Nobody will randomize acutely ill patients to delayed treatment to settle the question. TriEmulate writes the protocol for that trial anyway, then states exactly how each part of it would be carried out in the data you already hold.
Everything else is trying to replace it. A trial assigns treatment independently of prognosis. Routine care does not. Sicker patients get treated sooner, and that channel runs straight from confounder to outcome. Most of what goes wrong in observational research goes wrong before any model is fit: a start date that drifts, eligibility judged using information that only arrived later, a comparison no trial could have produced. Target trial emulation is the discipline of settling those questions first, in writing, and it is what TriEmulate automates.
A schematic, not a computation. The three assumptions named between the arms are what an emulation has to argue for, since randomization is not there to deliver them.
TriEmulate is enterprise infrastructure under active development and is not yet generally available, but it runs today and can be walked through end to end on a question of your own. A United States provisional patent application covering the system was filed in September 2026. If you work on causal inference in health data, bring a question and we will put it through.