Plan tracking: which plan are we following, and which simulation is reality closest to?
Pick a month to re-run the judgement using only actuals up to the end of that month. The later the month, the more evidence there is and the less room remains to switch plans.
Which simulation is reality closest to (probability over time)
Starts from the weights at planning time. Each month’s actuals (late returns, failures, unscheduled removals, kit lead-time quotes) update the weights according to which simulation (Monte Carlo) they resemble.
Evidence: actuals vs what each simulation expected
Green = the simulation closest to actuals. “—” = no evidence yet (e.g. no engine has returned).
Which candidate plan are we following?
If we switch now (expected cost for the rest of the period)
Compares plans assuming engines already inducted stay as they are and only engines not yet inducted follow the candidate plan. Weighted by each simulation’s probability. Changes past their decision deadline assume emergency arrangements.
Deviations from plan
Year-end outturn forecast ($M)
Candidate plans × simulations: expected cost if the plan is kept to the end ($k) and cancellation risk
A table fixed together with the baseline annual plan. It shows which plan holds up in which simulation, and how much is lost when reality turns out differently.
AI: 3 lines for the executive meeting
Answer check (ground truth of the synthetic data)
Data are synthetic values estimated from public information. Actuals are synthetic too (the baseline plan executed in one scenario of one simulation).