Cold-Start Markets

By the end of this article, you should be able to: say why a brand-new market breaks synthetic control and difference-in-differences, pick the right question to answer before reaching for a method, design a staggered rollout that avoids the problem, and build a launch-curve benchmark when the rollout order is already fixed.

Step 1: Meet the problem

A delivery marketplace opens its 41st city, Boise. Four weeks in, weekly gross order value sits at $180k. Leadership wants to know whether the launch is working.

Try the usual tools and every one of them dies in the same place:

Synthetic control builds a weighted blend of other cities that matches the treated city before treatment. Boise has no before.
Difference-in-differences subtracts the treated unit's own pre-period trend. There is no pre-period trend to subtract.
Causal impact forecasting forecasts the counterfactual from the unit's own history. There is no history to forecast from.

That is not three separate problems. Every one of these methods estimates what the treated unit would have done by learning the unit's own behavior first, and a cold-start market has given you nothing to learn from. The chart below is the entire difficulty: one point, no line behind it.

Boise (actual)

Boise, four weeks after launch. Illustrative. There is nothing to the left of week 1, which is exactly what every standard method needs.

Step 2: Work out which question you are actually being asked

Before choosing a method, separate three questions that get said in the same breath and need completely different answers.

"Should we have launched here?" This is a counterfactual about a decision already made. It is rarely a live question, and it is the hardest of the three. Say so rather than pretending you can answer it cleanly.

"Is this launch ahead or behind where it should be?" This is almost always the real question, and it is answerable. Note what changed: you are no longer comparing Boise to Boise without us. You are comparing Boise to what a launch like this normally looks like at week 4. That is a different and much easier counterfactual, and it is available.

"Did the thing we did inside Boise work?" A promotion, a fee change, a courier incentive. This is an ordinary causal question that happens to be inside a new city, and Step 7 handles it.

Getting an interviewer to agree which question is on the table, before you name a technique, is most of a strong answer.

Quick check. Leadership asks "did launching in Boise pay off?" and expects a number this quarter. What do you say?