Marginal returns, not averages, reveal where an extra dollar stops paying off. This matters because a single incremental ROAS number is still an average, missing the exact point where returns turn negative. Without pairing mix modeling with real experiments, that curve stays theoretical, and low-signal data can’t be fixed by tuning alone. This video traces measurement’s shift from barcodes and last-click to modern models built around marginal returns.
Why Marginal Returns, Not Averages, Are the Universal Truth of Marketing
Marginal returns, not averages, are the universal truth of marketing, because every dollar of ad spend eventually stops returning what the dollars before it did, and averages hide exactly where that happens. In this Curve Your Enthusiasm conversation, Jon Lorenzini and John Wallace host Dr. Dominique Hanssens, the UCLA professor widely known as the godfather of econometric models, for a wide-ranging look at the past, present, and future of marketing measurement.
Hanssens grounds the conversation in three sources of diminishing returns: audience saturation, the hard ceiling of market size for durable goods, and the influence of other marketing mix factors like distribution and pricing, all of which digital media and propensity scoring can accelerate further. He traces measurement’s own history, from the 1980s barcode revolution that first made real sales tracking possible, through the multi-touch attribution era, to today’s shift back toward marketing mix modeling blended with experimentation as third-party cookies disappear. Both speakers push hard against averages: even a well-run incremental ROAS number is still an average across every dollar spent, not a read on the next dollar specifically, and conversion lift tests, while a step forward, usually give the same kind of average rather than the full response curve a marketer actually needs. The conversation also covers why agencies need to be realigned around incremental rather than last-click metrics, the case for Bayesian methods when data is sparse, the well-documented link between brand health and long-term financial performance, and why Google Trends should never be mistaken for a genuine read on sales or brand health.
In this video, you’ll learn:
The fundamental economic concept of diminishing returns in advertising
Why average metrics often mislead marketing investment decisions
How to transition from last-click to marginal return measurement
The importance of supplementing models with controlled experiments
Strategies for aligning agency incentives with incremental business outcomes
TIME STAMPVideo Overview: Marginal Returns and the Universal Truths of Measurement
+–
[0:00 – 1:19] Introduction to the Series
Jon “JLo” Lorenzini, Chief Marketing Science Officer at LiftLab, introduces the series Curve Your Enthusiasm. The show features conversations with academics, practitioners, and strategists regarding the diminishing returns curves that all marketers face. Today’s episode covers the past, present, and future of measurement and marketing analytics, divided into three key topics: universal truths, the timeline of measurement, and evolving methodologies.
[1:19 – 2:38] Introducing the Guests
Jon introduces the speakers: John Wallace, co-founder of LiftLab and former founder of DataSong, and Dr. Dominique Hanssens, a Distinguished Research Professor of Marketing at UCLA, known as the “Godfather of econometric models.”
[2:38 – 7:47] Universal Marketing Truths
Dr. Hanssens explains foundational marketing concepts, specifically the diminishing returns phenomenon. He details three sources for this: saturation of the audience (repetition leads to fewer returns), the limit of market size for durable goods, and the impact of other marketing mix factors like distribution and pricing. John and Dr. Hanssens discuss how digital media and propensity scores can accelerate these diminishing returns.
[7:47 – 18:05] The Timeline of Measurement: Past to Present
Dr. Hanssens recounts the evolution of measurement, starting with the 1980s barcode revolution, which allowed for real sales tracking. This was followed by digital analytics and multi-touch attribution. With the loss of third-party cookies, the industry is shifting back toward sophisticated marketing mix modeling blended with experiments. John highlights the role of marketing science in creating new data through intentional testing rather than just observing.
[18:05 – 32:38] Misleading Metrics: Why Averages Are the Enemy
Dr. Hanssens and John argue that averages are misleading because they assume a linear response, which rarely exists due to diminishing returns. They explain that marketers should focus on the return on the last dollar (marginal returns) to find the point where spending maximizes profit. They emphasize that even though conversion lift tests are a step in the right direction, they often provide averages rather than the full response curve needed for precise optimization.
[32:38 – 45:28] Attributed vs. Incremental Metrics
The group discusses change management in getting agencies aligned with incremental measurement (iROAS) rather than last-click or platform-reported ROAS. They explain that last-click is often a reflection of how qualified an audience is rather than a proof of advertising effectiveness. They use the analogy of a shoe store: the person at the cash register didn’t do all the work to sell the shoes; the door did.
[45:28 – 48:47] Bayesian vs. Frequentist Methodologies
Dr. Hanssens defines the difference, noting that Bayesian methods are increasingly popular because modern computers can simulate complex outcomes. This allows analysts to incorporate prior beliefs or previous experimental results to refine models and reduce standard errors in situations where data may be missing or sparse.
[48:47 – 54:49] Long-Term vs. Short-Term Metrics
They discuss the relationship between brand health (long-term) and conversions (short-term). Dr. Hanssens cites studies showing that a 10% increase in brand health typically correlates to a 3% increase in long-term financial performance. John cautions that when calculating Customer Lifetime Value, marketers must not omit the cost of paid media required for customer retention.
[54:49 – 1:00:15] Seasonality, Google Trends, and Q&A
Dr. Hanssens explains that seasonality should be handled in models using indicator (dummy) variables to separate base fluctuations from marketing effects. Regarding Google Trends, the team notes it is a useful indicator of public interest but warns that it shouldn’t be confused with actual sales or brand health, as news events can cause misleading spikes.
Key Lessons: Building Measurement Around Marginal, Not Average, Returns
Diminishing returns: Marketing efficiency naturally declines as audiences saturate and spend increases, necessitating a marginal rather than average view of ROI.
Modern measurement: With the decline of third-party cookies, companies should pivot toward robust marketing mix models integrated with live experimentation.
Strategic optimization: Marketers must prioritize the return on the last dollar spent to identify the precise point where additional investment ceases to be profitable.
Data integrity: When existing data sets are fragmented or unreliable, organizations should proactively generate high-quality experimental data to inform their strategies.
