Click Generate…

You are a senior Power BI architect and performance specialist. I have shared a folder of TMDL files exported from a Power BI Desktop semantic model. The folder is available at: [paste your OneDrive URL here]

Please perform a full technical review of this semantic model and produce a structured report covering the following six areas:

SECTION 1 — BUSINESS CONTEXT
Before any technical analysis, explain in plain English:
- What business domain this model appears to serve (sales, finance, HR, operations, etc.)
- What types of analytical questions the model is designed to answer
- Who the likely audience is based on the measure names and table structure
- Any domain-specific patterns or assumptions you can identify

SECTION 2 — MODEL STRUCTURE REVIEW
Provide a complete inventory of:
- All tables (fact vs dimension, whether they follow star schema)
- All relationships (cardinality, direction, any many-to-many or bidirectional)
- All calculated columns (flag any that should be measures instead)
- All parameters and what they control
- Data sources and connection types
- Flag any tables that appear to be staging, intermediate, or unused

SECTION 3 — POWER QUERY PERFORMANCE REVIEW
Analyse every Power Query (M) script and identify:
- Steps that break query folding (list each one by query name and step)
- Unnecessary Table.Buffer, List.Buffer, or row-by-row operations
- Redundant merge or expand steps that could be eliminated
- Unused queries or staging queries that are loaded but not needed
- Transformations that should be pushed upstream to the source
- Provide a rewritten version of the worst-performing query as an example

SECTION 4 — DAX QUALITY AND PERFORMANCE REVIEW
Review every measure and calculated column for:
- Use of FILTER() as a table argument (should use predicate syntax instead)
- Iterator functions (SUMX, AVERAGEX) over large tables without necessity
- Calculated columns that add to model size and could be measures
- Missing use of DIVIDE() instead of direct division (divide-by-zero risk)
- Measures that use CALCULATE() with ALL() when REMOVEFILTERS() is cleaner
- Dependency chains that are unnecessarily long or complex
- Any time intelligence that will break if the date table is not marked
Provide an optimised rewrite for each problematic measure you find.

SECTION 5 — DATA MODEL BEST PRACTICES AUDIT
Score the model against these 10 best practices (Pass / Fail / Partial for each):
1. Star schema design with clear fact and dimension separation
2. All relationships use single-direction cross-filtering
3. Date table is present, marked, and covers the full data range
4. No high-cardinality text columns in fact tables
5. Numeric columns use correct data types (not stored as text)
6. Calculated columns minimised (prefer measures)
7. Unused columns are hidden or removed
8. Consistent naming convention across tables, columns and measures
9. Measures are organised in display folders
10. Auto date/time intelligence is disabled (custom date table used instead)

SECTION 6 — COPILOT AND Q&A READINESS
Evaluate the model for use with Power BI Copilot, Q&A, and AI-driven features:
- Which measure and column names are ambiguous or would confuse a language model
- Which tables or fields lack descriptions that should have them
- Which fields should be hidden from the Q&A surface
- Specific renaming recommendations with before/after examples
- Any relationship or schema ambiguity that would produce wrong AI-generated results

OUTPUT FORMAT
- Use clear section headers matching the six areas above
- Use bullet points for findings, not paragraphs
- Include specific DAX or M code snippets when referencing issues
- End each section with a priority-ordered action list (High / Medium / Low impact)
- Be specific — reference actual table names, measure names, and column names from the files

Using the assessment you just completed, create a prioritised step-by-step fix plan I can hand directly to a Power BI developer with intermediate skills.

Structure the plan as follows:

PRIORITY 1 — CRITICAL FIXES (Do these first — high impact on performance or correctness)
For each fix:
- Problem: What is wrong and why it matters
- Location: Exact table/measure/query name in the model
- Fix: Step-by-step instructions to resolve it in Power BI Desktop
- Code: Before and after DAX or M code where applicable
- Impact: What will improve after the fix (speed, size, accuracy)
- Refresh required: Yes / No

PRIORITY 2 — IMPORTANT IMPROVEMENTS (Do these second — meaningful quality gains)
[Same structure as Priority 1]

PRIORITY 3 — BEST PRACTICE ALIGNMENTS (Do these when time allows — polish and maintainability)
[Same structure as Priority 1]

Rules for this plan:
- Every fix must be actionable by someone who did not write the original model
- Do not use vague language like "consider refactoring" — be precise
- If a DAX rewrite is needed, provide the complete new measure, not a partial snippet
- If a Power Query fix is needed, show the complete revised step or query
- Order fixes within each priority group by effort — quickest wins first

Folder reference: [paste your OneDrive URL here]

You are preparing this Power BI semantic model for production use with Microsoft Copilot, Power BI Q&A, and future AI agent integrations. Using the model files at [paste your OneDrive URL here], produce a Copilot Readiness Checklist.

For each item below, give a status (✓ Ready / ✗ Action needed / ⚠ Partial) and specific instructions:

1. MEASURE NAMING
   - Are all measures named in plain English that a non-technical user would understand?
   - List every measure that uses abbreviations, codes, or technical terms
   - Provide the recommended plain-English rename for each

2. TABLE AND COLUMN NAMING
   - Are table names and column names human-readable?
   - Flag any PascalCase, snake_case, or coded names that should be display-name friendly
   - Provide rename recommendations

3. FIELD DESCRIPTIONS
   - Which tables, measures, and key columns are missing descriptions?
   - Write a ready-to-use description for each missing one (copy-paste into Model view)

4. HIDDEN FIELDS AUDIT
   - Which columns are currently visible that should be hidden (foreign keys, internal IDs, technical flags)?
   - Which measures or columns are hidden that should be visible for Q&A?
   - Provide the exact list to show/hide

5. SYNONYM RECOMMENDATIONS
   - For the top 10 most-used measures and dimensions, suggest 2–3 synonyms each that users might naturally type in Q&A (e.g. "revenue" → "sales", "income", "turnover")

6. RELATIONSHIP CLARITY FOR AI
   - Are there any relationships where the direction or cardinality could confuse an AI agent?
   - Flag any ambiguous join paths and recommend resolution

7. COPILOT VISUAL SUGGESTIONS READINESS
   - Will Copilot be able to generate meaningful visuals from this model?
   - What is the single most impactful change to make the model more Copilot-friendly?

Output as a checklist I can work through field by field in Power BI Desktop Model view.

Trend Arrow =
VAR _Current = [Gross Margin %]
VAR _Prior = [Prior Period Margin %]
VAR _Direction = IF( _Current >= _Prior, "up", "down" )
VAR _Color = IF( _Direction = "up", "#22C55E", "#EF4444" )
VAR _ArrowUp = "M10,2 L18,14 L13,14 L13,22 L7,22 L7,14 L2,14 Z"
VAR _ArrowDown = "M10,22 L18,10 L13,10 L13,2 L7,2 L7,10 L2,10 Z"
VAR _Path = IF( _Direction = "up", _ArrowUp, _ArrowDown )
VAR _Change = FORMAT( ABS( _Current - _Prior ), "0.0%" )
VAR _SVG =
    "data:image/svg+xml;utf8," &
    "<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 80 24'>" &
        "<path d='" & _Path & "' fill='" & _Color & "' transform='scale(0.8) translate(0,1)'/>" &
        "<text x='24' y='17' font-family='Segoe UI,sans-serif' font-size='12' font-weight='600' fill='" & _Color & "'>" &
            IF( _Direction = "up", "+" , "-" ) & _Change &
        "</text>" &
    "</svg>"
RETURN
    _SVG

YoY Color =
VAR _Growth = [YoY Growth %]
RETURN
    SWITCH(
        TRUE(),
        _Growth >= 0.10, "#15803D",   -- Dark green: 10%+ growth
        _Growth >= 0,    "#22C55E",   -- Green: positive growth
        _Growth >= -0.10, "#EF4444",  -- Red: negative growth
        "#991B1B"                      -- Dark red: 10%+ decline
    )
-- Usage: Conditional formatting > Font color > Field value > select this measure

{
  "name": "Enterprise KPI Theme",
  "dataColors": [
    "#2563EB", "#7C3AED", "#059669", "#D97706",
    "#DC2626", "#0891B2", "#4F46E5", "#15803D"
  ],
  "tableAccent": "#2563EB",
  "good": "#22C55E",
  "neutral": "#EAB308",
  "bad": "#EF4444",
  "maximum": "#15803D",
  "center": "#EAB308",
  "minimum": "#991B1B"
}

-- Naming convention: _SVG.[Type].[Context]
-- Examples:
_SVG.TrafficLight.SalesPerformance
_SVG.Sparkline.RevenueByMonth
_SVG.ProgressBar.TargetCompletion
_SVG.BulletChart.BudgetVsActual
_SVG.StarRating.CustomerSatisfaction

Traffic Light (Config-Driven) =
VAR _Score = [Score %]
VAR _KPIName = SELECTEDVALUE( KPI_Thresholds[KPI_Name] )
VAR _GreenMin =
    LOOKUPVALUE(
        KPI_Thresholds[Green_Min],
        KPI_Thresholds[KPI_Name], _KPIName
    )
VAR _YellowMin =
    LOOKUPVALUE(
        KPI_Thresholds[Yellow_Min],
        KPI_Thresholds[KPI_Name], _KPIName
    )
VAR _Color =
    SWITCH(
        TRUE(),
        _Score >= _GreenMin,  "#22C55E",
        _Score >= _YellowMin, "#EAB308",
        "#EF4444"
    )
VAR _SVG =
    "data:image/svg+xml;utf8," &
    "<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 20 20'>" &
        "<circle cx='10' cy='10' r='9' fill='" & _Color & "'/>" &
    "</svg>"
RETURN
    _SVG