The customer journey spans dozens of touchpoints across days or weeks. Privacy changes have decimated user-level tracking. Walled gardens report metrics optimized for their own interests.
Yet somehow, you still need to know which advertising actually drives business results.
Enter the measurement trinity: Multi-Touch Attribution (MTA), Marketing Mix Modeling (MMM), and Incrementality Testing. Each has strengths and limitations. Together—powered by AI—they provide the closest thing to truth available in modern advertising.
Almost half (46.9%) of US marketers plan to invest more in MMM over the next year. This isn't a trend—it's a fundamental shift in how sophisticated advertisers measure effectiveness.
The Measurement Problem
Why Attribution Fails
Traditional attribution—especially last-click—made sense when customer journeys were simpler. Someone clicked an ad, visited a site, purchased. Credit the click.
Today's reality is different:
- Consumers interact with 6+ touchpoints before converting
- Cross-device behavior fragments the journey
- Privacy controls break tracking chains
- Upper-funnel exposure influences lower-funnel conversion
- Many touchpoints are impression-based, not click-based
Last-click attribution ignores all the touchpoints that primed the purchase. It over-credits lower-funnel channels (search, retargeting) and under-credits upper-funnel channels (display, video, social).
The Platform Reporting Problem
Each advertising platform reports its own version of truth:
- Google says the conversion came from Google
- Meta says the same conversion came from Meta
- Both take full credit
This isn't fraud—it's how attribution works. But you can't add their numbers together without double-counting.
The Privacy Complication
iOS 14.5+ decimated app tracking. Cookie deprecation reduced cross-site visibility. GDPR and state privacy laws restrict data collection.
This creates opportunity for measurement approaches that don't require user-level tracking: aggregated modeling and controlled experiments.
The Three Measurement Approaches
Multi-Touch Attribution (MTA)
What it does: Assigns fractional credit to each touchpoint in the conversion path based on statistical analysis.
Strengths:
- Granular, campaign-level insights
- Fast feedback loops
- Identifies specific tactics that work
Limitations:
- Requires user-level tracking (increasingly difficult)
- Struggles with cross-device and offline touchpoints
- Can't measure impression-only exposure
Best for: Digital-heavy campaigns where user journeys are trackable, tactical optimization within channels.
Marketing Mix Modeling (MMM)
What it does: Uses statistical regression to measure how each marketing channel contributes to business outcomes over time.
Strengths:
- Privacy-safe (no user-level tracking required)
- Measures offline and impression-based channels
- Captures long-term and halo effects
- Provides holistic cross-channel view
Limitations:
- Requires significant historical data
- Slower feedback (weeks/months, not days)
- Model quality depends on data quality and expertise
Best for: Strategic budget allocation across channels, understanding long-term effects.
Incrementality Testing
What it does: Uses controlled experiments to prove whether advertising caused conversions that wouldn't have happened otherwise.
Strengths:
- Proves causation, not just correlation
- No attribution assumptions
- Works across all channels
- Privacy-safe
Limitations:
- Requires sufficient scale for statistical validity
- Loses revenue during holdout periods
- Point-in-time results may not persist
Best for: Validating channel value, calibrating MMM and MTA models.
How AI Transforms Measurement
Faster MMM
Traditional MMM required months of data collection and analysis by specialized statisticians. AI-powered MMM delivers:
- Rapid model updates. On-demand model refresh—re-run when inputs change rather than waiting for quarterly updates.
- Automated calibration. Incrementality tests automatically calibrate MMM outputs.
- Scenario planning. Real-time scenario testing—what happens if we shift budget from X to Y?
- Accessible insights. Natural language interfaces let marketers query measurement data without statistical expertise.
Automated Incrementality Testing
- Experiment design. AI automates the creation of test-and-control experiments.
- Geo-testing optimization. AI identifies optimal geographic test cells and control regions.
- Statistical analysis. Automated significance testing and confidence intervals.
- Continuous experimentation. Always-on incrementality measurement rather than periodic tests.
Unified Measurement Platforms
Modern measurement platforms combine approaches:
- Measured integrates incrementality testing with Causal MMM, automatically calibrating models with experimental data.
- TransUnion combines identity resolution with MMM, MTA, and incrementality.
- Google's Meridian provides open-source MMM that advertisers can run themselves.
Implementing a Triangulated Measurement Strategy
The Triangulation Principle
No single measurement approach tells the complete truth. The solution: triangulation—using multiple methodologies to validate findings.
When MTA, MMM, and incrementality testing all point the same direction, you have confidence. When they diverge, you've identified something worth investigating.
Implementation Framework
Phase 1: Foundation
- Establish data infrastructure. Clean, consistent data across channels is prerequisite.
- Document current state. What do platform reports show? What's your current allocation?
- Choose your tools based on budget and scale.
Phase 2: MMM Implementation
- Aggregate historical data. Typically 2+ years of marketing spend, sales/conversions, and external factors.
- Build initial model. Work with vendors or internal teams.
- Identify optimization opportunities.
Phase 3: Incrementality Validation
- Prioritize tests for channels with the most uncertainty or largest spend.
- Design experiments. Geo-tests, holdout groups, or platform-specific features.
- Use results to calibrate MMM.
Phase 4: Ongoing Optimization
- Continuous experimentation on key channels.
- Regular model updates as conditions change.
- Unified decision-making using triangulated insights.
What to Measure
- Incremental ROAS (iROAS). The return from advertising that wouldn't have happened without it.
- Channel contribution. What percentage of total conversions does each channel actually drive?
- Halo effects. How do upper-funnel channels influence lower-funnel performance?
- Diminishing returns. At what point does additional spend yield decreasing incremental returns?
Platform-Specific Insights
Meta
Research consistently shows Meta is underreported by last-click attribution. Upper-funnel exposure on Facebook and Instagram primes conversions that occur through other channels.
TikTok
Kochava research found MMM revealed TikTok campaigns generate 35% higher incremental impact compared to last-touch attribution. TikTok initiates journeys rather than ending them.
Google's position as last-click capture channel means it's often over-credited by attribution. MMM helps right-size Google's contribution.
CTV/Streaming
Connected TV is impression-based with no click path. Only MMM and incrementality testing can properly measure CTV's contribution.
Best Practices
- Combine approaches. Use MMM for strategic allocation, MTA for tactical optimization, incrementality for validation.
- Validate platform reporting. Never allocate budget based solely on what platforms tell you.
- Account for long-term effects. MMM captures brand-building and delayed conversions that attribution misses.
- Build feedback loops. Incrementality results should calibrate your models.
- Accept uncertainty. Perfect measurement doesn't exist. Use multiple methodologies to triangulate toward truth.
Common Mistakes
- Over-relying on platform metrics. Every platform reports favorably about itself.
- Last-click thinking. Cutting upper-funnel channels because they don't show conversions often hurts overall performance.
- Ignoring incrementality. High ROAS might just mean you're capturing sales that would have happened anyway.
- Insufficient data for MMM. Rushing implementation with insufficient data produces unreliable models.
- One-and-done testing. Incrementality changes over time. Continuous testing beats periodic validation.
The ROI of Proper Measurement
The stakes are significant:
- Efficiency gains. Customers increase marketing efficiency up to 30%+ through incrementality-based optimization.
- iROAS variance. Incrementality experiments show iROAS ranging from 253% to 1,609% across advertisers.
- Budget reallocation. Proper measurement typically shows some channels are over-invested and others under-invested.
- Trust gap. Only 6% of advertisers fully trust retailers' reported media metrics.
The Bottom Line
Attribution as we knew it is dead. User-level tracking across the full journey is increasingly impossible, and platform reporting is inherently biased.
The future of measurement is triangulation:
- MMM for strategic, holistic channel contribution
- Incrementality testing for causal proof of advertising value
- MTA for tactical, within-channel optimization where tracking remains possible
AI makes this practical at scale—faster modeling, automated experimentation, and accessible insights without requiring teams of statisticians.
The advertisers winning in 2025 aren't optimizing based on platform dashboards. They're proving which conversions wouldn't have happened without their advertising—and allocating budget accordingly.
That's incrementality. That's the new standard for measurement.
