A counterfactual is the alternative outcome that cannot be directly observed. If a company requires office attendance and employee resignations rise, the question is how many resignations would have occurred without that policy. The difference between the observed outcome and that hypothetical outcome is the effect researchers want to estimate.
Randomized trials approximate a counterfactual through groups assigned by chance. Observational studies may use matched organizations, difference-in-differences or synthetic control methods. Each approach offers a different way to estimate the missing comparison and comes with assumptions.
A before-and-after chart is not a complete counterfactual. Economic conditions, restructuring, compensation, team composition or the usual trend over time could change outcomes even if the policy had stayed the same.
When an article claims an RTO mandate caused employee departures, ask what comparison represents the world without that mandate. The credibility of that comparison determines how much causal weight the result can bear.