The chart below shows how cost accumulates as work waits. The x-axis is weeks of delay; the y-axis is accumulated cost as a proportion of the item’s total potential value.
Each profile has a different curve shape — and the shape determines how work should be sequenced. Select a profile using the buttons on the chart to see its curve and the decision rule it implies.
Expedite — pull now
Already generating a high cost per week. Pull ahead of the regular queue and work now. Keep the lane narrow — when too many items are expedited the class collapses into noise and the regular queue gains hidden competition.
Standard — sequence by rate
Constant cost per week — each week of delay costs the same as the last. Sequence by relative rate, highest first. FIFO is only correct when all items carry equal cost; in practice they don’t, and the difference compounds.
Fixed date — watch the runway
Near-zero cost until the deadline, then all value lost at once. Can safely sit lower in the queue than its apparent importance suggests — until remaining time equals remaining work, at which point it must jump the queue immediately.
Intangible — age with caution
Low visible cost for a long time. The profile can change suddenly when external conditions shift — a CVE, a deprecated dependency. When it does, the intangible becomes an expedite and disrupts everything in progress around it.
Delay has a rate
A queued item is not merely waiting. It is generating a cost. Every week it sits in the backlog is a week of value not yet captured — or a week closer to a missed deadline or a compounding technical risk.
The cost of delay is the value lost per week an item waits. Two items of identical size can have very different costs, depending on how their value decays over time. Sequencing that ignores those shapes leaves value on the table.
Expedite
Expedite items are already generating a high cost per week, and that rate continues for as long as they wait. A production incident, a blocking dependency — the cost of delay is high enough that working anything else first is the wrong trade.
Pulling an expedite item disrupts normal flow by definition — it displaces in-progress work and raises WIP temporarily. When the cost of delay on that item is higher than the disruption it causes to standard work, pulling it is the correct trade. The disruption is the lesser cost.
The risk with expedite is overuse. When too many items are expedited, the class loses its meaning and the queue fills with hidden competition. Expedite should be reserved for high-rate items — typically a small fraction of total flow, with a strict cap on how many can occupy the lane at once.
Standard
Standard items have a roughly constant cost of delay — each week of wait costs about the same as the last. Most work falls here. Sequencing it is a question of relative magnitude: a feature generating £20k of value per week should be pulled before one generating £5k, regardless of which arrived first.
Arrival-order sequencing gives correct results only when all items carry equal cost of delay.
Fixed date
Fixed-date work has almost no cost of delay right up to the deadline, then loses all value at once. A seasonal product launch, a contractual API migration — missing the date by a day carries the same cost as missing it by a month.
This profile inverts the usual urgency logic. Fixed-date items can safely sit lower in the queue than their apparent importance suggests, until the runway closes. Track remaining work against remaining time; when the margin is gone, the item must jump the queue. Prioritising it earlier displaces higher-cost work for no gain.
Intangible
Intangible items appear to have a low cost of delay. No user complains, no metric moves. The cost is real but invisible — so these items drift to the bottom of the queue and stay there.
The failure mode is a profile change. A database on an outdated major version generates a low visible cost of delay until a critical CVE is published, or a dependency update reveals the version is no longer supported. The cost spikes. The intangible becomes an expedite. The work it displaces pays the disruption cost.
Teams that never process intangibles don’t avoid that cost. They accumulate deferred items, each quietly aging, and pay a compounding penalty when any one escalates. When one escalates, the damage falls on everything it displaces.
Takeaways
- — Cost of delay is the value lost per week an item waits. Different work has different profiles.
- — Expedite items carry a high rate — pull them now and keep the lane narrow. Overusing the class destroys its meaning.
- — Standard work should be sequenced highest cost of delay first. FIFO is only correct when all rates are equal.
- — Fixed-date items can wait — until remaining time equals remaining work, then jump the queue. Prioritising them early displaces higher-cost work for no gain.
- — Intangibles look cheap to delay. Process them steadily. Left long enough, any one can change profile and become an expedite.