Operating model synthesis for split-authority delivery systems

2026-05-29 · governance-policy organisational-design cost-performance enterprise-adoption tools-infrastructure · synthesis medium · source → · wiki →

key claims

Governance-generated queueing, specifically approval latency, multi-party coordination overhead, and fragmented decision rights, is the dominant flow constraint in split-authority delivery systems, and adding headcount to non-constraints does not improve system-level throughputInstitute (2024)Mitchell (2026)DORA (2024)Github (n.d.)
Little's Law implies that excess Work in Progress (WIP) extends lead time and reduces predictability, so controlling the work entry rate into split-authority queues is a prerequisite for throughput improvement rather than a scheduling convenienceRepec (n.d.)Mitchell (2026)
A three-class demand segmentation that assigns standard reversible work to a fast path with post-hoc audit, routine bounded work to a standard path with lightweight pre-approval or bounded delegation, and novel or high-consequence work to an exception path with full pre-approval, prevents exception work from contaminating routine flow lanes without creating unsustainable classification overheadMitchell (2026)Mitchell (2026)
High-frequency execution decisions including daily sequencing, minor scope adjustment, local environment spend, and incident response should be delegated to delivery teams because centralised pre-approval of these decisions generates approval-queue congestion without a risk-reduction return proportionate to the coordination cost incurredDORA (2024)Mitchell (2026)Company (2022)
Pre-approval controls minimise coordination cost only when reserved for transactions with high uncertainty, high consequence, or high relationship-specific investment; applying pre-approval uniformly to all demand converts governance from a coordination aid into the primary throughput constraint and generates workaround and circumvention behaviour among teams that experience controls as coercive barriers to legitimate workUmn (n.d.)Mitchell (2026)Borys (1996)Github (n.d.)
Banking supervisory guidance requires control intensity to be proportionate to risk profile and all material responsibilities to have named owners with escalation paths, which is structurally consistent with the three-class segmentation and named-integrator requirement and sets the minimum compliance standard for the authority bundleSupervision (2015)Europa (n.d.)
An integrator without budget-recommendation rights, explicit escalation authority, and a benefits-reporting mandate remains only a coordinator and will reproduce queue-fragmentation and cost-shifting failure patterns observed across multiple completed repository items in this corpusMitchell (2026)Mitchell (2026)
Conway's Law and Team Topologies research jointly imply that authority fragmentation embeds itself in delivery-system dependency coupling, so a governance redesign that does not address team boundary alignment will face recurring re-emergence of the same coordination queues it was intended to eliminateMelconway (n.d.)Pais (2019)DORA (2024)

Research Question

What operating model improves throughput while reducing delivery risk in a split-authority environment, where "split-authority environment" means a delivery context in which authority is divided among at least two stakeholder groups with independent veto or approval power, typically a business owner, a technology delivery team, and an independent risk or compliance function?

Findings

Executive Summary

In a split-authority environment, the dominant flow constraint is governance-generated queueing rather than capacity shortage, and the operating model that most reliably improves throughput while reducing delivery risk combines three design elements: demand segmentation into risk-appropriate lanes, bounded delegation of high-frequency execution decisions to delivery teams, and a named integrator with real budget-recommendation, prioritisation, and escalation authority. The key mechanism is proportionate control intensity: pre-approval is reserved for novel or high-consequence decisions, while standard and reversible work travels fast-path lanes with post-hoc audit, and this pattern is structurally consistent with both flow theory and regulatory governance requirements. Monitoring four leading indicator families (queue health, exception volume, lead time trends, rework and incident load) enables trigger-based control regime shifts that prevent throughput collapse without requiring static control rules. The confidence level is medium because the synthesis is grounded in mechanism and direction evidence from completed repository items rather than primary cross-firm experimental comparisons of split-authority operating models.

Key Findings

Governance-generated queueing, specifically approval latency, multi-party coordination overhead, and fragmented decision rights, is the dominant flow constraint in split-authority delivery systems, and adding headcount to non-constraints does not improve system-level throughput.
Little's Law implies that excess Work in Progress (WIP) extends lead time and reduces predictability, so controlling the work entry rate into split-authority queues is a prerequisite for throughput improvement rather than a scheduling convenience.
A three-class demand segmentation that assigns standard reversible work to a fast path with post-hoc audit, routine bounded work to a standard path with lightweight pre-approval or bounded delegation, and novel or high-consequence work to an exception path with full pre-approval, prevents exception work from contaminating routine flow lanes without creating unsustainable classification overhead.
High-frequency execution decisions including daily sequencing, minor scope adjustment, local environment spend, and incident response should be delegated to delivery teams because centralised pre-approval of these decisions generates approval-queue congestion without a risk-reduction return proportionate to the coordination cost incurred.
Pre-approval controls minimise coordination cost only when reserved for transactions with high uncertainty, high consequence, or high relationship-specific investment; applying pre-approval uniformly to all demand converts governance from a coordination aid into the primary throughput constraint and generates workaround and circumvention behaviour among teams that experience controls as coercive barriers to legitimate work.
Banking supervisory guidance requires control intensity to be proportionate to risk profile and all material responsibilities to have named owners with escalation paths, which is structurally consistent with the three-class segmentation and named-integrator requirement and sets the minimum compliance standard for the authority bundle.
An integrator without budget-recommendation rights, explicit escalation authority, and a benefits-reporting mandate remains only a coordinator and will reproduce queue-fragmentation and cost-shifting failure patterns observed across multiple completed repository items in this corpus.
Conway's Law and Team Topologies research jointly imply that authority fragmentation embeds itself in delivery-system dependency coupling, so a governance redesign that does not address team boundary alignment will face recurring re-emergence of the same coordination queues it was intended to eliminate.
Four leading indicator families provide early warning of throughput destabilisation in split-authority systems: queue health as WIP age and blocked-work ratio, exception volume as the Class 3 share of total intake, lead time trends as a multi-week moving average, and rework and incident load as the share of team capacity consumed by remediation activities.
Trigger-based control regime adjustments, meaning shifts between fast-path, standard, and exception lanes governed by persistent indicator threshold breaches, are more defensible than fixed allocation rules because they respond to actual flow state rather than to a fixed schedule, consistent with the Theory of Constraints constraint-elevation step.

Assumptions

The dominant flow constraint in a specific organisation's split-authority environment is governance-generated queueing rather than raw capacity. Organisations where headcount is genuinely the bottleneck should address the capacity constraint first before redesigning governance.
The three demand classes map to work types that the organisation can distinguish at intake. If classification is infeasible due to information asymmetry at intake, the segmentation approach requires a dedicated triage step.
The named integrator can be credibly empowered with budget-recommendation rights in the organisation's governance structure. In organisations where budget authority is entirely held by a central finance committee, the integrator model would need to be adapted to whatever authority form is available.

Analysis

The causal chain in split-authority delivery runs from fragmented decision rights through approval-gate queueing to WIP accumulation, lead time extension, and throughput loss. The operating model recommendation directly targets each step in this chain: demand segmentation limits the proportion of work that enters the constrained approval gate; bounded delegation (delegation within defined guardrails with pre-agreed escalation paths) removes high-frequency low-consequence decisions from the constraint entirely; the named integrator prevents authority diffusion from recreating fragmentation; and the indicator bundle provides early warning before the chain reaches failure.

A plausible rival model is to increase the pre-approval gate's capacity by adding reviewers rather than redesigning the governance structure. The ToC evidence contradicts this as a complete solution: adding capacity to the constraint can help at the margin, but it does not address the root cause, which is that approval-by-exception (a control pattern where standard work is delegated with post-hoc audit and only exceptional work triggers pre-approval review) and demand segmentation could eliminate most of the queue rather than just widen the bottleneck.

A second rival model is to rely on stronger model-quality gates or automated testing to reduce delivery risk without changing the governance structure. This is complementary rather than rival: automated gates in the fast path are exactly how Class 1 work achieves acceptable risk at high speed. The recommendation does not exclude automated controls; it assigns them to the appropriate lane.

A third class of rival explanation treats technical quality factors — technical debt, automated testing coverage, trunk-based development, and deployment automation — as the dominant or co-primary delivery constraint. The DevOps Research and Assessment (DORA) research cited throughout this item identifies technical practices as significant contributors to delivery performance. In split-authority environments, technical quality deficits do constrain delivery, but through a different mechanism from governance queueing: technical debt and low test coverage raise the effective risk classification of individual changes, routing more work into Class 3 exception lanes and increasing the volume that requires full pre-approval. Technical quality improvement and governance structure change therefore address different links in the causal chain and are complementary rather than substitutable. The evidence base used here does not permit direct comparison of the magnitude of each effect, and the relative dominance of governance queueing over technical quality as the primary constraint has not been empirically measured across organisations.

The evidence weights toward the combined five-principle model because each element addresses a distinct failure mechanism identified across multiple completed repository items, and the elements reinforce each other: demand segmentation requires integrator authority to enforce classification discipline, and both bounded delegation and trigger-based regime shifts depend on the indicator bundle detecting threshold breaches before the flow state deteriorates to the failure point.

Risks, Gaps, and Uncertainties

Threshold values for the four leading indicator families must be calibrated empirically; the evidence base does not supply universal trigger numbers.
The Q1-Q6 dependency items have not been completed as primary research items; this synthesis uses completed repository items as proxies. A future cycle completing Q1-Q6 could validate, refine, or contradict these conclusions.
The evidence base is strongest for mechanism and direction; it does not supply cross-firm empirical comparisons of operating model variants that would support high-confidence claims about outcome magnitudes or transition timelines.
Technical quality factors (technical debt, automated testing coverage, and deployment automation) constrain delivery performance through a distinct mechanism from governance queueing; the relative magnitude of these two constraint types has not been empirically compared across organisations, and the dominant-constraint claim applies specifically to split-authority environments where governance structure is the primary source of queue formation.
The three-lane classification requires intake-stage information about work risk and reversibility. In environments where this information is not reliably available at intake, the segmentation model will misroute work and degrade throughput rather than improve it.

Open Questions

Which specific WIP-age and exception-volume thresholds best predict irreversible queue collapse in split-authority regulated enterprises?
What is the minimum viable integrator authority bundle in organisations where budget authority cannot be delegated below a central finance committee?
How does the three-lane segmentation interact with AI-assisted delivery acceleration: do AI throughput gains shift the proportion of work that reaches exception-path classification?
Should new backlog items be created for Q1-Q6 to answer the sub-questions as primary research items and validate this synthesis against empirical evidence?

Output

Type: knowledge
Description: Synthesis of flow-theory, governance, and decision-rights evidence into a five-principle operating model for split-authority delivery environments, showing that governance-generated queueing is the dominant constraint and that demand segmentation, bounded delegation, a named integrator, and trigger-based leading indicators form the minimum viable response.
Links: