Measure of Mind

Queue
Studio

Admin

Simulate

of turns

An empty board

Drop an emitter to start a stream of work, then service nodes to process it. Hover a node and drag from an edge dot to connect it to another node.

Workstreams

The board

Select a node or a connection to configure it. Figures below update live from the Kingman approximation as you edit.

Predicted pressure — Kingman

E / S add a node · Del removes the selection · hover a node and drag an edge dot to connect · scroll to zoom, drag the board to pan.

Emitter

Name

Workstreams emitted

Arrival rate ?How much work arrives over time — the engine of your flow. e.g. 200 tasks every 1 day means 200 tasks arrive each day on average.

tasks per

Tasks per arrival ?How many tasks come in each time work arrives. Use a range to vary it — e.g. 1 to 3 means each arrival brings 1–3 tasks at random.

–

Spacing between arrivals ?How far apart arrivals are in time — e.g. 2–8 minutes means each arrival comes 2–8 min after the last (drawn at random). A wider range = more uneven, bursty flow; a narrow range = steadier.

–

Service node

Name

Working hours ?Teams don’t work 24/7. Capacity is scaled by the share of the week the team is actually staffed, so an 8h/day 5-day team shows ~4× more utilized than a 24/7 team under the same load. Increase these for cross-regional teams that hand off around the clock.

Hours per day

Days per week

Inbound workstreams

Nothing reaches this node yet. Connect an emitter or an upstream service node to it.

Time to serve one task ?How long one task takes at this step. Set a range (e.g. 2–6 minutes). The app uses the average and the spread to predict queues — a bigger spread means more unpredictable waits.

–

Servers ?How many tasks this step can handle at the same time. e.g. 3 = three tasks served in parallel. More servers = shorter waits, but higher cost.

Target utilization ?How busy this step is allowed to get, as a %. e.g. 85% = OK if it’s busy 85% of the time. The app turns red above this. Keep it under 100% to avoid a growing backlog. (ρ is the queueing symbol for utilization.)

After service

% completed here ?Each finished task has this chance of ending here; the rest continue downstream. During a run each task is decided randomly, so the actual split varies around this percentage.

Connection

Choose which workstreams travel this line, and whether each one forks in parallel or is routed by share.

Line routing ?Pick which side of each node the line attaches to, and where along that side. Use this to untangle overlapping lines.

Leaves from

Arrives at

Start position %

End position %

Label size ?Scale the breadcrumb text on this line up or down. Long breadcrumbs are truncated with … and show the full path on hover.

Complexity passed downstream ?Tasks travelling this line pick a random multiplier in this range that scales their service time at every downstream step (multipliers compound across hops). Use >1 for escalations that get harder — e.g. 1.5–3 for work needing a support team. 1–1 = no effect.

× – ×

Press Run to simulate. Until then this panel shows the Kingman prediction, recomputed on every edit.

Node	Workstream	λ in	Servers	Svc time	CV a / CV s	ρ	Wq — Kingman	Wq — sim	Avg queue	Max queue
{{ r.node }}	{{ r.stream }}	{{ r.lam }}	{{ r.c }}	{{ r.tau }}	{{ r.cv }}	{{ r.rho }}	{{ r.wq }}	{{ r.wqSim }}	{{ r.lq }}	{{ r.maxq }}

Workstream	Released	Completions	Throughput	Lead time — avg	p50	p95	In system at end	Path
{{ r.name }}	{{ r.released }}	{{ r.done }}	{{ r.thr }}	{{ r.avg }}	{{ r.p50 }}	{{ r.p95 }}	{{ r.wip }}	{{ r.path }}

Queue-depth traces appear here once a simulation has run.

{{ c.title }} peak {{ c.peak }}

0{{ c.endLabel }}

Clear the whole board?

This removes every node, connection and workstream, along with any simulation results. You will be left with an empty board.

Undo will bring it back if you change your mind. Export first to keep a copy on disk.

Tutorial {{ tourCount }}

Access code required

Enter the access code you were given to unlock the full simulator.

{{ tourTitle }}