Inspect the comparison markets that make the synthetic control

Define the treated unit and the comparison opportunity

Identify the intervention, affected geography or business unit, outcome and exposure date. Include any earlier announcement or partial rollout that could change behavior before the formal launch. Check that the proposed treatment is sufficiently distinct from ordinary ongoing activity for the causal question to be meaningful.

Synthetic-control method coauthor Jens Hainmueller describes constructing a weighted combination of untreated donor units to match pre-treatment outcomes and predictors. His guide emphasizes donor credibility, pre-period fit and contamination risks. Those principles make the comparison inspectable; they do not make the later gap causal without the design’s assumptions.

Review donor eligibility before looking for the best effect

List potential donors and document why each is an appropriate comparison. Check outcome definitions, reporting coverage, business model, relevant shocks and exposure to related campaigns. A neighboring market may be similar but receive spillover media or customer travel from the treated area. Similarity is not enough if the comparison is partly treated.

Choose and preserve eligibility rules without using the desired post-launch outcome to select convenient donors. Record exclusions with their substantive reasons. If a donor experiences a later unique event, have the analyst assess the implications transparently rather than quietly removing it until the result improves. Keep the distinction between planned eligibility and post-event sensitivity analysis visible.

Inspect historical fit and weight concentration

Ask to see the treated and synthetic series before the intervention, not only a post-period gap chart. Inspect important subperiods and known seasonal changes. A close average can hide poor tracking where the business most needs comparability. Discuss what the observed mismatch means for the proposed claim.

Review the donor weights and the role of each heavily weighted unit. A result dominated by one donor depends strongly on that donor’s continuing relevance. Ask what happens under defensible changes to the donor set or predictor choices and how those checks were specified. Do not demand a particular visual fit by adding variables selected from post-treatment outcomes, which would contaminate the construction of the comparison.

Treat inference and business scope as separate review questions

Have the analyst explain the uncertainty or placebo procedure and why it fits this design. If placebo comparisons are used, inspect their pre-period fit and eligibility rather than reading one rank as a universal significance guarantee. A synthetic control can be informative while remaining fragile to a small donor pool or a unique concurrent event.

Connect any supported estimate to the actual unit, period and intervention. Do not assume the same effect will appear in every expansion market. Preserve the data version, donor rules, weights and analysis choices so the comparison can be reproduced. If pre-fit or contamination is unacceptable, state that the design cannot support the requested attribution instead of presenting the most attractive model as settled evidence.

Make the reasoning concrete.

A fictional home-goods brand launches a city campaign and receives a synthetic-control report. Most comparison weight comes from a neighboring city that also receives substantial spillover advertising, but the report shows only the final lift number.

The team identifies the spillover concern and requests a transparent design review. It does not remove donors merely to increase the estimate or generalize one city’s result to all markets.

Take the questions into your next brief.

Use these prompts in your own notes. Nothing is submitted from this page.

1. Unit

   What intervention and population does the treated series represent?

2. Donors

   Why is each candidate an appropriate untreated comparison?

3. Fit

   Where does the synthetic series fail before launch?

4. Weights

   Which units carry most of the result?

5. Sensitivity

   Which defensible changes alter the conclusion materially?

Watch for the wrong conclusion.

• Selecting donors for a favorable post-launch gap.
• Calling a spillover-exposed market untreated.
• Hiding poor historical fit behind one effect estimate.

Sources & scope

Workflows, worksheets and fictional examples are original editorial analysis. Named research guidance supports identified method principles. This is design-review preparation for a qualified analyst, not an estimation recipe or a claim of observed business results.

• Jens Hainmueller: Synthetic Control Methods, an Explainer →Primary research-method guidance · Supports: Weighted untreated donors, pre-period fit, donor weights and sensitivity or inference considerations. The marketing review pack is original editorial analysis.