Interpreting Model Outputs & ROI Analysis

Understanding and analyzing Cassandra’s model outputs ensures you make data-driven marketing decisions. This guide walks you through key insights, interpreting ROI, and validating attribution results.

Step 1: Understanding Model Outputs

1. Channel Contribution Analysis
   • Cassandra assigns a percentage contribution to each marketing channel, showing its role in driving conversions or revenue.
   • Contribution values are adjusted based on historical performance and statistical significance.
2. Baseline & Seasonality Effects
   • Baseline: Represents revenue that would have occurred even without marketing spend.
   • Seasonality: Captures predictable variations in demand (e.g., holiday spikes, summer slowdowns).
3. Confidence Intervals & Uncertainty
   • Each channel’s ROI includes confidence intervals to indicate statistical certainty.
   • A wider confidence interval means greater uncertainty, requiring validation through incrementality tests.
4. Multicollinearity & Impact on Results
   • Highly correlated marketing activities can lead to overlapping attributions.
   • Cassandra identifies potential multicollinearity issues and suggests aggregating campaign types.

Step 2: Evaluating ROI & Efficiency Metrics

1. ROI (Return on Investment)
   • Formula: ROI = (Attributed Revenue / Marketing Spend)
   • Higher ROI values suggest stronger efficiency, but compare against industry benchmarks.
2. ROAS (Return on Ad Spend)
   • Formula: ROAS = Revenue from Ads / Ad Spend
   • Use ROAS to evaluate paid media performance across platforms.
3. CPA (Cost per Acquisition)
   • Formula: CPA = Total Spend / Conversions
   • Lower CPA values indicate greater cost efficiency in acquiring customers.
4. Diminishing Returns & Spend Saturation
   • Increasing budget on a saturated channel yields smaller incremental gains.
   • Cassandra’s saturation curves show the optimal spend level per channel.

Step 3: Validating Model Accuracy

1. Comparing Model Predictions vs. Actual Performance
   • Cassandra forecasts revenue based on previous spend levels—compare predictions to real business outcomes.
   • Use budget allocator validation to see if past recommendations aligned with results.
2. Running Incrementality Tests
   • GeoLift Experiments: Validate a channel’s effect by testing holdout regions.
   • Conversion Lift Tests: Compare audiences exposed vs. not exposed to ads within platforms.
3. Analyzing Business Error & Confidence Scores
   • Business error indicates model accuracy in attributing spend to revenue.
   • Confidence scores measure certainty in results—higher is better.

Step 4: Making Data-Driven Decisions

1. Reallocating Budget to High-Performing Channels
   • Shift budget from low-ROI channels to higher ROI channels.
   • Consider reducing spend in over-saturated channels.
2. Adjusting Strategy Based on ROI Trends
   • If brand awareness channels show low ROI, focus on multi-touch attribution instead of last-click ROI.
   • High uncertainty in performance? Run experiments to validate impact.
3. Tracking Over Time
   • Run monthly ROI reviews to ensure marketing efficiency remains high.
   • Use Cassandra’s success tracker to compare quarterly improvements in ROI.

Summary & Next Steps

• Understand channel contributions, seasonality, and baseline impacts.
• Analyze ROI, ROAS, and CPA to measure efficiency.
• Use confidence intervals & business error scores to assess accuracy.
• Validate attributions with incrementality tests.
• Reallocate budget based on high vs. low-performing channels.