Power BI Campaign ROI Analyzer Skill

Skill by ara.so — Marketing Skills collection.

This skill enables AI agents to work with the Power BI Campaign ROI Analyzer, a comprehensive social media marketing analytics dashboard that transforms raw campaign data into actionable insights. The system tracks multi-channel performance, product sales attribution, demographic segmentation, and predictive trend scoring across Facebook, Instagram, LinkedIn, TikTok, and other platforms.

What This Project Does

The Campaign ROI Analyzer is a Power BI-based analytics platform that:

Multi-Dimensional Campaign Analysis: Tracks reach, engagement, conversion, and retention metrics with weighted business value scoring
Product Sales Attribution: Breaks down sales by social channel with quality metrics (AOV, repeat purchase rate, time to conversion)
Demographic Segmentation: Analyzes audience clusters by psychographics, device preference, engagement patterns, and sentiment
Predictive Trend Scoring: Assigns forward-looking scores to campaigns using historical patterns and seasonal variations
Real-Time Collaboration: Supports team annotations, customizable alerts, and multilingual reporting
Unified Data Pipeline: Normalizes data from social APIs, CRM systems, and e-commerce platforms into a single analytical model

Installation

Prerequisites

Power BI Desktop (latest version recommended)
Windows 10/11 or macOS (via browser for Power BI Service)
Minimum 8 GB RAM (16 GB+ for large datasets)
500 MB storage for application files
Active internet connection for data source APIs

Setup Steps

Clone the Repository

git clone https://github.com/thanhtu150/powerbi-campaign-roi-analyzer.git
cd powerbi-campaign-roi-analyzer

Open the Power BI File

# On Windows
start "campaign-roi-analyzer.pbix"

# On macOS (Power BI Service)
open https://app.powerbi.com
# Then upload the .pbix file

Configure Data Sources

Power BI will prompt for data source credentials. Configure each connector:

// Example: Facebook/Instagram connector configuration
let
    Source = Facebook.Graph("https://graph.facebook.com/v18.0/me/insights"),
    AccessToken = Environment.GetEnvironmentVariable("FACEBOOK_ACCESS_TOKEN"),
    Headers = [#"Authorization" = "Bearer " & AccessToken],
    Data = Json.Document(Web.Contents(Source, [Headers=Headers]))
in
    Data

Set Environment Variables

Configure API credentials via environment variables (never hardcode):

# Windows PowerShell
$env:FACEBOOK_ACCESS_TOKEN="your_token_here"
$env:LINKEDIN_CLIENT_ID="your_client_id"
$env:LINKEDIN_CLIENT_SECRET="your_client_secret"
$env:TIKTOK_API_KEY="your_api_key"
$env:GOOGLE_ANALYTICS_PROPERTY_ID="your_property_id"
$env:SHOPIFY_API_TOKEN="your_shopify_token"

# Linux/macOS
export FACEBOOK_ACCESS_TOKEN="your_token_here"
export LINKEDIN_CLIENT_ID="your_client_id"
export LINKEDIN_CLIENT_SECRET="your_client_secret"
export TIKTOK_API_KEY="your_api_key"
export GOOGLE_ANALYTICS_PROPERTY_ID="your_property_id"
export SHOPIFY_API_TOKEN="your_shopify_token"

Data Model Structure

The dashboard uses a star schema with the following key tables:

Fact Tables

FactCampaigns: Campaign performance metrics (impressions, clicks, conversions, spend)
FactSales: Product sales transactions attributed to campaigns
FactEngagement: Detailed engagement events (likes, comments, shares, saves)

Dimension Tables

DimDate: Time dimension with fiscal calendars and holidays
DimCampaign: Campaign metadata (name, platform, type, budget)
DimProduct: Product catalog with categories and pricing
DimAudience: Demographic and psychographic segments
DimChannel: Social media platform definitions

Key DAX Measures

1. Campaign ROI Calculation

Campaign ROI = 
VAR TotalRevenue = SUM(FactSales[Revenue])
VAR TotalSpend = SUM(FactCampaigns[AdSpend])
VAR ROI = DIVIDE(TotalRevenue - TotalSpend, TotalSpend, 0)
RETURN
    ROI

2. Weighted Engagement Rate

Weighted Engagement Rate = 
VAR Likes = SUM(FactEngagement[Likes]) * 1
VAR Comments = SUM(FactEngagement[Comments]) * 2
VAR Shares = SUM(FactEngagement[Shares]) * 3
VAR Saves = SUM(FactEngagement[Saves]) * 2.5
VAR TotalImpressions = SUM(FactCampaigns[Impressions])
VAR WeightedEngagements = Likes + Comments + Shares + Saves
RETURN
    DIVIDE(WeightedEngagements, TotalImpressions, 0)

3. Cost Per Acquisition (CPA)

CPA = 
VAR TotalSpend = SUM(FactCampaigns[AdSpend])
VAR TotalConversions = COUNTROWS(FILTER(FactSales, FactSales[IsFirstPurchase] = TRUE))
RETURN
    DIVIDE(TotalSpend, TotalConversions, 0)

4. Channel Attribution Score

Channel Attribution Score = 
VAR ChannelRevenue = SUM(FactSales[Revenue])
VAR ChannelSpend = SUM(FactCampaigns[AdSpend])
VAR AvgOrderValue = AVERAGE(FactSales[OrderValue])
VAR RepeatPurchaseRate = 
    DIVIDE(
        COUNTROWS(FILTER(FactSales, FactSales[IsRepeatPurchase] = TRUE)),
        COUNTROWS(FactSales),
        0
    )
VAR Score = 
    (DIVIDE(ChannelRevenue, ChannelSpend, 0) * 0.4) + 
    (AvgOrderValue / 100 * 0.3) + 
    (RepeatPurchaseRate * 100 * 0.3)
RETURN
    Score

5. Predictive Trend Score

Trend Score = 
VAR CurrentPeriod = CALCULATE(SUM(FactEngagement[TotalEngagements]), DATESINPERIOD(DimDate[Date], MAX(DimDate[Date]), -7, DAY))
VAR PreviousPeriod = CALCULATE(SUM(FactEngagement[TotalEngagements]), DATESINPERIOD(DimDate[Date], MAX(DimDate[Date]), -14, DAY))
VAR TrendSlope = DIVIDE(CurrentPeriod - PreviousPeriod, PreviousPeriod, 0)
VAR SeasonalFactor = 
    SWITCH(
        TRUE(),
        MONTH(MAX(DimDate[Date])) IN {11, 12}, 1.2,  // Holiday season
        MONTH(MAX(DimDate[Date])) IN {1, 2}, 0.8,    // Post-holiday slump
        MONTH(MAX(DimDate[Date])) IN {7, 8}, 0.9,    // Summer slowdown
        1.0
    )
VAR AdjustedScore = TrendSlope * SeasonalFactor * 100
RETURN
    AdjustedScore

Power Query Data Transformation

Normalize Social Media Data

let
    // Import raw data from multiple sources
    FacebookData = Facebook.Graph(Environment.GetEnvironmentVariable("FACEBOOK_PAGE_ID")),
    LinkedInData = LinkedIn.Campaigns(Environment.GetEnvironmentVariable("LINKEDIN_ACCOUNT_ID")),
    TikTokData = Json.Document(Web.Contents("https://open-api.tiktok.com/v1.3/", [Headers=[#"Access-Token"=Environment.GetEnvironmentVariable("TIKTOK_API_KEY")]])),
    
    // Standardize Facebook data
    NormalizedFacebook = Table.TransformColumns(FacebookData, {
        {"impressions", each _, Int64.Type},
        {"reach", each _, Int64.Type},
        {"clicks", each _, Int64.Type},
        {"spend", each _, Currency.Type},
        {"date_start", each Date.From(_), type date},
        {"campaign_name", each Text.Proper(_), type text}
    }),
    AddedPlatformFacebook = Table.AddColumn(NormalizedFacebook, "Platform", each "Facebook", type text),
    
    // Standardize LinkedIn data
    NormalizedLinkedIn = Table.TransformColumns(LinkedInData, {
        {"impressions", each _, Int64.Type},
        {"clicks", each _, Int64.Type},
        {"costInLocalCurrency", each _, Currency.Type},
        {"startDate", each Date.From(_), type date},
        {"name", each Text.Proper(_), type text}
    }),
    RenamedLinkedIn = Table.RenameColumns(NormalizedLinkedIn, {
        {"costInLocalCurrency", "spend"},
        {"startDate", "date_start"},
        {"name", "campaign_name"}
    }),
    AddedPlatformLinkedIn = Table.AddColumn(RenamedLinkedIn, "Platform", each "LinkedIn", type text),
    
    // Combine all sources
    CombinedData = Table.Combine({AddedPlatformFacebook, AddedPlatformLinkedIn}),
    
    // Add derived metrics
    AddEngagementRate = Table.AddColumn(CombinedData, "EngagementRate", each 
        if [impressions] > 0 then ([clicks] / [impressions]) else 0, type number),
    AddCPC = Table.AddColumn(AddEngagementRate, "CPC", each 
        if [clicks] > 0 then ([spend] / [clicks]) else 0, Currency.Type),
    
    // Sort by date descending
    SortedData = Table.Sort(AddCPC, {{"date_start", Order.Descending}})
in
    SortedData

Calculate Demographic Segments

let
    Source = FactEngagement,
    
    // Group by user demographics
    GroupedByDemo = Table.Group(Source, {"UserID", "AgeGroup", "Gender", "Location"}, {
        {"TotalEngagements", each List.Sum([Engagements]), type number},
        {"AvgSessionDuration", each List.Average([SessionDuration]), type number},
        {"PreferredContentType", each List.Mode([ContentType]), type text}
    }),
    
    // Calculate engagement intensity
    AddIntensityScore = Table.AddColumn(GroupedByDemo, "IntensityScore", each 
        ([TotalEngagements] * 0.6) + ([AvgSessionDuration] / 60 * 0.4), type number),
    
    // Assign psychographic cluster
    AddCluster = Table.AddColumn(AddIntensityScore, "PsychographicCluster", each 
        if [IntensityScore] >= 75 then "Power Users"
        else if [IntensityScore] >= 50 then "Active Engagers"
        else if [IntensityScore] >= 25 then "Casual Browsers"
        else "Passive Viewers", type text)
in
    AddCluster

Common Dashboard Configurations

1. Set Custom Alert Thresholds

Navigate to the dashboard settings and configure alerts using DAX:

Engagement Alert = 
VAR CurrentRate = [Weighted Engagement Rate]
VAR Threshold = 0.02  // 2% threshold
VAR IsUnderperforming = CurrentRate < Threshold
RETURN
    IF(IsUnderperforming, 
        "⚠️ Engagement below 2%", 
        "✓ Performance normal")

2. Configure Time Granularity

// Dynamic time grouping based on date range
Time Grouping = 
VAR DateRange = DATEDIFF(MIN(DimDate[Date]), MAX(DimDate[Date]), DAY)
VAR Grouping = 
    SWITCH(
        TRUE(),
        DateRange <= 7, "Hour",
        DateRange <= 31, "Day",
        DateRange <= 90, "Week",
        "Month"
    )
RETURN
    Grouping

3. Set Metric Weighting

// Custom weighted performance score
Custom Performance Score = 
VAR ReachWeight = 0.15
VAR EngagementWeight = 0.30
VAR ConversionWeight = 0.40
VAR RetentionWeight = 0.15

VAR NormalizedReach = DIVIDE([Total Reach], MAX(FactCampaigns[Reach]), 0)
VAR NormalizedEngagement = DIVIDE([Weighted Engagement Rate], 0.10, 0)  // Assume 10% is max
VAR NormalizedConversion = DIVIDE([Conversion Rate], 0.05, 0)  // Assume 5% is max
VAR NormalizedRetention = DIVIDE([Retention Rate], 0.80, 0)  // Assume 80% is max

VAR Score = 
    (NormalizedReach * ReachWeight) +
    (NormalizedEngagement * EngagementWeight) +
    (NormalizedConversion * ConversionWeight) +
    (NormalizedRetention * RetentionWeight)

RETURN
    Score * 100  // Convert to 0-100 scale

Real-World Usage Patterns

Pattern 1: Product Launch Analysis

// Compare product performance across channels
Product Channel Performance = 
SUMMARIZE(
    FactSales,
    DimProduct[ProductName],
    DimChannel[ChannelName],
    "Total Revenue", SUM(FactSales[Revenue]),
    "Units Sold", SUM(FactSales[Quantity]),
    "Avg Order Value", AVERAGE(FactSales[OrderValue]),
    "Conversion Rate", DIVIDE(
        COUNTROWS(FactSales),
        CALCULATE(SUM(FactCampaigns[Clicks]), USERELATIONSHIP(FactCampaigns[CampaignID], FactSales[CampaignID])),
        0
    )
)

Pattern 2: Budget Optimization

// Calculate optimal budget allocation based on ROI
Optimal Budget Allocation = 
VAR TotalBudget = SUM(FactCampaigns[Budget])
VAR ChannelROI = [Campaign ROI]
VAR TotalROI = CALCULATE([Campaign ROI], ALL(DimChannel))

VAR OptimalAllocation = 
    DIVIDE(ChannelROI, TotalROI, 0) * TotalBudget

RETURN
    OptimalAllocation

Pattern 3: Audience Segment Targeting

// Identify high-value audience segments
High Value Segments = 
CALCULATETABLE(
    SUMMARIZE(
        DimAudience,
        DimAudience[SegmentName],
        "Avg LTV", AVERAGE(FactSales[CustomerLifetimeValue]),
        "Purchase Frequency", COUNTROWS(FactSales) / DISTINCTCOUNT(FactSales[CustomerID]),
        "Engagement Score", [Weighted Engagement Rate]
    ),
    FILTER(
        DimAudience,
        [Avg LTV] > 500 && [Engagement Score] > 0.03
    )
)

Pattern 4: Campaign Fatigue Detection

// Detect campaigns showing declining performance
Campaign Fatigue Score = 
VAR DaysActive = DATEDIFF(MIN(FactCampaigns[StartDate]), MAX(DimDate[Date]), DAY)
VAR InitialEngagement = CALCULATE([Weighted Engagement Rate], DATESINPERIOD(DimDate[Date], MIN(FactCampaigns[StartDate]), 7, DAY))
VAR CurrentEngagement = CALCULATE([Weighted Engagement Rate], DATESINPERIOD(DimDate[Date], MAX(DimDate[Date]), -7, DAY))
VAR PerformanceDecline = DIVIDE(InitialEngagement - CurrentEngagement, InitialEngagement, 0)

VAR FatigueScore = 
    IF(
        DaysActive > 14 && PerformanceDecline > 0.20,
        "High Fatigue",
        IF(
            DaysActive > 7 && PerformanceDecline > 0.10,
            "Moderate Fatigue",
            "Normal"
        )
    )

RETURN
    FatigueScore

Data Refresh Configuration

Scheduled Refresh via Power BI Service

Publish the dashboard to Power BI Service
Configure data source credentials in the Power BI Service settings
Set refresh schedule:

{
  "refreshSchedule": {
    "days": ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"],
    "times": ["06:00", "12:00", "18:00"],
    "localTimeZoneId": "UTC",
    "enabled": true
  },
  "notifyOption": "MailOnFailure"
}

Manual Refresh via Power BI Desktop

// Add refresh timestamp to track data freshness
let
    Source = #table(
        {"RefreshTimestamp", "Status"},
        {{DateTime.LocalNow(), "Success"}}
    )
in
    Source

Troubleshooting

Issue: API Rate Limit Exceeded

Solution: Implement exponential backoff in Power Query

let
    FetchWithRetry = (url as text, maxRetries as number) as any =>
        let
            TryFetch = (attempt as number) as any =>
                try
                    Json.Document(Web.Contents(url))
                otherwise
                    if attempt < maxRetries then
                        let
                            WaitTime = Number.Power(2, attempt) * 1000,
                            Wait = Function.InvokeAfter(() => TryFetch(attempt + 1), #duration(0, 0, 0, WaitTime / 1000))
                        in
                            Wait
                    else
                        error "Max retries exceeded"
        in
            TryFetch(0)
in
    FetchWithRetry

Issue: Data Source Authentication Failure

Solution: Refresh OAuth tokens programmatically

let
    RefreshToken = Environment.GetEnvironmentVariable("OAUTH_REFRESH_TOKEN"),
    ClientId = Environment.GetEnvironmentVariable("OAUTH_CLIENT_ID"),
    ClientSecret = Environment.GetEnvironmentVariable("OAUTH_CLIENT_SECRET"),
    
    TokenEndpoint = "https://oauth.platform.com/token",
    
    Response = Json.Document(Web.Contents(TokenEndpoint, [
        Content = Text.ToBinary(Uri.BuildQueryString([
            grant_type = "refresh_token",
            refresh_token = RefreshToken,
            client_id = ClientId,
            client_secret = ClientSecret
        ])),
        Headers = [#"Content-Type" = "application/x-www-form-urlencoded"]
    ])),
    
    NewAccessToken = Response[access_token]
in
    NewAccessToken

Issue: Metric Calculation Discrepancies

Solution: Validate data quality with diagnostic queries

// Data quality check
Data Quality Report = 
SUMMARIZE(
    FactCampaigns,
    DimChannel[ChannelName],
    "Total Records", COUNTROWS(FactCampaigns),
    "Missing Impressions", CALCULATE(COUNTROWS(FactCampaigns), ISBLANK(FactCampaigns[Impressions])),
    "Negative Spend", CALCULATE(COUNTROWS(FactCampaigns), FactCampaigns[AdSpend] < 0),
    "Future Dates", CALCULATE(COUNTROWS(FactCampaigns), FactCampaigns[Date] > TODAY()),
    "Duplicate CampaignIDs", COUNTROWS(FactCampaigns) - DISTINCTCOUNT(FactCampaigns[CampaignID])
)

Issue: Slow Dashboard Performance

Solution: Optimize measures with variables and reduce cardinality

// Optimized measure using variables
Optimized ROI = 
VAR SalesTable = SUMMARIZE(FactSales, FactSales[CampaignID], "Revenue", SUM(FactSales[Revenue]))
VAR SpendTable = SUMMARIZE(FactCampaigns, FactCampaigns[CampaignID], "Spend", SUM(FactCampaigns[AdSpend]))
VAR JoinedTable = NATURALLEFTOUTERJOIN(SpendTable, SalesTable)
VAR TotalRevenue = SUMX(JoinedTable, [Revenue])
VAR TotalSpend = SUMX(JoinedTable, [Spend])
RETURN
    DIVIDE(TotalRevenue - TotalSpend, TotalSpend, 0)

Export and Reporting

Export Data to CSV

// Export summarized campaign data
let
    Source = FactCampaigns,
    Grouped = Table.Group(Source, {"CampaignID", "CampaignName", "Platform"}, {
        {"Total Impressions", each List.Sum([Impressions]), type number},
        {"Total Clicks", each List.Sum([Clicks]), type number},
        {"Total Spend", each List.Sum([AdSpend]), Currency.Type},
        {"ROI", each List.Sum([Revenue]) / List.Sum([AdSpend]) - 1, type number}
    }),
    Export = Csv.Document(Table.ToColumns(Grouped))
in
    Export

Generate Automated Email Reports

Use Power Automate integration to send scheduled reports:

{
  "trigger": {
    "type": "Recurrence",
    "recurrence": {
      "frequency": "Week",
      "interval": 1,
      "schedule": {
        "weekDays": ["Monday"],
        "hours": [8]
      }
    }
  },
  "action": {
    "type": "ExportPowerBIReport",
    "inputs": {
      "workspaceId": "ENV:POWERBI_WORKSPACE_ID",
      "reportId": "ENV:POWERBI_REPORT_ID",
      "format": "PDF"
    }
  }
}

Best Practices

Never Hardcode Credentials: Always use environment variables or Azure Key Vault
Implement Incremental Refresh: For large datasets, configure incremental refresh to load only new/changed data
Use Aggregations: Pre-aggregate data at appropriate granularity to improve query performance
Document Measure Logic: Add descriptions to all custom measures for team collaboration
Version Control: Store .pbix files in Git with clear commit messages for change tracking
Test Data Quality: Build automated data validation checks before visualization
Optimize Relationships: Use bidirectional filters sparingly and prefer one-way relationships
Monitor Performance: Use Performance Analyzer in Power BI Desktop to identify bottlenecks

Additional Resources

Sample Data Generator: Create synthetic campaign data for testing
Custom Visuals: Install additional visuals from AppSource for advanced charts
DAX Studio: Use for advanced query profiling and optimization
Power BI REST API: Automate report distribution and embedding

This skill provides comprehensive guidance for working with the Campaign ROI Analyzer dashboard, from initial setup through advanced analytics patterns and troubleshooting.