Topic 59 of 64· Software engineering and digital delivery connection
Flow Metrics in Government Delivery
Flow metrics — Little's Law, work-in-progress (WIP) limits, and flow efficiency — describe how quickly work moves through a system with limited capacity. A sprint board is one such system; a benefit-claims queue, a planning-application register, or a visa casework backlog is exactly the same mathematics wearing a different uniform.
Why it matters
Government caseloads are queueing systems, and queueing systems obey queueing laws whether or not anyone measures them. Statutory determination periods make this explicit: under the Town and Country Planning regime, most minor planning applications carry an 8-week statutory determination target and major applications 13 weeks — a cycle-time commitment baked directly into law. The Home Office's asylum casework backlog, scrutinized repeatedly by the National Audit Office and the Home Affairs Select Committee, is a well-documented case of a public system where work-in-progress grew faster than throughput for a sustained period, driving cycle times far beyond any statutory or service expectation. Flow metrics give engineers and caseworking managers alike a shared, quantitative vocabulary for exactly this failure mode, rather than leaving it as a qualitative "backlog problem".
The maths
Little's Law: WIP = Throughput × Cycle Time
→ Cycle Time = WIP / Throughput
Flow efficiency = active (touch) time / total cycle time (Vacanti)
WIP limit effect: for fixed throughput, halving WIP roughly halves
average cycle time (Little's Law rearranged) — the lever available
without adding headcount.
See DORA metrics for public value for the equivalent maths applied to software deployment pipelines rather than casework.
Worked example
Local authority planning department: 400 applications open at any time (WIP), the team resolves 50 applications/week (throughput).
Cycle time = WIP / Throughput = 400 / 50 = 8 weeks
That lands exactly at the statutory 8-week target for minor applications — with no slack, meaning any variability in incoming demand or consultee response time pushes determinations over the legal deadline.
Flow efficiency: of those 8 weeks (56 calendar days), an application typically has around 6 hours of actual caseworker processing time.
Flow efficiency = 6 hours / (56 days × 8 working hours/day)
= 6 / 448 ≈ 1.3%
Vacanti's benchmark for software teams puts typical flow efficiency at 15–20%; government casework, with multiple statutory consultee handoffs and public consultation windows, often runs an order of magnitude lower. The 98.7% of "wait" time is where the eight weeks actually goes — not in caseworker capacity.
WIP-limit intervention: capping open applications per caseworker at 15 instead of an unbounded 25 (holding throughput constant) shifts WIP from 400 to roughly 240 across a 16-person team:
New cycle time = 240 / 50 = 4.8 weeks
A near-halving of cycle time from a policy change, not a staffing increase — the same lever DORA-style delivery teams pull when they cap sprint WIP.
Software engineering connection
Flow metrics are the shared language between a delivery team's Kanban board and the casework floor it is building software for: a caseworker's queue and a pull-request queue are both governed by Little's Law, and both blow their cycle-time targets the same way — too much WIP relative to throughput. This matters directly for cost of delay in public programmes: cycle time × CoD is the pounds sitting in the queue at any moment, and it matters for service standards and transaction metrics, where a published turnaround target is a cycle-time commitment that only flow metrics can diagnose when it is missed. A casework system's software should expose WIP and cycle time as first-class operational metrics, not bury them inside a case-management system nobody queries.
Pitfalls
- Adding WIP limits without fixing the real bottleneck: if the constraint is an external statutory consultee's response time, capping caseworker WIP just moves the queue upstream rather than shortening it.
- Treating flow efficiency as a target to game: rushing the 1.3% of active time barely moves cycle time; the leverage is almost always in the wait states, which usually means process redesign, not caseworker speed.
- Ignoring variability: Little's Law describes averages; a caseload with high demand variance needs buffer capacity, not just a tighter WIP limit, or statutory deadlines will still be missed on the volatile tail even as the average improves.
- Measuring WIP inconsistently: a case "open" in the system of record but actually stalled awaiting a third party is still WIP; excluding it flatters the numbers without changing citizen-facing reality.
Sources
- Vacanti D, Actionable Agile Metrics for Predictability: An Introduction, Actionable Agile Press, 2015.
- Reinertsen DG, The Principles of Product Development Flow, Celeritas Publishing, 2009.
- Ministry of Housing, Communities and Local Government, planning application statutory timescales. https://www.gov.uk/guidance/making-an-application
- National Audit Office, reports on Home Office asylum casework and accommodation. https://www.nao.org.uk/