Synthetic Control Method
A causal inference technique that constructs an artificial control group by weighting untreated units to closely match the treated unit's pre-intervention characteristics and trajectory.
What Is the Synthetic Control Method?
The synthetic control method (SCM) constructs a "synthetic" version of a treated unit by taking a weighted combination of untreated units that best matches the treated unit's pre-intervention trajectory. Developed by Abadie and Gardeazabal (2003), SCM extends the logic of difference-in-differences to cases where no single control unit is comparable enough, building a bespoke counterfactual from a pool of candidates.
Also Known As
- Marketing team: "synthetic control," "synthetic counterfactual"
- Sales team: "constructed control region"
- Growth team: "synthetic market test"
- Data team: "SCM," "Abadie synthetic control"
- Finance team: "constructed counterfactual analysis"
- Product team: "synthetic baseline comparison"
How It Works
You launch a major brand campaign in San Francisco and want to measure impact on sales. No single city matches SF perfectly. SCM constructs a synthetic SF from a weighted combination: Seattle (35%), Austin (30%), Portland (20%), Boston (15%). These weights are chosen so the synthetic pre-campaign sales track real SF's sales almost exactly. After the campaign, real SF rises to $5.2M/week while synthetic SF (unchanged control markets weighted the same way) stays at $4.5M/week. The $700K/week gap is the estimated campaign effect.
Best Practices
- Require a strong pre-intervention fit — synthetic and real should track closely for at least 12-18 months pre-treatment.
- Use placebo tests (apply SCM to untreated units) to validate the method's false-positive rate.
- Limit the donor pool to units that could plausibly be similar — don't include New York as a donor for rural Wyoming.
- Report confidence intervals via permutation tests, not just point estimates.
- Combine with other causal methods (DiD, regression) when stakes are high.
Common Mistakes
- Ignoring poor pre-period fit and still trusting the post-period comparison.
- Letting extreme weights dominate (e.g., 90% on one donor unit — effectively just a paired comparison).
- Applying SCM to short time series where noise swamps signal.
Industry Context
Ecommerce and DTC teams use SCM for market-level campaigns (TV, OOH) where a single matched control doesn't exist. SaaS and B2B use SCM less often but apply it to enterprise cohort studies and policy evaluations. Lead gen operations use SCM for local-market sales enablement tests where individual markets are too unique for direct comparison.
The Behavioral Science Connection
SCM embodies context-specificity — a principle behavioral economists champion. Rather than assuming a one-size-fits-all control, SCM constructs a bespoke comparison that respects the heterogeneity of markets, segments, or cohorts. This mirrors how behavioral interventions themselves need to be tailored to specific decision contexts rather than applied universally.
Key Takeaway
Synthetic control shines when you have one treated unit and a pool of candidates — use it when no single control is good enough.