← Back to Digital Labs
Asher Krane Labs

Resource Fencer
Why your 75% is actually 103%

Managers allocate hours on spreadsheets. Humans live inside switching costs, fragmented attention, and queueing theory. This tool shows what happens in the gap between the two.

Spreadsheet view
Human reality Showing the optimistic version
Scenario inputs
Weekly capacity 40h
Contractual hours in the working week. The total container.
Operational reserve 15%
The non-negotiable slice that's structurally unavailable — email, admin, bio breaks, corridor conversations, the time between meetings that's too short to use. Goldratt called this protective capacity. You cannot plan into it.
Concurrent projects 2
Each additional project forces a context reload — and that reload costs real hours.
Cognitive mode mixing Low
Mostly similar work — e.g. coding across two repos, or writing two reports. Low gear-shift penalty.
Meetings per week 4
Each meeting doesn't just cost its duration — it fragments the blocks around it.
Interruptions per day 3
Slack pings, "quick questions", drive-bys. Each one costs ~15 min of refocus.
FLOW ZONE
24.0h
usable hours out of 40
Throughput
1.00x
Delivery speed vs calm baseline
Lead-time stretch
1.00x
How much longer things actually take
Hours lost
16.0h
Reserve + invisible overhead combined
Throughput degradation curve
Human-adjusted
Textbook (machine)
You are here
Where the week actually goes
Usable hours
Operational reserve
Switching tax
Asymmetry tax
Fragmentation
Operational reserve
The structural overhead of existing in an organisation. Email, admin, bio breaks, hallway conversations, the gaps between meetings too short to use. Goldratt's Critical Chain calls this protective capacity — it's not slack, it's shock absorption. Plan into it and you guarantee failure.
Switching tax
The cognitive cost of juggling multiple projects. Every context switch forces your brain to unload one mental model and reload another. That reload isn't instant — it eats real hours that never appear on any timesheet.
Asymmetry tax
The penalty for mixing unlike types of work. Hopping between two coding tasks is one thing. Hopping between deep code and a stakeholder workshop is cognitively violent — the mental gears don't just shift, they grind.
Fragmentation
Meetings and interruptions don't just cost their own duration. They shatter focus blocks into pieces too small for deep work. A 30-minute meeting in the middle of the morning destroys the productive blocks around it.
Kingman's formula
The spreadsheet says
40h
available for work
The human actually has
24.0h
usable for real work
Hour breakdown
Weekly capacity
40.0h
What the contract says.
Operational reserve
−6.0h
Structurally unavailable.
Switching tax
−4.0h
Context-switching cost.
Asymmetry tax
−0.5h
Mixed work types.
Fragmentation
−2.6h
Meetings + interruptions.

The science underneath

⚠ Humans are not machines

Little's Law, Kingman's Formula, and Weinberg's Guideline were originally modelled on mechanical and digital systems — servers, factory lines, telecom switches — where service times are predictable and the processor doesn't degrade under load. Humans break both assumptions. Our task-completion times vary wildly (fatigue, emotion, illness, life), and we get slower per task as utilisation rises. A router at 90% processes each packet at the same speed. A person at 90% is also making more errors, taking longer to decide, and losing working memory. This tool shifts the degradation curve 10–15 points earlier than the textbook versions to reflect this reality. When you see "Strain" at 60%, that's not pessimism — it's the honest gap between silicon and grey matter.

Little's Law

Lead time = Work in Progress ÷ Throughput. When overload slows throughput, lead time doesn't creep — it lurches. A person at 90% utilisation isn't "busy," they're a queue that's starting to back up. For machines, this happens closer to 100%. For humans, the maths turns ugly much sooner.

Kingman's Formula

Waiting time rises non-linearly as utilisation climbs. The textbook curve assumes low service-time variability — which is true for machines, not for people. Human variability is high and worsens under pressure, so the "wait time explosion" that machines experience at 90–95% starts hitting knowledge workers around 75–80%.

Weinberg's Guideline

Gerald Weinberg showed that each additional project costs roughly 20% of a person's productive capacity — not from the work itself, but from the mental overhead of reloading context. Machines can run parallel threads. Human brains serialise, and each context switch carries a recovery cost that compounds.

Goldratt's Protective Capacity

Eliyahu Goldratt's Theory of Constraints established that every system needs protective capacity — a buffer that absorbs normal variation. Schedule a resource to 100% and you've eliminated the shock absorber. One hiccup cascades through the entire chain. For knowledge workers, that buffer covers email, admin, bio breaks, and the organisational friction that no spreadsheet accounts for.

Asher Krane Labs · Micro-tools for people who manage humans, not spreadsheets.
Human-adjusted model. Degradation curves shifted 10–15 points earlier than textbook queueing theory. Coefficients informed by Weinberg, Goldratt, DeMarco & Lister, and applied queueing heuristics.