The importance of the other why in marketing
There’s the famous book on why, one that fills keynotes and carousels. And then there’s the other one. The unfriendly one. ‘The book of why’ has statistics, maths and formulas. It wasn’t a bestseller. It’s not on your “marketing must-reads” lists. And yet, it’s the book that changed how I think and work.
‘Start with why’ is memorable. The golden circle has been drawn a gazillion times since Sinek wrote his book. The book on why is far more demanding. You don’t quote it. You use it to ask the question that makes dashboards nervous: What would have happened if we hadn’t done X?
From “what happened” to “what caused it”
Pearl’s big idea is a ladder with three rungs:
Association: noticing that X and Y move together. (Our website visits increase when we run this DOOH campaign)
Intervention: asking what happens if we do something. (What if we run longer campaigns?)
Counterfactuals: asking what would have happened otherwise. (Would this customer have visited the site if they hadn’t seen the ad?)
Most marketing sits on the first rung. Dashboards glow with associations: trends, clusters, anomalies. The trouble is that association can’t settle the only question that matters for decisions: “If we change this, will reality change in the way we expect?” That’s the second rung. And the third rung is where we answer the CFO’s favorite line: “What did the spend truly add?”
Correlation doesn’t imply causation
True. And incomplete. Pearl’s more useful point for marketers: “where there’s smoke, there might be fire”. A correlation can be a real effect, a shared cause or nonsense. The data won’t decide this for you, your model will.
So treat correlations as hypotheses, not headlines. Sketch the causal story, name the usual suspects (season, promos, paydays, news, trends), then test if the hypotheses still holds true. Without a model, a spike is gossip. With a model, it’s evidence you can verify.
Every attribution report is a story about cause
Whether it admits it or not. If display “drives” conversions only because it spikes on paydays when people buy anyway, you’re measuring a halo, not a lift. The right question isn’t “Who gets credit?” but “What changed because of us and what would have happened anyway?” That’s incrementality. That’s uplift. That’s budget well spent.
Drawing causal maps
The simplest upgrade is a pencil and a page (or a Remarkable like I like to do). One diagram, arrows only: what could cause what, what else is moving, where confounding might creep in. You’ll argue about the arrows.. That argument tells you what to measure, what to control, and what to leave alone.
Marketing is not science but could use some
You don’t need a PhD. You need curiosity, a diagram, and the discipline to ask for the counterfactual. Pearl’s toolkit won’t turn your job into a math exam; it will make you harder to fool. By dashboards, by neat narratives, and by yourself.
