual (advancedSlicerVisual)

Slicers can display SVG in header images and custom slicer items. This is less common but useful for branded slicer headers.

Binding

Set SVG in slicer header via header.image:

{
  "objects": {
    "header": [{
      "properties": {
        "image": {
          "expr": {
            "Measure": {
              "Expression": {"SourceRef": {"Schema": "extension", "Entity": "Table"}},
              "Property": "Header SVG"
            }
          }
        }
      }
    }]
  }
}

---

Design Considerations

Card SVGs Should Be Small

Card visuals have limited space. Keep SVGs compact:

• Arrow indicators: viewBox='0 0 20 20'
• Mini gauges: viewBox='0 0 100 60'
• Progress bars: viewBox='0 0 100 20'

Classic Card Does NOT Work

The classic card visual (card) does NOT support SVG measures. Only cardVisual (the "new card") works. If SVG renders as text or blank, verify the visual type is cardVisual.

Color Rules

Same as all SVG visuals: always use hex codes with # (e.g., fill='#2196F3'). Never use %23 URL encoding.

Conditional Colors for KPI States

Common color scheme for KPI indicators:

VAR _Color = SWITCH(TRUE(),
    _Performance >= 1.0,  "#4CAF50",   -- Green (exceeds target)
    _Performance >= 0.8,  "#FF9800",   -- Amber (near target)
    "#F44336"                           -- Red (below target)
)

Or using the SpaceParts sentiment pattern with 4 levels:

VAR _Color = SWITCH(TRUE(),
    _Performance < -0.05, "#f4ae4c",   -- Dark yellow (very bad)
    _Performance < -0.025, "#ffe075",  -- Light yellow (bad)
    _Performance > 0.05,  "#2D6390",   -- Dark blue (very good)
    _Performance > 0.025, "#74B2FF",   -- Light blue (good)
    "#CCCCCC"                           -- Grey (neutral)
)

SVG Elements Reference for DAX

Quick reference for SVG elements commonly used in DAX measures. All examples use single quotes for SVG attributes to avoid DAX double-quote escaping.

Rectangle

"<rect x='10' y='5' width='50' height='10' fill='#2196F3' rx='2'/>"

Circle

"<circle cx='50' cy='10' r='5' fill='#F44336'/>"

Line

"<line x1='0' y1='10' x2='100' y2='10' stroke='#333333' stroke-width='2'/>"

Polyline (Sparklines)

"<polyline fill='none' stroke='#01B8AA' stroke-width='3' points='0,50 10,30 20,40 30,10'/>"

Build points with CONCATENATEX:

VAR Lines = CONCATENATEX(Table, [X] & "," & [Y], " ", [SortColumn])

Text

"<text x='50' y='10' font-size='12' fill='#333333' font-weight='bold' text-anchor='middle' dominant-baseline='middle'>Label</text>"

Path (Arcs, Curves)

"<path d='M 10,10 L 50,10 L 30,30 Z' fill='#4CAF50'/>"

Arc for gauge/donut:

"<path d='M 10 50 A 40 40 0 0 1 90 50' fill='none' stroke='#2196F3' stroke-width='8'/>"

Group

"<g transform='translate(10,10)'>" & _Shape1 & _Shape2 & "</g>"

Gradient Definition

"<defs><linearGradient id='grad' x1='0' y1='0' x2='0' y2='1'><stop offset='0' stop-color='#0000FF'/><stop offset='1' stop-color='#00000000'/></linearGradient></defs>"

Reference in fill: fill='url(#grad)'

SVG Container

"<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'>"

Common Patterns

Responsive sizing

Use viewBox instead of fixed width/height:

"<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 30'>"

Hex colors

Always use # directly in SVG attributes -- e.g., fill='#01B8AA'. Do not use %23 URL encoding -- this causes VisualDataProxyExecutionUnknownError in image visuals and is unreliable in other visual types. Avoid named colors (blue, red) -- always use hex.

VAR LineColor = "#01B8AA"

Coordinate inversion

SVG Y=0 is at top; invert for charts:

VAR _Y = 100 - [NormalizedValue]  -- Flips so higher values go up

Render order

Elements render in document order (first = back, last = front). Place backgrounds before foreground elements.

SVG Patterns for Image Visuals

Image visuals (image visual type) render SVG measures as standalone graphics on the report canvas. Unlike table/matrix SVGs (which are inline micro-charts in rows), image visuals occupy their own visual container and can be any size.

Critical: sourceType Must Be 'imageData'

For data:image/svg+xml;utf8,... data URIs, the image visual must use sourceType = 'imageData'. Using 'imageUrl' (which is for HTTP URLs) causes VisualDataProxyExecutionUnknownError and renders black.

Binding an SVG Measure to an Image Visual

Via JSON

Create the image visual.json file manually (see pbir-format skill in the pbip plugin for JSON structure). Set sourceType to 'imageData' and bind the sourceField to the SVG measure reference (e.g., _Fmt.SparklineSVG).

Via Python API

Create the visual.json file manually (see the pbir-format skill in the pbip plugin for JSON structure).

# Example using pbir_object_model (if available):
report.add_extension_measure(
    table="Orders",
    name="KPI Header SVG",
    expression='''...SVG DAX expression...''',
    data_type="Text",
    data_category="ImageUrl",
    display_folder="SVG Charts",
)

# Create image visual and bind to the measure
page = report.pages[0]
visual = page.add_visual("image", x=100, y=50, width=400, height=200)
visual.set_image_source("measure", measure_ref="Orders.KPI Header SVG")
report.save()

JSON Structure

The image visual uses objects.image with these properties:

{
  "objects": {
    "image": [{
      "properties": {
        "sourceType": {"expr": {"Literal": {"Value": "'imageData'"}}},
        "transparency": {"expr": {"Literal": {"Value": "0D"}}},
        "sourceField": {
          "expr": {
            "Measure": {
              "Expression": {
                "SourceRef": {"Schema": "extension", "Entity": "Orders"}
              },
              "Property": "KPI Header SVG"
            }
          }
        },
        "effects": {"expr": {"Literal": {"Value": "false"}}}
      }
    }]
  }
}

Note: image visuals need no query block -- only objects.image with sourceType, sourceField, and optionally transparency/effects.

---

Pattern: KPI Header Card

A standalone SVG that shows a metric value, label, and trend indicator. Designed for image visuals.

KPI Header SVG =
VAR _Value = [Total Revenue]
VAR _PY = [Total Revenue PY]
VAR _Change = DIVIDE(_Value - _PY, _PY)
VAR _ChangeLabel = FORMAT(_Change, "+#,##0.0%;-#,##0.0%")
VAR _ValueLabel = FORMAT(_Value, "$#,##0,, M")
VAR _ChangeColor = IF(_Change >= 0, "#2D6A2E", "#982F2F")
VAR _Arrow = IF(_Change >= 0,
    "<polygon points='170,28 175,20 180,28' fill='" & _ChangeColor & "'/>",
    "<polygon points='170,20 175,28 180,20' fill='" & _ChangeColor & "'/>"
)

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 300 60'>" &
"<text x='10' y='20' font-family='Segoe UI' font-size='11' fill='#666' font-weight='600'>TOTAL REVENUE</text>" &
"<text x='10' y='48' font-family='Segoe UI' font-size='28' fill='#333' font-weight='700'>" & _ValueLabel & "</text>" &
_Arrow &
"<text x='185' y='28' font-family='Segoe UI' font-size='12' fill='" & _ChangeColor & "' font-weight='600'>" & _ChangeLabel & " vs PY</text>" &
"</svg>"

Design notes:

• Use a wide viewBox (e.g., 300x60) since image visuals are typically wider than table cells
• Include all text labels inside the SVG -- no separate visual title needed
• Font sizes can be larger than table SVGs (28px value vs 10-12px in tables)

---

Pattern: Sparkline with Endpoint Dot

A clean sparkline with a highlighted endpoint, suitable for image visuals in dashboards.

Sparkline with Endpoint =
VAR _Values = ADDCOLUMNS(
    CALCULATETABLE(VALUES('Date'[Month]),
        DATESINPERIOD('Date'[Date], MAX('Date'[Date]), -12, MONTH)),
    "@Value", [Sales Amount]
)
VAR _XMin = MIN('Date'[Month])
VAR _XMax = MAX('Date'[Month])
VAR _YMin = MINX(_Values, [@Value])
VAR _YMax = MAXX(_Values, [@Value])

VAR _Points = CONCATENATEX(
    ADDCOLUMNS(_Values,
        "@X", INT(280 * DIVIDE('Date'[Month] - _XMin, _XMax - _XMin)) + 10,
        "@Y", INT(50 * DIVIDE([@Value] - _YMin, _YMax - _YMin))
    ),
    [@X] & "," & (55 - [@Y]),
    " ",
    'Date'[Month]
)

VAR _LastX = INT(280 * 1) + 10
VAR _LastVal = MAXX(FILTER(_Values, 'Date'[Month] = _XMax), [@Value])
VAR _LastY = 55 - INT(50 * DIVIDE(_LastVal - _YMin, _YMax - _YMin))

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 300 60'>" &
"<polyline fill='none' stroke='#448FD6' stroke-width='2' points='" & _Points & "'/>" &
"<circle cx='" & _LastX & "' cy='" & _LastY & "' r='4' fill='#448FD6'/>" &
"</svg>"

---

Pattern: Multi-Metric Dashboard Tile

Combines multiple values and a mini trend line in a single image visual.

Dashboard Tile SVG =
VAR _Revenue = [Total Revenue]
VAR _Target = [Revenue Target]
VAR _Pct = DIVIDE(_Revenue, _Target)
VAR _RevLabel = FORMAT(_Revenue, "$#,0,, M")
VAR _PctLabel = FORMAT(_Pct, "0%")
VAR _BarW = _Pct * 180

-- Colors
VAR _BarColor = SWITCH(TRUE(), _Pct < 0.5, "#F44336", _Pct < 0.8, "#FF9800", "#4CAF50")

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 200 80'>" &

-- Label
"<text x='10' y='16' font-family='Segoe UI' font-size='10' fill='#999' font-weight='600'>REVENUE vs TARGET</text>" &

-- Value
"<text x='10' y='42' font-family='Segoe UI' font-size='22' fill='#333' font-weight='700'>" & _RevLabel & "</text>" &

-- Progress bar
"<rect x='10' y='52' width='180' height='8' fill='#E8E8E8' rx='4'/>" &
"<rect x='10' y='52' width='" & _BarW & "' height='8' fill='" & _BarColor & "' rx='4'/>" &

-- Percentage label
"<text x='10' y='72' font-family='Segoe UI' font-size='10' fill='" & _BarColor & "' font-weight='600'>" & _PctLabel & " of target</text>" &

"</svg>"

---

Design Considerations for Image Visuals

viewBox Sizing

Image visuals are flexible in size. Use a viewBox that matches the visual's aspect ratio:

Colors

Always use hex codes with # (e.g., fill='#2196F3'). Never use %23 URL encoding -- this causes rendering failures in image visuals. Named colors like blue or red are unreliable; always use hex.

Transparency and Effects

Disable default image effects to prevent unwanted styling:

"transparency": {"expr": {"Literal": {"Value": "0D"}}},
"effects": {"expr": {"Literal": {"Value": "false"}}}

No Query Block

Image visuals bound to SVG measures do not need a query block. The measure is evaluated directly via sourceField. Adding a query block can cause duplicate evaluations.

---

Dynamic and Responsive SVG

Responsive width (image and card only)

Set the SVG root to width='100%' height='100%' and drive geometry off the viewBox; the host scales the rendered image to fill its container:

VAR _Prefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' width='100%' height='100%' viewBox='0 0 300 60' preserveAspectRatio='xMidYMid meet'>"

This works in image visuals and card callouts. Table and matrix cells ignore percentage width; they use grid.imageWidth from the visual's objects.grid. Fix the viewBox geometry to the configured imageWidth when targeting table cells.

SVG can scale to fill the host, but it cannot read the container width as a number, so true reflow (re-binning bars to available width) is not possible in a DAX SVG measure. Reach for Deneb when the layout must adapt to container width.

Conditional layout based on context

Branch the entire element assembly on filter context so the measure emits a different shape depending on what is selected, keeping each branch under the 32K string ceiling:

Conditional SVG =
VAR _Mode =
    SWITCH(
        TRUE(),
        ISBLANK([Sales Amount]),       "empty",
        HASONEVALUE('Date'[Month]),    "single",
        "trend"
    )

-- Assemble the data-URI prefix once, outside the branch
VAR _Prefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 300 60'>"
VAR _Suffix = "</svg>"

-- Build only the branch that will be returned
VAR _EmptyMarkup  = "<text x='10' y='35' font-family='Segoe UI' font-size='11' fill='#999'>n/a</text>"
VAR _BarMarkup    = -- single-bar assembly (HASONEVALUE context)
    "<rect x='10' y='20' width='" & INT(DIVIDE([Sales Amount], [Sales Target]) * 200) & "' height='20' fill='#448FD6'/>"
VAR _SparkMarkup  = -- sparkline assembly (multi-period context)
    "<polyline fill='none' stroke='#448FD6' stroke-width='2' points='" &
    CONCATENATEX(
        ADDCOLUMNS(VALUES('Date'[Month]), "@V", [Sales Amount]),
        INT(RANKX(VALUES('Date'[Month]), 'Date'[Month], , ASC) * 25) & "," &
        INT(50 - DIVIDE([@V], MAXX(VALUES('Date'[Month]), [Sales Amount])) * 40),
        " ", 'Date'[Month]
    ) & "'/>"

VAR _Body =
    SWITCH(
        _Mode,
        "empty",  _EmptyMarkup,
        "single", _BarMarkup,
        _SparkMarkup
    )

RETURN _Prefix & _Body & _Suffix

Key rules for conditional layout measures:

• Always include a blank/empty-state branch so filter contexts with no data emit a readable fallback rather than a degenerate zero-width SVG or BLANK
• Assemble the data-URI prefix and </svg> suffix once outside the SWITCH so the empty branch cannot accidentally omit them
• Switch on report/page-level state for the whole measure (e.g., HASONEVALUE over a slicer's dimension); switching per row on a different condition confuses readers who see different chart metaphors in adjacent cells
• Subtotal and total rows should usually return the empty markup or BLANK() rather than a coarced single-value chart

SVG Patterns for Table and Matrix Visuals

Table (tableEx) and Matrix (pivotTable) visuals are the primary target for DAX SVG measures. Configure grid.imageHeight and grid.imageWidth in visual objects to control rendering size (default: 25px height, 100px width).

Setup

Image Size Configuration

Set in the visual's objects.grid:

"grid": [{
  "properties": {
    "imageHeight": {"expr": {"Literal": {"Value": "25D"}}},
    "imageWidth": {"expr": {"Literal": {"Value": "100D"}}}
  }
}]

Sort Trick

Embed a <desc> tag to enable sorting the SVG column by bar length:

VAR _Sort = "<desc>" & FORMAT(_Actual, "000000000000") & "</desc>"

Power BI uses the <desc> content as the sort key for the image column.

Axis Normalization (Required for All Bar-Based Charts)

All bar, bullet, and dumbbell charts need a shared axis maximum so bars are comparable across rows:

VAR _BarMax = 100          -- max pixel width of the bar area
VAR _BarMin = 20           -- left offset (space for labels/dots)
VAR _Scope = ALLSELECTED('Table'[GroupColumn])

VAR _MaxActual = CALCULATE(
    MAXX(_Scope, [Actual]),
    REMOVEFILTERS('Table'[GroupColumn])
)
VAR _MaxTarget = CALCULATE(
    MAXX(_Scope, [Target]),
    REMOVEFILTERS('Table'[GroupColumn])
)

VAR _AxisMax =
    IF(
        HASONEVALUE('Table'[GroupColumn]),
        MAX(_MaxActual, _MaxTarget),
        CALCULATE(MAX([Actual], [Target]), REMOVEFILTERS('Table'[GroupColumn]))
    ) * 1.1   -- 10% headroom

VAR _AxisRange = _BarMax - _BarMin
VAR _ActualNormalized = DIVIDE(_Actual, _AxisMax) * _AxisRange
VAR _TargetNormalized = (DIVIDE(_Target, _AxisMax) * _AxisRange) + _BarMin - 1

Key points:

• REMOVEFILTERS on the group column ensures _AxisMax is consistent across all rows
• Multiply by 1.1 for headroom so bars never hit the edge
• Target position = normalized value + left offset

Number Formatting (Adaptive Scale)

VAR _Label = SWITCH(TRUE(),
    _Actual <= 1E3,  FORMAT(_Actual, "#,0"),
    _Actual <= 1E6,  FORMAT(_Actual, "#,0, K"),
    _Actual <= 1E9,  FORMAT(_Actual, "#,0,, M"),
    FORMAT(_Actual, "#,0,,, B")
)

---

Pattern: Data Bar

Simple proportional bar for table columns. The most basic SVG pattern.

Data Bar =
VAR _Value = [Sales Amount]
VAR _Max = CALCULATE(MAX([Sales Amount]), REMOVEFILTERS('Product'[Category]))
VAR _Pct = DIVIDE(_Value, _Max)
VAR _W = _Pct * 100
VAR _Color = "#5B8DBE"

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 16'>" &
"<rect width='" & _W & "' height='16' fill='" & _Color & "' opacity='0.7' rx='2'/>" &
"<text x='" & (_W + 3) & "' y='12' font-size='10' fill='#333'>" &
FORMAT(_Value, "#,0") & "</text></svg>"

Variants:

• Rounded corners: add rx='4' to the rect
• Rounded tops only: use rx equal to half the height
• Conditional color: VAR _Color = IF(_Value > _Threshold, "#4CAF50", "#F44336")

---

Pattern: Bullet Chart with Action Dots

Combines actual bar, target line, baseline, and a sentiment-colored dot. From the SpaceParts production model.

SVG Bullet Chart =
-- Config
VAR _Actual = [MTD Turnover]
VAR _Target = [MTD Turnover 1YP]
VAR _Performance = DIVIDE(_Actual - _Target, _Target)

-- Sentiment thresholds
VAR _VeryBad  = -0.05
VAR _Bad      = -0.025
VAR _Good     =  0.025
VAR _VeryGood =  0.05

-- Chart dimensions
VAR _BarMax = 100
VAR _BarMin = 20
VAR _Scope = ALL('Customers'[Key Account Name])

-- Colors
VAR _BackgroundColor = "#F5F5F5"
VAR _BarFillColor    = "#CFCFCF"
VAR _BaselineColor   = "#737373"
VAR _TargetColor     = "#000000"

VAR _ActionDotFill =
    SWITCH(TRUE(),
        _Performance < _VeryBad, "#f4ae4c",
        _Performance < _Bad,     "#ffe075",
        _Performance > _VeryGood, "#2D6390",
        _Performance > _Good,    "#74B2FF",
        "#FFFFFF00"
    )

-- Axis normalization
VAR _MaxActual = CALCULATE(MAXX(_Scope, [MTD Turnover]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _MaxTarget = CALCULATE(MAXX(_Scope, [MTD Turnover 1YP]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _AxisMax =
    IF(HASONEVALUE('Customers'[Key Account Name]),
        MAX(_MaxActual, _MaxTarget),
        CALCULATE(MAX([MTD Turnover], [MTD Turnover 1YP]), REMOVEFILTERS('Customers'[Key Account Name]))
    ) * 1.1

VAR _AxisRange = _BarMax - _BarMin
VAR _ActualNormalized = DIVIDE(_Actual, _AxisMax) * _AxisRange
VAR _TargetNormalized = (DIVIDE(_Target, _AxisMax) * _AxisRange) + _BarMin - 1

-- SVG construction
VAR _SvgPrefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'>"
VAR _Sort = "<desc>" & FORMAT(_Actual, "000000000000") & "</desc>"

VAR _ActionDot    = "<circle cx='10' cy='11' r='5' fill='" & _ActionDotFill & "'/>"
VAR _BarBg        = "<rect x='" & _BarMin & "' y='2' width='" & _BarMax & "' height='80%' fill='" & _BackgroundColor & "'/>"
VAR _ActualBar    = "<rect x='" & _BarMin & "' y='5' width='" & _ActualNormalized & "' height='50%' fill='" & _BarFillColor & "'/>"
VAR _Baseline     = "<rect x='" & _BarMin & "' y='4' width='1' height='60%' fill='" & _BaselineColor & "'/>"
VAR _TargetLine   = "<rect x='" & _TargetNormalized & "' y='2' width='2' height='80%' fill='" & _TargetColor & "'/>"

VAR _SvgSuffix = "</svg>"

RETURN
    _SvgPrefix & _Sort & _ActionDot & _BarBg & _ActualBar & _Baseline & _TargetLine & _SvgSuffix

Key elements:

• Action dot at left edge shows sentiment at a glance (colored circle)
• Background rect provides visual context for the bar area
• Target line as a thin rect (width 2px) at the target position
• Baseline at the left edge of the bar area
• <desc> for sort ordering

---

Pattern: Overlapping Bars with Variance

Two bars (actual on top, target behind) with a colored variance indicator and label. From SpaceParts.

SVG Overlapping Bars =
-- Config
VAR _Actual = [Actuals MTD]
VAR _Target = [Budget MTD]
VAR _Performance = DIVIDE(_Actual - _Target, _Target)

-- Font
VAR _Font = "Segoe UI"
VAR _FontSize = 10
VAR _FontWeight = 600

-- Chart dimensions
VAR _BarMax = 100
VAR _BarMin = 30
VAR _Scope = ALLSELECTED('Customers'[Key Account Name])

-- Colors
VAR _ActualColor   = "#686868"
VAR _TargetColor   = "#e1dfdd"
VAR _VarianceColor = IF(_Performance < 0, "#fab005", "#2094ff")

-- Axis normalization
VAR _MaxActual = CALCULATE(MAXX(_Scope, [Actuals MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _MaxTarget = CALCULATE(MAXX(_Scope, [Budget MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _AxisMax =
    IF(HASONEVALUE('Customers'[Key Account Name]),
        MAX(_MaxActual, _MaxTarget),
        CALCULATE(MAX([Actuals MTD], [Budget MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
    ) * 1.1

VAR _AxisRange = _BarMax - _BarMin
VAR _ActualNormalized = DIVIDE(_Actual, _AxisMax) * _AxisRange
VAR _TargetNormalized = DIVIDE(_Target, _AxisMax) * _AxisRange

-- SVG construction
VAR _SvgPrefix = "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'>"
VAR _Sort = "<desc>" & FORMAT(_Actual, "000000000000") & "</desc>"

VAR _Icon  = "<text x='" & _BarMin - 3 & "' y='13.5' font-family='Segoe UI' font-size='6' font-weight='700' text-anchor='end' fill='" & _VarianceColor & "'>" & FORMAT(_Performance, "^^^;vvv;") & "</text>"
VAR _Label = "<text x='" & _BarMin - 10 & "' y='15' font-family='" & _Font & "' font-size='" & _FontSize & "' font-weight='" & _FontWeight & "' text-anchor='end' fill='" & _VarianceColor & "'>" & FORMAT(_Performance, "#,##0%;#,##0%;#,##0%") & "</text>"

VAR _TargetBar   = "<rect x='" & _BarMin & "' y='10' width='" & _TargetNormalized & "' height='12' stroke='" & _ActualColor & "' fill='" & _TargetColor & "'/>"
VAR _ActualBar   = "<rect x='" & _BarMin & "' y='3' width='" & _ActualNormalized & "' height='12' stroke='" & _ActualColor & "' fill='" & _ActualColor & "'/>"
VAR _VarianceBar = "<rect x='" & _BarMin + MIN(_ActualNormalized, _TargetNormalized) + 1 & "' y='" & IF(_Target > _Actual, 2.9, 9) & "' width='" & ABS(_ActualNormalized - _TargetNormalized) - 1 & "' height='6' stroke='" & _VarianceColor & "' fill='" & _VarianceColor & "'/>"

VAR _SvgSuffix = "</svg>"

RETURN
    _SvgPrefix & _Sort & _Icon & _Label & _TargetBar & _ActualBar & _VarianceBar & _SvgSuffix

Key elements:

• Variance label + directional icon at the left
• Target bar behind (wider height offset), actual bar on top
• Variance highlight bar spans the gap between actual and target
• Variance bar position (y) shifts based on whether actual > target

---

Pattern: Dumbbell Chart

Compares two values as circles connected by a line. From SpaceParts.

SVG Dumbbell Chart =
-- Config
VAR _Actual = [Actuals MTD]
VAR _Target = [Sales Target MTD]

-- Chart dimensions
VAR _SvgWidth = 100
VAR _SvgHeight = 25
VAR _Scope = ALLSELECTED('Customers'[Key Account Name])

-- Axis normalization
VAR _MaxActual = CALCULATE(MAXX(_Scope, [Actuals MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _MaxTarget = CALCULATE(MAXX(_Scope, [Sales Target MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
VAR _AxisMax =
    IF(HASONEVALUE('Customers'[Key Account Name]),
        MAX(_MaxActual, _MaxTarget),
        CALCULATE(MAX([Actuals MTD], [Sales Target MTD]), REMOVEFILTERS('Customers'[Key Account Name]))
    ) * 1.1

VAR _AxisRange = _SvgWidth
VAR _ActualNormalized = DIVIDE(_Actual, _AxisMax) * _AxisRange
VAR _TargetNormalized = DIVIDE(_Target, _AxisMax) * _AxisRange

-- Colors (conditional: blue if on target, red if off)
VAR _AxisColor = "#C7C7C7"
VAR _Fill   = IF(_Actual > _Target, "#448FD6", "#D64444")
VAR _Stroke = IF(_Actual > _Target, "#2F6698", "#982F2F")

-- SVG construction
VAR _SvgPrefix = "data:image/svg+xml;utf8,<svg width='" & _SvgWidth & "' height='" & _SvgHeight & "' xmlns='http://www.w3.org/2000/svg'>"
VAR _Sort = "<desc>" & FORMAT(_Actual, "000000000000") & "</desc>"

VAR _Axis         = "<line x1='0' y1='" & _SvgHeight / 2 & "' x2='" & _SvgWidth & "' y2='" & _SvgHeight / 2 & "' stroke='" & _AxisColor & "'/>"
VAR _Origin       = "<circle cx='2' cy='" & _SvgHeight / 2 & "' r='2' fill='" & _AxisColor & "'/>"
VAR _DumbbellLine = "<line x1='" & _ActualNormalized & "' y1='" & _SvgHeight / 2 & "' x2='" & _TargetNormalized & "' y2='" & _SvgHeight / 2 & "' stroke='" & _Fill & "' stroke-width='3'/>"
VAR _TargetCircle = "<circle cx='" & _TargetNormalized & "' cy='" & _SvgHeight / 2 & "' r='4' fill='#F5F5F5' stroke='#C7C7C7' stroke-width='1.5'/>"
VAR _ActualCircle = "<circle cx='" & _ActualNormalized & "' cy='" & _SvgHeight / 2 & "' r='4' fill='" & _Fill & "' stroke='" & _Stroke & "' stroke-width='1.5'/>"

VAR _SvgSuffix = "</svg>"

RETURN
    _SvgPrefix & _Sort & _Axis & _Origin & _DumbbellLine & _TargetCircle & _ActualCircle & _SvgSuffix

Key elements:

• Horizontal axis line spanning the full width
• Origin dot at left edge
• Connecting line between actual and target (colored by performance)
• Target circle: neutral grey fill, grey stroke
• Actual circle: blue (on target) or red (off target)
• Render order matters: line first, then target circle, then actual circle on top

---

Pattern: Status Pill

Colored pill with text label for categorical status. From MacGuyver Toolbox.

Status Pill =
VAR _Status = [Status Category]
VAR _BgColor = SWITCH(_Status,
    "On Track",  "#CEE5D0",
    "At Risk",   "#FFF3CD",
    "Late",      "#EBD2CE",
    "#F0F0F0"
)
VAR _TextColor = SWITCH(_Status,
    "On Track",  "#2D6A2E",
    "At Risk",   "#856404",
    "Late",      "#982F2F",
    "#333333"
)

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'>" &
"<rect x='0.5' y='0.5' width='98%' height='95%' rx='15%' fill='" & _BgColor & "' stroke='" & _TextColor & "'/>" &
"<text x='50%' y='58%' font-family='Segoe UI' font-size='12' font-weight='700' fill='" & _TextColor & "' text-anchor='middle' dominant-baseline='middle'>" & UPPER(_Status) & "</text></svg>"

---

Pattern: Lollipop Chart

Thin line with proportionally sized dot at the end. From MacGuyver Toolbox.

Lollipop Chart =
VAR _Actual = [Sales Amount]
VAR _Target = [Sales Target]
VAR _BarMax = 75
VAR _BarMin = 20
VAR _Scope = ALL('Product'[Category])

VAR _MaxVal = CALCULATE(MAX([Sales Amount], [Sales Target]), REMOVEFILTERS('Product'[Category])) * 1.1
VAR _AxisRange = _BarMax - _BarMin

VAR _ActualNormalized = DIVIDE(_Actual, _MaxVal) * _AxisRange
VAR _DotRadius = MAX(DIVIDE(_Actual, _MaxVal) * 7.5, 3.5)

VAR _Performance = DIVIDE(_Actual - _Target, _Target)
VAR _Color = IF(_Performance >= 0, "#448FD6", "#D64444")

VAR _Label = SWITCH(TRUE(),
    _Actual <= 1E3,  FORMAT(_Actual, "#,0"),
    _Actual <= 1E6,  FORMAT(_Actual, "#,0, K"),
    FORMAT(_Actual, "#,0,, M")
)

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg'>" &
"<desc>" & FORMAT(_Actual, "000000000000") & "</desc>" &
"<rect x='" & _BarMin & "' y='12' width='" & _ActualNormalized & "' height='1.5' fill='" & _Color & "'/>" &
"<circle cx='" & (_BarMin + _ActualNormalized) & "' cy='12.5' r='" & _DotRadius & "' fill='" & _Color & "'/>" &
"<text x='" & _BarMin - 3 & "' y='15' font-family='Segoe UI' font-size='9' font-weight='600' text-anchor='end' fill='" & _Color & "'>" & _Label & "</text></svg>"

Key elements:

• Dot radius scales proportionally with value: MAX(DIVIDE(_Actual, _MaxVal) * 7.5, 3.5)
• Thin line (height 1.5px) connects the left edge to the dot
• Label positioned to the left of the bar area

---

Pattern: Sparkline (Polyline)

Line sparkline showing trend over time using CONCATENATEX.

Sparkline SVG =
VAR _Values = ADDCOLUMNS(
    CALCULATETABLE(
        VALUES('Date'[Month]),
        DATESINPERIOD('Date'[Date], MAX('Date'[Date]), -12, MONTH)
    ),
    "@Value", [Sales Amount]
)
VAR _XMin = MIN('Date'[Month])
VAR _XMax = MAX('Date'[Month])
VAR _YMin = MINX(_Values, [@Value])
VAR _YMax = MAXX(_Values, [@Value])

VAR _Points = CONCATENATEX(
    ADDCOLUMNS(_Values,
        "@X", INT(100 * DIVIDE('Date'[Month] - _XMin, _XMax - _XMin)),
        "@Y", INT(30 * DIVIDE([@Value] - _YMin, _YMax - _YMin))
    ),
    [@X] & "," & (30 - [@Y]),
    " ",
    'Date'[Month]
)

RETURN
IF(HASONEVALUE('Product'[Category]),
    "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 30'>" &
    "<polyline fill='none' stroke='#5B8DBE' stroke-width='2' points='" & _Points & "'/></svg>",
    BLANK()
)

Key technique: Y is inverted (30 - [@Y]) because SVG Y=0 is at the top.

---

Pattern: Area Sparkline with Gradient

Area Sparkline =
VAR Defs = "<defs><linearGradient id='grad' x1='0' y1='0' x2='0' y2='50' gradientUnits='userSpaceOnUse'><stop stop-color='navy' offset='0'/><stop stop-color='transparent' offset='1'/></linearGradient></defs>"

VAR XMin = MIN('Table'[Date])
VAR XMax = MAX('Table'[Date])
VAR YMin = MINX(VALUES('Table'[Date]), CALCULATE([Measure]))
VAR YMax = MAXX(VALUES('Table'[Date]), CALCULATE([Measure]))

VAR Points = ADDCOLUMNS(
    SUMMARIZE('Table', 'Table'[Date]),
    "X", INT(150 * DIVIDE('Table'[Date] - XMin, XMax - XMin)),
    "Y", INT(50 * DIVIDE([Measure] - YMin, YMax - YMin))
)

VAR Lines = CONCATENATEX(Points, [X] & "," & (50 - [Y]), " ", 'Table'[Date])

RETURN IF(HASONEVALUE('Table'[Category]),
    "data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 150 50'>" &
    Defs &
    "<polyline fill='url(#grad)' fill-opacity='0.3' stroke='navy' stroke-width='3' points='0 50 " & Lines & " 150 50 Z'/></svg>",
    BLANK())

---

Pattern: Bar Sparkline

Bar Sparkline =
VAR _Values = ADDCOLUMNS(
    CALCULATETABLE(VALUES('Date'[Month]),
        DATESINPERIOD('Date'[Date], MAX('Date'[Date]), -12, MONTH)),
    "@Value", [Sales Amount]
)
VAR _Count = COUNTROWS(_Values)
VAR _Max = MAXX(_Values, [@Value])
VAR _Min = MINX(_Values, [@Value])
VAR _Range = _Max - _Min
VAR _W = 100
VAR _H = 30
VAR _BarW = _W / _Count
VAR _Scale = _H / _Range

VAR _Bars = CONCATENATEX(
    ADDCOLUMNS(_Values, "@Idx", RANKX(_Values, 'Date'[Month], , ASC)),
    VAR _X = ([@Idx] - 1) * _BarW
    VAR _BarH = ([@Value] - _Min) * _Scale
    VAR _Y = _H - _BarH
    RETURN "<rect x='" & _X & "' y='" & _Y & "' width='" & (_BarW * 0.8) &
           "' height='" & _BarH & "' fill='#2196F3' opacity='0.7'/>",
    "", [@Idx], ASC
)

RETURN
"data:image/svg+xml;utf8,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 " & _W & " " & _H & "'>" & _Bars & "</svg>"

---

UDF Pattern

For reusable SVG charts, use DAX UDFs (User-Defined Functions) that accept parameters. These are defined as extension measures in a dedicated __SVGs or similar table.

-- Calling a UDF
SVG.Chart.BulletChart.ActionDot(
    [MTD Turnover],                        -- Actual measure
    [MTD Turnover 1YP],                    -- Target measure
    'Customers'[Key Account Name],         -- Group column
    -0.050,                                -- Very bad threshold
    -0.025,                                -- Bad threshold
     0.025,                                -- Good threshold
     0.050,                                -- Very good threshold
    "#f4ae4c",                             -- Very bad color
    "#ffe075",                             -- Bad color
    "#74B2FF",                             -- Good color
    "#2D6390"                              -- Very good color
)

Available UDF libraries:

• SVG.Chart.BarChart.OverlappingBars -- overlapping bars with variance label
• SVG.Chart.BarChart.OverlappingBarsSimple -- overlapping bars without label
• SVG.Chart.BulletChart.ActionDot -- bullet chart with sentiment dots
• SVG.Chart.DumbbellPlot -- dumbbell comparison chart
• SVG.Chart.Waterfall -- waterfall chart with connectors

See the PowerBI MacGuyver Toolbox and DaxLib.SVG libraries for more UDFs.

---

Performance in Table and Matrix Cells

An SVG measure is a string-building DAX expression evaluated once per visible cell (per row in a table; per row x column intersection in a matrix), all in the single-threaded formula engine with no storage-engine acceleration for the string concatenation. Cost scales with iteration count, not byte size.

Pre-aggregate in model measures, not inside the SVG string

A sparkline that runs CONCATENATEX over 24 monthly points for every row is ~720 point-iterations of string assembly per page refresh, re-evaluated on every cross-filter. Push the base aggregations into reusable model measures or a precomputed table so the storage-engine cache absorbs that work; the SVG measure should only map computed numbers to coordinates.

Prefer one <polyline> over N individual elements

A single CONCATENATEX producing a points='...' string for <polyline> iterates once and produces one element. Replacing it with N <circle> or <line> elements runs N iterations and emits N elements for the parser. Use <polyline> for line/area sparklines; reserve individual shapes for endpoints or markers only.

Round coordinates to integers

Shorter strings mean cheaper FORMAT calls and less markup:

-- Preferred: integer coordinates
VAR _X = INT(DIVIDE(_Month - _XMin, _XMax - _XMin) * 100)
-- Avoid: floating-point strings like "47.3829..."

The 32K limit is per rendered cell, not per expression

The ceiling applies to the string returned by the measure for each individual cell. Diagnose by running LEN([Your SVG Measure]) via pbir model -q for a worst-case category member. Near the ceiling, the cell silently drops to blank text. Options when approaching the limit:

• Reduce the time-series window (12 months instead of 24)
• Split into two simpler measures rendered in adjacent columns
• Move the visualization to Deneb, which has no comparable string ceiling

The HASONEVALUE/ISINSCOPE total guard matters for both correctness and cost

A matrix evaluates SVG measures at subtotal and grand-total rows too. Without a guard, the measure runs at a coarser grain and may return a meaningless or oversized SVG. Gate explicitly:

IF(
    HASONEVALUE('Product'[Category]),
    -- SVG assembly
    BLANK()
)

For nested matrices, ISINSCOPE('Product'[SubCategory]) targets a specific hierarchy level and avoids building SVGs at every subtotal band.

Caching and volatile inputs

Identical SVG strings for identical filter contexts hit the formula engine result cache. Inputs that change every evaluation defeat caching: NOW(), TODAY(), and unseed random jitter. The jitter-plot example uses a hash-based pseudo-random from a key column; preserve that pattern rather than introducing RAND().

grid.imageHeight/grid.imageWidth and column width

Setting large image dimensions in the visual's objects.grid does not change the formula-engine cost of generating the SVG, but it adds rasterization and paint cost in the renderer. Keep dimensions proportional to the cell content; avoid inflating them for visual effect. Auto-size-width ON with a wide SVG column triggers horizontal scroll and re-layout churn; set an explicit columnWidth in the visual formatting.