Marketing’s dangerous obsession
Models are a quantification of what has already happened. That is useful in many circumstances, writes Jen Davidson, but it won’t tell you where you want to go.
The author Jen Davidson
The surge in interest around Market Mix Modelling (MMM) says less about the brilliance of the models and more about the uncertainty inside marketing teams. Despite being flooded with data, many clients feel less certain than ever about what’s really working.
In our world, we’ve fielded more questions about MMMs this year than any other topic. But a close second are questions on the blunt use of frameworks (60/40 a particular favourite!) to justify marketing investment.
A data point, not a strategy
The problem isn’t the models themselves; the problem is what we’ve allowed them to become. An MMM will tell you what happened when you did ‘X’. It measures efficiency. It quantifies historical performance. That’s valuable. But somewhere along the way, we started treating the model as the answer, rather than an input.
We’ve created an industry where the algorithm has veto power over the strategist. When you and your three biggest competitors all use similar MMMs, trained on similar category data, optimising toward similar KPIs, you inevitably converge on the same “efficient” answer. The model has turned strategy into a math problem with (seemingly) one optimal solution. And everyone’s solving for the same variables. Which goes some way to explain why so many campaigns within a category look the same.
I think Jen makes a great point – there can often be an over reliance on tools to ‘tell you the answer’ rather than using all the data and experience at a marketers disposal to make informed and critically, nuanced decisions based on the dynamics that exist in your category.
I don’t think the mansplaining of the first comment is helpful or appropriate.
Grab the MMM bull, get the (Innis) horns!
Jen, think we need some updated assumptions here!
1) MMMs are not trained on category data
2) Brands often supply competitive data for their models (which will detect a change)
3) I highly doubt an MMM would say the same thing per brand: they are often a reflection of how that brand generates outcomes in the real world
As you say, they are an input to a discussion and a diagnosis as to what’s working. But I think the statements made about MMMs also is a misunderstanding about some of the core things happening.
And as I’ve offered over email to you lately, happy to take you through some of the new research and work and looking forward to helping show you some of this in the New Year.
Much MMM is a lot like the campaigns decades ago when “Same again then?” was the norm.