Each simulated batch travels six sequential real stages of an autologous (patient-matched) cell-therapy pipeline: Collection (apheresis) → Ship Out (cold-chain to facility) → Manufacturing/Expansion → QC Release Testing → Ship Back (cold-chain to clinic) → Infusion. A batch's total vein-to-vein time T is the sum of six independent random stage durations, each drawn from a log-normal distribution built from its own mean and variability (coefficient of variation):
T = Σ stage_i, stage_i ~ LogNormal(mean_i, cv_i)
σ² = ln(1+cv²), μ = ln(mean) − σ²/2, draw = exp(μ + σ·z), z ~ N(0,1)
The log-normal is the standard model for biomanufacturing lead times: durations are strictly positive and right-skewed (a stage rarely finishes early, but can run long — a failed QC test, a stalled expansion culture).
Patient health decline uses the same logistic dose-response as the 3D vein-to-vein model, evaluated at the elapsed waiting time:
health(days) = 100 × [1 − 1 / (1 + exp(−k·(days − T₀)))] k≈0.09, T₀≈24 days
Bottleneck insight (verified numerically over 200,000 simulated batches): Manufacturing/Expansion has by far the largest mean duration, so it is the pipeline's bottleneck. Halving its variability (CV) shrinks the 95th-percentile total turnaround by roughly 1.5–1.6 days. Halving the variability of a short, non-bottleneck stage (e.g. Collection) moves the P95 by well under a tenth of a day — confirming that tightening variability pays off most when applied to the longest-mean stage, not just any stage.
- Mean sliders — the average duration of each stage in days.
- Variability sliders — coefficient of variation (spread) of that stage's random duration; higher = more day-to-day unpredictability.
- Run 1 batch — animates a single token through the pipeline in real time.
- Run 200 (stats) — instantly draws 200 more batches and folds them into the live histogram, mean, P50 and P95 without animating them.
- Tighten bottleneck / Tighten Collection — one-click presets that halve the CV of Manufacturing (the bottleneck) or Collection (a non-bottleneck) so you can directly compare their effect on P95.