Expanding the Causal Toolkit | New Method: Synthetic Control Method (SCM) - Feasibility & Requirements
- Maria Alice Maia
- Mar 24
- 3 min read
What was the ROI of that massive Olympics ad campaign? Or the real impact of your new flagship store?
When you make a huge, one-of-a-kind bet, you can’t run a simple A/B test. There’s no parallel universe where you didn’trun the campaign. So, how do you prove it worked?
Most companies guess. They compare their results to a "similar" region or a previous time period and call it a day. This isn't just "Doing Data Wrong"—it's a catastrophic failure of imagination and rigor that leaves millions in value unaccounted for.
The "Doing Data Wrong" Scenario: The Marketing Mirage
A CPG company launches a massive, nation-wide advertising campaign during the World Cup. To measure its impact, the marketing team compares sales growth in their country to sales growth in a neighboring country where the campaign didn't run.
They see a 5% lift and declare victory. The problem? The neighboring country's economy was slumping, their currency was volatile, and a local competitor was running a promotion. The comparison was meaningless. They didn't isolate the campaign's effect; they just measured noise.
The "Right Way": Building a Data-Driven Doppelgänger
When you can't find a control group, you have to build one. This is the power of the Synthetic Control Method (SCM), a game-changing technique for exactly these situations.
Instead of picking one flawed comparison, SCM creates the perfect counterfactual—a "synthetic" twin of your treated unit, built from a weighted average of multiple untreated units.
In our marketing example, we wouldn't use one neighboring country. We’d use data to find the perfect recipe—say, 40% of Country A, 25% of Country B, 15% of Country C, and 20% of Country D—that, when combined, perfectly matched our target country's pre-campaign sales trajectory, consumer confidence, and other key predictors.
This "synthetic country" shows us what would have happened without the campaign. The difference between our actual sales and the synthetic sales is the true, causal impact. No more guessing.
This isn't a cute trick; it's a rigorous method detailed in top journals like the Journal of Economic Literature. It requires a long history of pre-intervention data and a "donor pool" of comparable units to work, but when the conditions are right, it's one of the most powerful tools we have.
What You Need to Do:
Managers: Your biggest, most strategic bets often have the murkiest ROI. Stop accepting flimsy comparisons. Ask your team: "We can't A/B test this, so how can we build a synthetic control to prove the impact?" This is how you justify high-stakes investments.
Tech & Data Professionals: This is where you elevate your game from reporting to strategic counsel. Identify the big, one-off events in your business—a new market entry, a major policy change, a national campaign. Propose using SCM. You have the power to deliver the definitive answer when everyone else is just guessing.
My career has been built on finding the ground truth, whether at Itaú, Stone, or in my own ventures. The knowledge I’ve gathered from FGV, Berkeley, HEC, and Cass Business School isn't a private asset; it’s a toolkit to be shared. My purpose is to give it back, so we can all get better.
If you're done with data theater and want to join a community of practitioners focused on real-world impact, subscribe to my email list for more no-nonsense, research-backed insights.
And if you have a real, high-stakes business case you're struggling with, let's talk. I'm opening up a few 20-minute slots for no-nonsense consultations. Let's solve it together.
