After leading media, digital and ecommerce at various large and well-known brands what were some of the reasons I started a marketing consultancy with a significant focus on Marketing Mix Modelling?
As I’ve mentioned in a few of the ISBA trade body sessions we’ve run for their members, I think my interest in measurement began when I chaired their digital group for around five years (quite some time ago now). Back then, I was heavily involved in what many of their members will be doing at this time of year: a process that has since grown into something far more comprehensive across all their groups. We ran a survey to understand advertiser priorities for the upcoming year.
Every year except one, when I chaired the group, measurement was the top priority. That shouldn’t be a surprise, of course. Any marketer worth their salt wants to understand the impact of their spend and, crucially, be able to explain it to internal stakeholders – including finance.
As new media channels emerged, there was always a need to balance innovation (to ensure you are where your audience is) with the need to prove their return.
And yet over the following years in different roles I kept coming across a “blind spot” of how badly measurement was set up at brands investing enormous amounts in marketing. It was also pretty obvious to me that this could often be one of the biggest wins for their P&Ls.
The fact that much of the industry relied purely on last-click attribution for so long, and that many brands still do, is amazing once you understand its implications on potential performance.
MMM, when executed properly as part of your broader effectiveness and marketing programme, can unlock significant growth by giving organisations a clearer perspective on what drives performance and how to allocate investment.
But … as I’ve dug deeper into the discipline and we’ve developed our own client-side oriented point of view, it’s become clear how unaware many (I’d say most) brands have another measurement “blind spot”. Legacy MMM companies don’t like to talk about it … but it is the extent to which small methodological choices can have significant consequences on the results.
Various vendors that have come to the market over the last few years have fuelled the misconception that Marketing Mix Modelling is plug-and-play: feed in the data, press the button,” and out pops your ROI. In reality, MMM isn’t just a software platform or algorithms; it’s judgement, context, and expertise.
Every MMM study involves decisions that shape your results – often invisibly to the client. That’s why my call is to marketing budget holders, as I used to be to, as well as trying to get your head round how to make the most of AI or your tech stack add to your list some simple questions and develop some curiosity about how your MMM is set up.
Bayesian MMMs, such as Google’s Meridian, incorporate what are called “priors” (what the model assumes is true about channel performance before it sees the real evidence). While these can unintentionally reflect historical biases, they don’t inherently favour any particular channel by default. The risk lies in how those priors are set: poorly chosen or unchallenged priors can reinforce existing assumptions and skew results. Cynics point out indeed that may well have been why Google selected it of course.
How you account for the lingering effects of campaigns with ad stocks (the concept that ads continue to influence people for a period of time) or decay rates (the speed at which that effect fades) fundamentally changes the output. Get the decay curve wrong and you’ll either over-credit channels for sales they didn’t drive, or under-credit them for their long-term brand impact.
Not all MMMs are created equal. Different methodologies can produce materially different results – even on identical datasets.
Ignoring methodology isn’t academic nitpicking. It leads to:
- Misallocated budgets in putting money into channels that aren’t actually driving the return you think.
- Missed opportunities in failing to uncover the real significant levers that could transform your business
Yes MMMs can be incredibly powerful – but your team needs to understand the methodology and its assumptions.
Don’t just go with “dashboard easy” getting transparency and some education will lead to actionable understanding you can trust.
Alex Tait, Founder, Entropy
