Marketing Analytics with Fexingo: Data, Attribution, and Measuring Campaign Performance

Episode 32 of Marketing Analytics with Fexingo dives into a foundational flaw in many attribution models: the lack of a true control group. Lucas and Luna examine how Unilever's 2023 ice cream campaign in Brazil used a geo-based holdout to prove that their digital spend drove only 12% incremental lift, not the 35% their last-click model claimed. They walk through the mechanics of matched-market ex...

Most marketing teams still judge ad performance by last-click attribution — and they're overpaying by billions. In this episode, Lucas and Luna break down why incrementality testing is the only reliable way to measure true ad lift, using a real case from a major retailer that ran a geo-level experiment and discovered 40% of its digital spend was wasted. They explain the difference between 'was thi...

Lucas and Luna dig into a 2026 meta-analysis of 147 brand lift studies that found creative quality accounts for nearly 60 percent of ad effectiveness — double the contribution of targeting precision. They walk through how Unilever ran a controlled experiment pitting two different creatives against identical audience segments and got a 3X lift difference, then discuss why most marketing analytics t...

Lucas and Luna explore how marketing analytics teams routinely confuse correlation with causation—and why it costs millions in wasted ad spend. They unpack a 2025 experiment from a mid-size e-commerce brand that ran a geo-lift test on its highest-performing Facebook ads. The ads looked brilliant in the attribution dashboard: 12x return on ad spend. But the geo test revealed the campaign drove zero...

Episode 28 of Marketing Analytics with Fexingo. Lucas and Luna dive into a common headache: what happens when your multi-touch attribution model and your marketing mix model give you completely contradictory answers about which channel drove a sale. They walk through a real case—a mid-size e-commerce brand that saw a 30% gap between MTA and MMM on paid search—and discuss why this isn't a bug but a...

Episode 27 of Marketing Analytics with Fexingo. Lucas and Luna dive into survival analysis — a technique borrowed from medical research — and why it outperforms traditional lead scoring for B2B sales cycles. They break down a real example: a SaaS company that used Kaplan-Meier curves to discover that leads who attended a demo within 7 days had a 34% higher conversion probability than those who wai...

In this episode, Lucas and Luna explore why cross-validation is a missing ingredient in many marketing attribution and media mix models. They walk through a concrete example: a DTC brand that trusted its default model's ROAS estimates, only to discover the model overfit to a noisy holiday period. When the brand ran a simple k-fold cross-validation, it found the model's predictions were 30% less re...

Episode 25 of Marketing Analytics with Fexingo dives into incrementality testing — the gold standard for measuring whether an ad actually causes a sale or just captures one that would have happened anyway. Lucas and Luna walk through a real 2025 case from a DTC apparel brand that ran a geo-based incrementality test across Facebook, Google, and TV. They explain why last-click and even multi-touch m...

In this episode, Lucas and Luna dive into the practical use of holdout groups to validate multi-touch attribution models. Using a real-world case from a mid-market e-commerce brand that ran a six-month geo holdout test, they explore how withholding a portion of the market from advertising reveals true incrementality. Lucas explains why most attribution models overstate performance by 30-50% and ho...

In this episode, Lucas and Luna dive into the difference between incrementality testing and correlation-based attribution. Lucas explains how the rise of privacy changes — like Apple's App Tracking Transparency and Google's deprecation of third-party cookies — has made traditional last-click and multi-touch models unreliable. He shares a concrete example from a DTC brand that ran a geo-based incre...

Lucas and Luna explore the hidden interactions between marketing channels that standard attribution models miss. Using a real example from a direct-to-consumer brand that ran TV and paid search simultaneously, they show how media mix models can reveal that search was stealing credit from TV, and how adjusting the budget allocation boosted overall ROAS by 18%. They discuss why simple additive model...

Episode 21 of Marketing Analytics with Fexingo digs into why ignoring last year’s data can break your media mix model. Lucas and Luna explore a 2024 case from a mid-sized DTC brand—let’s call it FreshStep—that ran a full MMM without a 2023 baseline. The result? It over-credited social video by 22% and under-credited search by 15%, leading to a $340,000 misallocation in Q1 2025 alone. They walk thr...

In this episode, Lucas and Luna challenge the obsession with direct response attribution for podcast advertising. They examine a case study from a DTC mattress company that ran 12 weeks of podcast ads and found zero last-click conversions but a 14% lift in branded search and a 9-point increase in purchase intent according to a brand lift study. They discuss why podcast ads operate more like tradit...

Episode 19 of Marketing Analytics with Fexingo. Lucas and Luna break down how brand lift studies work, why they're the gold standard for measuring ad effectiveness, and how companies like Netflix use them to prove causation, not just correlation. They walk through a real 2025 case where a CPG brand ran a geo-based brand lift test on a new snack launch, revealing a 12% lift in purchase intent that...

Episode 18 of Marketing Analytics with Fexingo dives into the critical but often overlooked step of calibrating media mix models against real-world experiments. Lucas and Luna explore why even sophisticated MMMs can produce misleading results without calibration, using the example of a DTC brand that ran a geo lift test only to discover their model had over-allocated 25% of budget to paid search....

Lucas and Luna dig into geo lift testing — the overlooked sibling of incrementality measurement. Using a real 2024 case from a national quick-service chain, they explain why splitting test and control by geography reveals true ad effectiveness that user-level attribution misses. You'll learn how geo experiments solve the 'leaky bucket' problem of digital attribution, why Facebook and Google can't...

Episode 16 of Marketing Analytics with Fexingo digs into Amazon's unique approach to marketing mix modeling. While most brands treat MMM as a top-down budget allocation tool applied annually, Amazon runs it as a continuous, channel-level optimization loop — updating models weekly and feeding results directly into programmatic bidding algorithms. Lucas and Luna break down the structural reasons beh...

Episode 15 of Marketing Analytics with Fexingo tackles the problem of sparse data in marketing attribution. Lucas and Luna explore how a Bayesian prior — starting with a baseline assumption — can stabilize models when conversion events are rare. They walk through a concrete example: a B2B SaaS company with a long sales cycle and only a handful of closed-won deals per quarter. The hosts explain how...

Episode 14 of Marketing Analytics with Fexingo digs into a blind spot most attribution models miss: interaction effects between marketing channels. Lucas and Luna explore a real-world example where a brand's paid search ads only worked because of a podcast campaign running at the same time — and how a marketing mix model with interaction terms caught what last-click and multi-touch attribution com...

Lucas and Luna explore why traditional attribution models—even multi-touch—fail to answer the most critical question in marketing analytics: what would have happened if you hadn't run that campaign at all? They dive into the concept of counterfactual reasoning, using concrete examples from e-commerce and B2B SaaS. Lucas explains how companies like Amazon and Booking.com use holdout groups and synt...

Lucas and Luna explain why holdout groups—control groups that receive zero ad exposure—are the missing piece in most marketing attribution setups. They walk through a real example from a DTC skincare brand that ran a six-week geo-holdout test and discovered its Facebook campaigns were actually cannibalizing organic sales. The episode covers how holdout groups differ from incrementality testing, th...

Episode 11 of Marketing Analytics with Fexingo: Lucas and Luna explore the hidden cost of optimizing every campaign for short-term attribution metrics. Using the cautionary tale of a major European retailer that slashed brand awareness spend to hit ROAS targets — only to see same-store sales decline 8% within two quarters — they unpack why over-attribution can hollow out a brand. They discuss the...

Episode 10 of Marketing Analytics with Fexingo dives into one of the most frustrating gaps in modern marketing: the mismatch between ad-server reported conversions and CRM-attributed sales. Lucas and Luna walk through a real case from a mid-market e-commerce brand that discovered its Facebook Ads Manager showed 30 percent more conversions than its Salesforce pipeline. They explain how cookie depre...

This episode dives into a problem most marketers overlook: the attribution window. Lucas and Luna explain why the default 7-day click and 28-day view windows in platforms like Google Analytics and Facebook are often misleading. Using a real example from a DTC skincare brand, they show how extending the window to 90 days changed their understanding of what drove sales—and saved them from killing th...

Lucas and Luna dig into why traditional marketing mix models (MMM) often fail to guide real budget decisions. Lucas explains how Bayesian statistics update prior beliefs with new data, producing probability distributions instead of point estimates. He walks through a specific example: a CPG brand that used Bayesian MMM to reallocate $15 million from TV to digital and saw 18% higher incremental rev...