By Bhavish Lekh, Co-founder & CEO
In a Phase III program, the decisions that carry the most capital and risk are made at the very beginning — at feasibility, long before the first patient is enrolled. How many sites, in which countries, at what enrolment rate, to reach Last Patient In by when. Those choices set the timeline the whole program is built around, and by the time a study is in the field they are extraordinarily expensive to unwind.
And they are almost always locked in on a single number.
The false comfort of a single-point estimate
The enrolment curve usually arrives as one confident line: one rate per site, one projected Last Patient In. It looks precise. But recruitment is one of the most unpredictable variables in study execution — sites over- and under-perform, countries move at different speeds, enrolment rises and falls with holidays and competing studies. None of it behaves like a fixed value, yet the single-number plan presents it as one. So the timeline reads as a confident promise while masking the risk that decides whether the program hits its date. And when key stakeholders ask how confident you really are, "it's our best estimate" is not a defensible answer.
I built these plans by hand
For ten years in feasibility, I built these enrolment models myself — point estimates, one assumption per site, stitched into a single projected timeline. And I know how quickly they get away from you. A single plan can span more than a dozen countries and hundreds of sites, each with its own variables, and within a few layers you are hunting for an assumption you made three tabs ago, no longer sure why you set it where you did.
Then the program moves. Six months on you are sitting across from key stakeholders explaining the gap between what was planned and where you actually are, reconstructing from memory the reasoning behind a number you set half a year earlier. Someone asks "what if the German sites come in slow?" and you rebuild the scenario by hand. Ask twenty of those questions and you have spent a week rebuilding twenty spreadsheets. And the tools made it worse: single-user by nature — one expert, one file, no clean way to collaborate or trace how the plan had changed.
A method finance and aerospace settled decades ago
When the inputs to a decision are genuinely uncertain, betting everything on one number is the weakest thing you can do. Finance and aerospace worked this out decades ago and converged on the same answer: instead of running the model once, run it thousands of times, varying the uncertain inputs within realistic ranges, and study the distribution of outcomes. That method is Monte Carlo simulation, and enrolment in a live study is a textbook case for it.
What that looks like in Simulator-L
Simulator-L™ is the scenario and modelling engine of the Aurora Suite™, built alongside active clinical teams working on real opportunities. Instead of rebuilding a scenario by hand twenty times, you set your assumptions once and Simulator-L runs ten thousand iterations, returning a probability distribution of enrolment outcomes rather than a single best guess. Strategy scenarios sit side by side: change the countries, the sites, the ramp-up, the seasonality, and watch the picture move.
The figure that changes the conversation is Probability of Success: the likelihood of hitting your target by your target date, as a percentage rather than a yes or no. It turns "we think we'll make it" into "this plan has this chance — and here is the range if it doesn't."
And it was built for a team, not a lone expert. Shared team access, version control and a full audit log keep the reasoning behind every projection visible, traceable and preserved — not locked in one person's spreadsheet and one person's memory.
De-risk the program decision — and prove it
It is a probability engine, not a crystal ball. Every projection is grounded entirely in the assumptions you provide — an honest, defensible picture of the range your plan implies, so you can plan for the realistic outcomes rather than the convenient one.
Monte Carlo has underpinned high-stakes decisions in finance and aerospace for decades. With Simulator-L™, it has come to clinical study execution — and the clearest way to see that is on a program you already know.
Want to explore a current program through our simulation engine? Book a demo now.
Bhavish Lekh is the Co-founder and CEO of Aurora Analytica. Connect with him on LinkedIn or reach out at bhav@aurora-analytica.com.