The Last-Mile Governance Problem: When Public Outcomes Depend on Distributed Decisions
Governance becomes difficult when the outcome is collective but the decisions that create it are dispersed across actors who carry different costs and incentives.
Some risks can be controlled at a single asset. Others are produced by thousands of small decisions: how households heat their homes, how clinics dispose of waste, how suppliers handle data, how local teams maintain equipment, or how business units apply standards. In these systems, leadership can set policy centrally and still fail operationally because the decisive behaviour occurs at the edge.
Two 2017 studies illustrate opposite responses to this problem. Dzikuć examined urban air pollution in Zielona Góra, Poland, where emissions were linked to distributed household boilers, fuel quality and road transport. Hariz, Donmez and Sennaroglu examined healthcare-waste management in Kilifi County, Kenya, and evaluated a more centralised treatment architecture using GIS and multi-criteria decision analysis. One problem called for changing many local decisions; the other considered reducing the number of decisions that had to be governed.
Together they expose an executive question that appears in both business and government:
Should the organisation improve distributed behaviour, constrain it through standards, or redesign the system so control is concentrated in fewer places?
The Strategic Context
Distributed systems are attractive because they can be responsive, locally adapted and resilient. They are also difficult to govern when outcomes depend on equipment quality, operator capability, incentives and monitoring that vary widely across the network.
In the Zielona Góra source, the historical air-quality challenge was associated particularly with individual heating systems and older boilers using low-quality solid fuels. The paper argued that the cleaner district-heating network could reduce harmful emissions, but many households remained outside it. It also observed that financial support was directed more strongly towards public buildings and housing associations than towards owners of single-family houses, even though household investment decisions were important to the outcome.
The governance problem is clear: the city may value cleaner air, but the household may face the capital cost, disruption and fuel-price consequences of replacing a boiler. The public benefit and private decision are not automatically aligned.
The Kilifi study starts from a different configuration. The authors described weak healthcare-waste practices across a large number of facilities and considered whether a central modern incinerator could pool capital, concentrate technical capability and create a clearer monitoring point for authorities. Their GIS analysis screened land against economic, environmental and social criteria, leaving about 8.2 per cent of the county as suitable. Eight candidate sites were then ranked using AHP, VIKOR and PROMETHEE, with all three methods selecting the same leading site in the study.
The important lesson is not that central incineration is always preferable. It is that governance architecture itself is a design variable.
What Leaders Commonly Misread
The first mistake is to assume that a rule is equivalent to control. A rule may define required behaviour, but distributed actors still need the capability, equipment, incentives and supervision to comply.
The second mistake is to frame non-compliance as an attitude problem when it may be an economic design problem. In the Polish case, householders faced the costs of replacing heating equipment. If the social value of cleaner air is larger than the private return from replacement, relying on voluntary private investment may predictably underperform.
The third mistake is to centralise too quickly. Centralisation can improve standardisation and monitoring, but it creates transport dependencies, single-point capacity risks, community acceptance issues and greater consequences if the central asset fails. The Kilifi paper itself treated road access, proximity to healthcare facilities, land use, water and social considerations as part of siting precisely because centralisation moves rather than eliminates constraints.
The fourth mistake is to treat every edge location as equally risky. Distributed networks usually contain clusters. A minority of assets, households, suppliers or business units may produce most of the exposure. Uniform controls can therefore spend heavily while barely changing the system outcome.
Related article: Environmental Governance Is a Control System: Integrity, Capacity and Citizen Participation
Reframing the Issue
The last-mile governance problem can be reframed as a choice among three architectures.
Influence
Keep decisions distributed but alter incentives, information, support and feedback so local actors choose differently.
Standardise
Keep execution distributed but narrow the allowable equipment, methods, suppliers, procedures or operating ranges.
Centralise
Move the activity to fewer assets or specialist teams where capability, monitoring and investment can be concentrated.
These are not mutually exclusive. A mature program may use all three. The challenge is deciding where each belongs.
For example, a hypothetical enterprise with hundreds of sites may centralise cyber identity management, standardise endpoint configurations and still influence local user behaviour through training and reporting. A maintenance organisation may centralise specialised overhaul work, standardise preventive-maintenance procedures and allow local teams to perform routine inspections.
The strategic task is to place each decision at the level where the enterprise can govern it most effectively.
Strategic Analysis: Design the Control Architecture Around Failure Modes
Start with the outcome that cannot be compromised
Air quality and healthcare-waste control are public outcomes, but the same logic applies to safety, data integrity, quality and regulated operations. Leadership should first define the outcome whose failure creates material enterprise or societal harm.
Then identify the decisions that create that outcome. If there are thousands, the governance burden may itself be a design problem.
Examine the economics of compliance
Distributed governance becomes fragile when the actor expected to change bears the cost while another party captures much of the benefit.
This is a classic incentive mismatch. Possible responses include co-funding, taxes, pricing, technical assistance, shared services or minimum standards. The correct instrument depends on whether the barrier is affordability, information, capability or preference.
The Polish source is useful because it distinguishes between cleaner central infrastructure and the private investment decisions required to connect or change heating technology. The gap between system availability and actor adoption is the last mile.
Determine whether capability can realistically be replicated
If a process requires specialised operators, continuous emissions control, sophisticated maintenance and regulatory oversight, replicating the full capability at hundreds of small locations may be uneconomic or unreliable. Centralisation becomes more attractive when the fixed cost of competent operation is high relative to local volume.
That was part of the logic behind the Kilifi study: rather than expecting every healthcare facility to own and operate compliant incineration equipment, a central facility could pool investment and concentrate supervision.
However, the same test can point the other way. If logistics are hazardous, service interruptions are unacceptable, or local responsiveness is critical, centralisation can reduce resilience.
Use geography and network structure explicitly
The Kilifi study's use of GIS is strategically important. It converts siting from a political or intuitive choice into a spatial decision involving transport, utilities, environmental constraints and population.
Enterprise equivalents include network models for service centres, warehouses, data centres, maintenance depots and shared laboratories. The location of centralised capability changes both economics and risk.
Test robustness across decision methods
The Kenyan study compared three MCDA methods rather than assuming one ranking technique was authoritative. All three selected the same leading site, which increased confidence in that result within the assumptions of the model.
The broader principle is valuable: for irreversible or contentious decisions, leadership should test whether the choice survives reasonable changes in method, weighting and assumptions.
If the preferred option changes whenever weights move slightly, the decision is not robust enough to present as obvious.
Related article: Environmental Decisions Need Confidence Ranges, Not Just Precise Scores
Decision Framework
A practical governance-architecture test can be built around seven questions.
1. How many decision points create the outcome? A small number favours direct control. A very large number raises coordination cost.
2. Who bears the cost of compliance and who receives the benefit? Misalignment signals the need for incentive design.
3. Can the required technical capability be replicated economically? If not, centralisation or shared services become stronger options.
4. What is the consequence of local failure? High-consequence activities justify tighter standards and monitoring.
5. What new dependency would centralisation create? Test transport, capacity, outage, security and single-point-failure risks.
6. How heterogeneous are local conditions? High variation may require controlled flexibility rather than one rigid standard.
7. Is the preferred architecture robust to different evaluation methods and stakeholder priorities? If not, treat the choice as provisional.
These questions create a design choice rather than an ideological preference for centralisation or decentralisation.
From Strategy to Execution
Immediate action should map the network. Identify the actors, assets and locations whose decisions create the outcome. Quantify where failures are concentrated and what local barriers are present.
Medium-term capability building should separate problems of willingness from problems of capability. If households, suppliers or business units lack finance, solve finance. If they lack approved technology, standardise technology. If monitoring is too fragmented to be credible, consider shared services or centralised control.
Long-term strategic positioning requires embedding the architecture in capital allocation. Centralised models need logistics capacity, redundancy, specialist workforce and service-level governance. Distributed models need incentive systems, auditability, local competence and escalation paths. Hybrid models need clear interfaces so responsibility does not disappear between the centre and the edge.
Programs should also define transition states. Moving from hundreds of local assets to a shared system rarely happens instantly. During migration, the organisation may carry both old and new risks at once.
Signals to Monitor
Warning signs include persistent non-compliance despite repeated communication, low adoption of a technically available alternative, high variation in asset condition across locations, regulatory effort growing faster than risk reduction, and central facilities becoming capacity bottlenecks.
Leaders should also watch the distribution of costs and benefits. If the system depends on private actors making investments that mainly create public value, adoption may stall unless the economics are redesigned.
Another signal is the emergence of informal workarounds. When local actors bypass a standard because it is too slow, expensive or impractical, the governance model is telling leadership something about the operating reality.
Questions for the Leadership Team
- Which outcomes currently depend on thousands of small local decisions being made correctly?
- Are local actors unwilling to comply, or have we designed an economically or operationally unrealistic requirement?
- Which capabilities should be centralised because competent replication is too costly?
- What risks would centralisation introduce into logistics, resilience or community acceptance?
- Where should we standardise the method while retaining local execution?
- Would our preferred architecture remain the same if stakeholder weights or modelling assumptions changed materially?
- What transition risks arise while old and new governance models operate together?
Closing Perspective
The hardest governance problems are often not caused by the absence of policy. They arise because responsibility is dispersed across a system whose actors face different incentives, capabilities and constraints.
Leaders have three broad levers: influence decisions, constrain decisions, or reduce the number of decisions that need to be governed. The strongest architecture is the one that matches the consequence of failure, the economics of compliance and the practical capability of the network.
Governance improves when the organisation stops asking only whether the rule is clear and begins asking whether the system is designed to make the required behaviour realistically achievable.
Source References
- Dzikuć, M. (2017). Problems associated with the low emission limitation in Zielona Góra (Poland): Prospects and challenges. Journal of Cleaner Production, 166, 81–87.
- Hariz, H.A., Donmez, C.Ç. & Sennaroglu, B. (2017). Siting of a central healthcare waste incinerator using GIS-based Multi-Criteria Decision Analysis. Journal of Cleaner Production, 166, 1031–1042.
About EraNorth Insights
EraNorth Insights publishes practical analysis on strategy, projects, operations, transformation and decision intelligence for professional and organisational use. About EraNorth.