Operating model simulator

Generated 2026-07-14 19:11 · engine-backed interactive report · QA PASS · slider-formula max error 7.7 percentage points vs held-out engine runs

Plain-language legend - what the abbreviations mean (nothing here assumes prior knowledge)

How the operating model can evolve - pick a stage to explore it below

Each stage keeps the same work and the same facts - only the way of working changes. Arrows show the direction of travel. Green delta = better than today, red = worse.

What this stage looks like - the org chart and the process it runs

The three numbers that matter - they recolour live as you move the levers

Ramps the workload up and shows exactly where each KPI turns red. Watch the tiles.
KPI = Key Performance Indicator. Green = comfortable, amber = getting tight, red = breached. Each tile names the point where it turns red.

Tune the world - untick a lever to switch it off; every chart recomputes instantly

Drag a slider or type an exact number. Untick a lever's box to hold it at "no change" and take it out of play. Dots on the lever charts further down are real engine runs; values between them are estimates (QA-measured error quoted at the top).

The stages, at your current settings

Shaded band = the realistic best-to-worst spread across re-runs with different luck (P10 to P90); the line is the middle (P50). Level-adjusted to your sliders (approximation, QA-measured).

Which levers matter most

Each lever, one at a time

Green zone = above your target. Dashed red line = where you are now. Each chart ends with the takeaway in plain terms and the point where it breaches.

What could break this - the risk tree

The story, in plain English

Assumptions ledger

AssumptionStatusValue
Arrival volumes, intraday and weekly shapesasked / stated in specorgspec baseline
Handle times and rework rates per stepasked / stated in specorgspec steps
Agent throughput and error rateuser estimateorgspec agent_pools
Capacity tile (busiest-team load) scales base utilisation by the load factorstated-default (directional estimate)dashboard
Agents process in parallel at fixed per-item timestated-default (simplification)engine
Human service may finish after shift end (overtime)stated-default (simplification)engine
Max one rework per step per itemstated-default (simplification)engine
Costs accrue per rostered weekday FTE; agent items flat coststated-defaultengine
Sliders combine via an engine-measured LOAD curve; stress levers via load equivalenceQA-validated approximationmax err 7.7 SLA pp
Trajectory band level-shifted from engine base runsapproximationfan chart note

Quality checks - every check, individually

Verdict: PASS - 49 individual checks: 42 green, 7 amber, 0 red. Green = comfortably inside tolerance, amber = inside but close, red blocks release. Slider-formula max error 7.7 SLA points at held-out engine validation runs.
What was checkedOwning agentInputMethodResult observedThreshold
● greenReplication count - baselineSimulation Runnerengine configurationcount seeded replications of the base run10 replications, distinct seeds by construction>= 5 replications
● amberFlow conservation - baselineCalculation QAengine completions (run r4242x10-91407796) vs orgspec arrival ratesLittle's-law style: completed items must equal spec-implied arrivalsengine 11215 vs analytical 12026 (6.7% apart)<= 10%
● greenCapacity math - baseline / Intake (APAC)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 79% vs analytical 80% (diff 1%)<= 18%
● greenCapacity math - baseline / Repair (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 88% vs analytical 89% (diff 1%)<= 18%
● greenCapacity math - baseline / Investigations (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 94% vs analytical 95% (diff 1%)<= 18%
● greenCapacity math - baseline / Verification (AMER)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 78% vs analytical 80% (diff 1%)<= 18%
● greenQueueing law: load raises delay - baselineCalculation QAvolume sweep turnaround (run r4242x10-91407796)Operational Research (queueing theory): turnaround must rise, not fall, as offered load increases - Kingman / Pollaczek-Khinchine delay grows with utilisationlargest downward reversal 0% of peak turnaround<= 25%
● greenSurrogate fidelity - baseline @ {'volume_mult': 1.15, 'fte_mult': 1.15}Visualizerheld-out engine run r4242x8-92386084 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 40.8% vs engine 41.6% (error 0.8 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - baseline @ {'volume_mult': 1.3, 'handle_mult': 1.15}Visualizerheld-out engine run r4242x8-89119459 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 0.8% vs engine 1.0% (error 0.2 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - baseline @ {'volume_mult': 0.85, 'fte_mult': 0.85}Visualizerheld-out engine run r4242x8-62696133 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 40.8% vs engine 40.4% (error 0.4 SLA points)<= 8.0 SLA points
● amberSurrogate fidelity - baseline @ {'volume_mult': 1.15, 'error_mult': 1.5}Visualizerheld-out engine run r4242x8-47853680 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 24.9% vs engine 18.5% (error 6.5 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - baseline @ {'volume_mult': 1.1, 'geo_outage_days': 3}Visualizerheld-out engine run r4242x8-4329489 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 7.8% vs engine 9.3% (error 1.5 SLA points)<= 8.0 SLA points
● greenReplication count - two_hubSimulation Runnerengine configurationcount seeded replications of the base run10 replications, distinct seeds by construction>= 5 replications
● amberFlow conservation - two_hubCalculation QAengine completions (run r4242x10-91407796) vs orgspec arrival ratesLittle's-law style: completed items must equal spec-implied arrivalsengine 11208 vs analytical 12026 (6.8% apart)<= 10%
● greenCapacity math - two_hub / Intake (APAC)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 79% vs analytical 80% (diff 1%)<= 18%
● greenCapacity math - two_hub / Repair (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 88% vs analytical 89% (diff 1%)<= 18%
● greenCapacity math - two_hub / Investigations (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 94% vs analytical 95% (diff 2%)<= 18%
● greenCapacity math - two_hub / Verification (moved to EMEA hub)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 58% vs analytical 59% (diff 1%)<= 18%
● greenQueueing law: load raises delay - two_hubCalculation QAvolume sweep turnaround (run r4242x10-91407796)Operational Research (queueing theory): turnaround must rise, not fall, as offered load increases - Kingman / Pollaczek-Khinchine delay grows with utilisationlargest downward reversal 0% of peak turnaround<= 25%
● greenSurrogate fidelity - two_hub @ {'volume_mult': 1.15, 'fte_mult': 1.15}Visualizerheld-out engine run r4242x8-92386084 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 77.0% vs engine 77.0% (error 0.0 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - two_hub @ {'volume_mult': 1.3, 'handle_mult': 1.15}Visualizerheld-out engine run r4242x8-89119459 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 1.1% vs engine 0.9% (error 0.2 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - two_hub @ {'volume_mult': 0.85, 'fte_mult': 0.85}Visualizerheld-out engine run r4242x8-62696133 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 77.0% vs engine 76.8% (error 0.1 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - two_hub @ {'volume_mult': 1.15, 'error_mult': 1.5}Visualizerheld-out engine run r4242x8-47853680 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 45.2% vs engine 41.8% (error 3.3 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - two_hub @ {'volume_mult': 1.1, 'geo_outage_days': 3}Visualizerheld-out engine run r4242x8-4329489 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 14.0% vs engine 17.2% (error 3.2 SLA points)<= 8.0 SLA points
● greenReplication count - agent_intakeSimulation Runnerengine configurationcount seeded replications of the base run10 replications, distinct seeds by construction>= 5 replications
● amberFlow conservation - agent_intakeCalculation QAengine completions (run r4242x10-91407796) vs orgspec arrival ratesLittle's-law style: completed items must equal spec-implied arrivalsengine 11415 vs analytical 12026 (5.1% apart)<= 10%
● greenCapacity math - agent_intake / Intake (APAC, exceptions)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 55% vs analytical 56% (diff 1%)<= 18%
● greenCapacity math - agent_intake / Repair (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 90% vs analytical 89% (diff 0%)<= 18%
● greenCapacity math - agent_intake / Investigations (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 95% vs analytical 95% (diff 0%)<= 18%
● greenCapacity math - agent_intake / Verification (AMER)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 79% vs analytical 80% (diff 0%)<= 18%
● greenQueueing law: load raises delay - agent_intakeCalculation QAvolume sweep turnaround (run r4242x10-91407796)Operational Research (queueing theory): turnaround must rise, not fall, as offered load increases - Kingman / Pollaczek-Khinchine delay grows with utilisationlargest downward reversal 0% of peak turnaround<= 25%
● greenSurrogate fidelity - agent_intake @ {'volume_mult': 1.15, 'fte_mult': 1.15}Visualizerheld-out engine run r4242x8-92386084 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 67.6% vs engine 68.1% (error 0.5 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - agent_intake @ {'volume_mult': 1.3, 'handle_mult': 1.15}Visualizerheld-out engine run r4242x8-89119459 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 2.0% vs engine 2.0% (error 0.0 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - agent_intake @ {'volume_mult': 0.85, 'fte_mult': 0.85}Visualizerheld-out engine run r4242x8-62696133 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 67.6% vs engine 67.1% (error 0.5 SLA points)<= 8.0 SLA points
● amberSurrogate fidelity - agent_intake @ {'volume_mult': 1.15, 'error_mult': 1.5}Visualizerheld-out engine run r4242x8-47853680 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 43.8% vs engine 36.5% (error 7.3 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - agent_intake @ {'volume_mult': 1.1, 'geo_outage_days': 3}Visualizerheld-out engine run r4242x8-4329489 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 43.5% vs engine 41.4% (error 2.0 SLA points)<= 8.0 SLA points
● greenReplication count - ai_verifySimulation Runnerengine configurationcount seeded replications of the base run10 replications, distinct seeds by construction>= 5 replications
● amberFlow conservation - ai_verifyCalculation QAengine completions (run r4242x10-91407796) vs orgspec arrival ratesLittle's-law style: completed items must equal spec-implied arrivalsengine 11212 vs analytical 12026 (6.8% apart)<= 10%
● greenCapacity math - ai_verify / Intake (APAC)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 79% vs analytical 80% (diff 1%)<= 18%
● greenCapacity math - ai_verify / Repair (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 88% vs analytical 89% (diff 1%)<= 18%
● greenCapacity math - ai_verify / Investigations (EMEA)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 94% vs analytical 95% (diff 1%)<= 18%
● greenCapacity math - ai_verify / Verification exceptions desk (AMER)Calculation QAengine utilization (run r4242x10-91407796) vs orgspec demand mathindependent re-derivation: demand hours / supply hours from the specengine 12% vs analytical 12% (diff 0%)<= 18%
● greenQueueing law: load raises delay - ai_verifyCalculation QAvolume sweep turnaround (run r4242x10-91407796)Operational Research (queueing theory): turnaround must rise, not fall, as offered load increases - Kingman / Pollaczek-Khinchine delay grows with utilisationlargest downward reversal 0% of peak turnaround<= 25%
● greenSurrogate fidelity - ai_verify @ {'volume_mult': 1.15, 'fte_mult': 1.15}Visualizerheld-out engine run r4242x8-92386084 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 75.1% vs engine 75.2% (error 0.1 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - ai_verify @ {'volume_mult': 1.3, 'handle_mult': 1.15}Visualizerheld-out engine run r4242x8-89119459 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 1.1% vs engine 0.9% (error 0.1 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - ai_verify @ {'volume_mult': 0.85, 'fte_mult': 0.85}Visualizerheld-out engine run r4242x8-62696133 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 75.1% vs engine 75.2% (error 0.1 SLA points)<= 8.0 SLA points
● greenSurrogate fidelity - ai_verify @ {'volume_mult': 1.15, 'error_mult': 1.5}Visualizerheld-out engine run r4242x8-47853680 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 36.7% vs engine 36.5% (error 0.2 SLA points)<= 8.0 SLA points
● amberSurrogate fidelity - ai_verify @ {'volume_mult': 1.1, 'geo_outage_days': 3}Visualizerheld-out engine run r4242x8-4329489 (never used to fit the surrogate)dashboard slider formula (load equivalence) vs a real engine run at the same settingsslider formula 61.4% vs engine 69.2% (error 7.7 SLA points)<= 8.0 SLA points
● greenData hygiene scrubCalculation QAevery .md/.yaml/.html file in the reporegex scan for PII patterns (emails, phone numbers) and a blocklist of real institution namesno findingszero findings

Settings used for this export

- prints the report exactly as tuned (choose "Save as PDF").