04/Simulation

Find out if a redesign works before anyone lives with it.

Run the new process thousands of times before you roll it out.

Build an input set from your real run data or from assumptions you can edit, then run design options at volume and compare what each one does.

Simulation · Onboarding v2.4
4 variants · 10,000 runs each 340 arrivals / week
A Current design
6d 04h median
B Single approval Best
2d 11h median
C Parallel review
3d 02h median
D Auto-approve under $5k
2d 18h median
All 4 branches exercised · 0 unreached steps · variant B clears the backlog at 340/week
SOC 2 Type 2 compliant Trust center

Setting up a run

Start from your real numbers, or defaults you can edit

Sapeum constructs a starting input set from live run data where you have it, and from sensible assumptions where you do not. Every value stays editable.

  • Inputs derived from live telemetry or stated assumptions
  • Volumes, durations and branch rates all adjustable
Input set · Onboarding v2.4
Arrival rate 340 / week
Credit check duration 4h ± 1h20
Escalation branch rate 14%
Approver availability 2 FTE, business hours
Seeded from 1,533 observed runs · every field editable

Comparing options

Compare designs on throughput and cycle time, not opinion

Try the versions you are actually choosing between: one approver or two, parallel review or sequential. Compare what each does to throughput and cycle time.

  • Multiple design variants against one input set
  • Expected output, throughput and cycle time per variant
Variants · 10,000 runs each
A · current design
6d 04h median
B · single approval Best
2d 11h median
C · parallel review
3d 02h median
D · auto-approve under $5k
2d 18h median
Variant B clears the backlog at 340/week; A does not

How much was tested

See which branches and edge cases were actually covered

Run at scale rather than walking through one happy path, so coverage across branches and edge cases is measured rather than assumed.

  • High-volume runs for statistical confidence
  • Branch and edge-case coverage reported, not assumed
Coverage · variant B
Happy path 78% of runs
Escalation branch 14% of runs
Rejection and rework 7% of runs
Approver unavailable 1% of runs
All four branches exercised · 10,000 runs · 0 unreached steps

Know the outcome before you commit to it

The expensive test stops being production.

Choose a design on numbers, not argument

Four options ranked by median cycle time.

Edge cases found before customers find them

Coverage is measured, so gaps show up first.

Common questions

What if we have no live data to base inputs on?

Sapeum builds a starting input set from reasonable assumptions and shows you every one it made. You adjust the ones you disagree with, which is usually a much faster conversation than starting from a blank model.

How is this different from just estimating?

An estimate gives a single number for a happy path. A simulation runs the branches, the queueing and the edge cases thousands of times, so you get a distribution and a coverage report rather than one optimistic figure.

Can we simulate a design we have not built yet?

Yes, that is the point. Simulation runs against a future-state design, so a variant can be tested and discarded before anyone configures a system or retrains a team.

How many runs is enough?

Enough that the numbers stop moving. Runs are cheap, so simulations are sized for confidence rather than for patience, and the coverage report shows which branches were actually exercised.

Does simulation feed into anything else?

Yes. A variant that wins in simulation becomes a candidate future state, and the result is one of the inputs to process improvements alongside analysis gaps and AI suggestions.