Synthetic control

Synthetic control constructs a counterfactual for a single treated unit, often one city, state, or market that received a policy, campaign, or price change, by combining several untreated units into a weighted composite that closely tracked the treated unit before the treatment began. The gap between the treated unit's actual outcome and the synthetic composite's outcome after treatment is the estimated effect.

It extends difference-in-differences logic to cases with only one or a few treated units and many possible untreated comparisons, choosing weights so the synthetic control's pre-treatment trend matches the treated unit as closely as possible, rather than relying on a single comparison unit chosen by judgment. The fit of the synthetic control in the pre-treatment period is the main evidence of the method's credibility: a poor pre-period fit undermines confidence in the post-period comparison.

The method is common in geo experiments and market-level marketing measurement, where a brand tests a new campaign in one city and needs a credible counterfactual for what that city's sales would have been without it. Its main limitation is data-hungriness, it needs a reasonably long pre-treatment history and several plausible comparison units, and its results can be sensitive to which candidate units are included in the donor pool.