Marketing Mix Modelling: Measure What Actually Drives Revenue
Marketing mix modelling helps senior leaders move beyond platform-reported metrics to understand the true revenue contribution of each marketing channel, ena…
Topic: Sales Efficiency
Why It Matters
Marketing budgets are under scrutiny in most organisations. Senior leaders are asked not just to spend responsibly but to justify every allocation with evidence of return.
The challenge is that most of the metrics readily available (cost per lead, click-through rates, platform-reported conversions) are produced by the very channels being measured. They are, in other words, self-reported. That makes them useful for operational decisions but inadequate for strategic ones.
When budgets are modest and channels are few, this limitation is manageable. As organisations grow, channel complexity increases. Paid search, programmatic display, television, radio, sponsorship, email, social: each channel reports its own numbers, and those numbers rarely account for how channels interact, overlap, or cannibalise each other.
A manager relying solely on last-click attribution or individual platform metrics is not measuring what they think they are measuring.
A more rigorous approach uses historical data to isolate the contribution of each channel to actual revenue outcomes, while also separating what the business would have earned regardless of marketing activity (base sales) from what marketing specifically drove (incremental sales). This gives commercial leaders a clearer, more defensible basis for budget decisions.
To assess whether this type of analysis is relevant to your current situation, consider the following questions.
Are you managing marketing investment across three or more channels, including at least one that is difficult to track digitally?
Do you find it hard to justify channel budget decisions to your CFO or board without relying on platform-reported data?
Have you noticed that reducing spend in one channel seems to affect results in another, but you cannot quantify that relationship reliably?
If any of these feel familiar, the approach described in this article is likely worth exploring.
Origin and Underlying Logic
The statistical method now widely used to evaluate marketing channel performance has its roots in econometric modelling, a branch of economics focused on using statistical techniques to analyse economic relationships.
The approach was first applied to consumer goods marketing in the 1970s and 1980s, when large FMCG companies began building regression-based models to understand how advertising, pricing, distribution, and promotions interacted to drive sales volumes.
At its core, the method works by treating revenue (or another chosen business outcome) as the dependent variable, and marketing spend across each channel alongside non-marketing factors such as seasonality, pricing, macroeconomic conditions, and competitor activity as independent variables.
The model uses historical data (typically two to three years of weekly or monthly records) to estimate the relationship between each input and the outcome. The result is a set of coefficients that describe how much each variable has historically contributed to the business result.
What makes this approach powerful is not just that it produces channel-level contribution estimates. It is that it does so while controlling for factors outside marketing's control. Two campaigns that appear to have delivered similar returns in isolation may look very different once the model accounts for the fact that one ran during a seasonal peak and the other did not.
The framework sits within the broader field of revenue attribution and complements other measurement approaches such as digital attribution models and controlled experiments. Managers familiar with A/B testing or incrementality studies will find it occupies a different but complementary space in the measurement toolkit.
Structure and Application
The model is built from historical input data across several categories: marketing spend by channel and time period, a consistent measure of business outcome (typically revenue or unit sales), and contextual variables including pricing, distribution, seasonality indices, and relevant external factors. Each data stream is structured into a time-series dataset (usually at weekly granularity) and used to fit a regression model.
The output is typically a decomposition chart: a visual representation of how much of the business outcome is attributable to base performance (what the organisation would have achieved without any marketing), and how much is attributable to each marketing channel and contextual factor.
From this output, two key metrics are usually derived.
The return on investment for each channel, comparing the revenue contribution attributed to that channel against spend invested in it.
The marginal return curves for each channel, which show how the incremental return from each additional unit of spend changes as total investment in that channel increases or decreases. These curves are central to the budget optimisation use case, where the goal is to identify the spend allocation across channels that maximises total return.
The analysis is most commonly delivered as a periodic study, typically run annually or biannually, though organisations with more mature data infrastructure sometimes build more continuous modelling capability.
The primary format is a model output report or dashboard, often supplemented by scenario planning tools that allow budget holders to model the expected effect of reallocating spend before committing.
Before reading further, pull two years of weekly marketing spend data by channel and the corresponding revenue or sales volume for the same period into a single spreadsheet. Note immediately where the gaps are: missing weeks, channels with inconsistent reporting, or periods where spend data is held by an external agency rather than internally.
Those gaps are the first practical problem this type of analysis will surface, and understanding them upfront will shape how seriously you can rely on any model that follows.
FAQ
Q: What is marketing mix modelling?
A: Marketing mix modelling is a statistical technique that uses historical data to estimate how much each marketing channel and external factor has contributed to a business outcome, usually revenue or sales volume. It separates what the business would have earned regardless of marketing spend (base sales) from what marketing activity specifically generated (incremental sales), giving commercial leaders a more defensible basis for budget decisions than platform-reported metrics alone.
Q: How is marketing mix modelling different from digital attribution modelling?
A: Digital attribution models track individual user journeys through digital touchpoints, assigning credit to channels based on how users interacted with them before converting. This works well within digital channels but cannot account for offline media such as television, radio, or out-of-home advertising. Marketing mix modelling uses aggregate data rather than user-level data and can incorporate all channels, including offline. The two approaches answer related but different questions and are often used together.
Q: How much data do you need to run a marketing mix model?
A: Most practitioners recommend a minimum of two years of weekly data, which gives the model sufficient observations to identify patterns and control for seasonal effects. Data needs to include marketing spend by channel, a consistent measure of business outcome, and relevant contextual variables such as pricing changes and promotional activity. Gaps, inconsistencies, or very short time horizons significantly reduce model reliability.
Q: How do I know if marketing mix modelling is the right approach for my situation?
A: This approach is most suited to organisations with meaningful spend across multiple channels (particularly a mix of online and offline), sufficient historical data, and a need to make defensible budget decisions at a strategic level. It is less relevant for early-stage businesses with limited data history, organisations running fewer than three or four active channels, or teams whose primary need is optimising digital campaigns in near-real-time rather than strategic budget allocation. The deep dive covers fit criteria in more detail.
Q: Is marketing mix modelling only for large companies?
A: It was historically associated with large FMCG and retail businesses due to the data and analytical resource required, but accessibility has increased significantly. Mid-sized organisations with structured data and access to analytical capability, whether in-house or through a specialist agency, can now build credible models. The key threshold is not company size but data quality and the complexity of the channel mix being modelled.