Citizens do not experience departments. They experience life events — a birth, a bereavement, a job loss, a business registration — and every one of those events cuts across four or five agencies that have never met. That structural mismatch, not staff effort or budget, is the reason public services feel harder than they should. Design thinking is the discipline that works on the mismatch rather than on the individual counters. This guide covers what changes when human-centred design meets a public-sector context: the constraints that make it genuinely different from private-sector practice, the evidence base, the life-event operating model, and the metrics worth reporting.
The short answer: Design thinking for public service excellence means organising services around the citizen’s life events rather than the government’s org chart, grounding decisions in observed citizen behaviour rather than assumed compliance, and testing service changes cheaply before legislating or procuring them. The OECD’s most recent guidance is explicit that this is now mainstream: of 28 countries surveyed in Government at a Glance 2025, 20 are adopting service design approaches based on citizens’ life events. The hard part is not the method. It is that public services carry three constraints private services do not — universal access, no exit option for the citizen, and procurement cycles longer than the design cycle.
Why public service design is genuinely different
Practitioners moving from corporate to government work usually underestimate this, and the resulting projects fail in predictable ways. The differences are structural, not cultural.
| Dimension | Private sector | Public service |
|---|---|---|
| The user’s options | Dissatisfied customers leave | Citizens cannot exit; dissatisfaction becomes complaint, non-compliance or disengagement |
| Who counts as the user | The profitable segment | Everyone eligible, including the hardest-to-serve — who are often the highest-need |
| Definition of success | Revenue, retention, margin | Uptake among eligible population, equity of access, trust, cost to serve |
| Constraint on solutions | Commercial viability | Statute, precedent, audit, ministerial accountability |
| Speed of change | Product release cycle | Procurement and, sometimes, legislative cycle |
| Cost of failure to the user | Inconvenience, switching | Lost income, denied entitlement, legal jeopardy |
| Non-users matter | Rarely | Critically — those who should be using a service and are not are the core problem |
That last row is the one that most changes practice. In a commercial project, you research users. In a public service project, the people who never appear in your service data are frequently the most important population you have. A scheme with 90% satisfaction among applicants and 30% uptake among eligible citizens is a failing scheme.
What the evidence says
The OECD has become the most useful evidence base here, because it aggregates practice across systems rather than promoting a single vendor’s method.
Its Global Trends in Government Innovation 2024 analysed nearly 800 case studies of public service innovation from 83 countries, and concluded that a focus on people’s needs, open collaboration between service providers and users, and a human-rights framing were central to driving change in public services. The report identifies co-design with users and anticipation of future needs as a defining trend rather than an experimental one.
Government at a Glance 2025 records that OECD countries are increasingly redesigning delivery to place users at the centre, with 20 of 28 surveyed countries adopting life-event-based service design.
The gap is in what happens after measurement. The OECD’s Digital Government Outlook 2026 reports that only half of countries incorporate user experience and performance data into decision-making to improve services, and names the core challenge directly: civil servants lack incentives to act on the evidence they gather. Dashboards exist but do not drive change; reports are filed but not discussed.
That is the honest diagnosis. Most public bodies now collect citizen experience data. Far fewer have a pathway from that evidence to a decision, and building that pathway is more consequential than running more research.
The life-event operating model
The single most transferable idea in public service design is to reorganise around life events. It is worth being precise about what this does and does not mean.
A life event is a moment in a citizen’s life that triggers interaction with the state: having a child, losing a job, starting a business, moving house, a death in the family, turning 60. Each typically involves multiple agencies, sequential dependencies, and repeated submission of the same information.
What life-event design changes:
- The unit of analysis shifts from a transaction to a journey. You map the citizen’s path from trigger to resolution, including the parts that happen outside government.
- Ownership becomes explicit. Someone is accountable for the whole journey, not just their segment of it. This is usually the hardest change and is organisational rather than digital.
- Data reuse becomes a design principle. The “tell us once” standard — a citizen provides information to the state once, not once per department — is the practical test of whether a life-event redesign is real.
- Proactive delivery becomes possible. If the state already knows a citizen has become eligible, the service can initiate rather than wait for an application. This is where the largest uptake gains sit.
What it does not mean: building a new portal that aggregates links to the same unchanged services. Front-end aggregation over unchanged back-end processes is the most common failure in this field. It improves findability and leaves the actual burden untouched.
The five stages, translated for public service
The method is the same. The technique within each stage changes considerably.
| Stage | Public-service-specific practice | What goes wrong |
|---|---|---|
| Empathise | Research must deliberately include non-users, refused applicants and those who abandoned midway. Frontline staff are a primary research population, not a stakeholder group. | Research recruited from existing service users; the excluded population stays invisible |
| Define | Frame around the citizen’s outcome, not the department’s process. Name the equity dimension explicitly: who is currently worst served, and why? | Problem framed as “low digital adoption” when the real problem is that the digital route assumes a document people do not have |
| Ideate | Generate options across the full spectrum — policy change, process change, communication change, digital change. Digital is one column, not the exercise. | Ideation constrained to what the current IT contract permits |
| Prototype | Paper forms, revised letters, scripted counter conversations and Wizard-of-Oz journeys are legitimate and fast. Test in one office or one district before procuring. | Prototyping skipped because procurement rules make it feel like it must be a real system |
| Test | Measure completion and uptake, not satisfaction. Test with the hardest-to-serve segment first; if it works for them it usually works for everyone. | Piloted with digitally confident volunteers, then rolled out to a population that looks nothing like them |
The prototyping row deserves emphasis. The most valuable prototypes in public service are often a rewritten letter and a changed question on a form. These are cheap, testable within weeks, and frequently move completion rates more than a platform rebuild. Redesigning a single notice so that recipients understand what action is required, by when, and what happens if they do nothing is unglamorous work with disproportionate returns.
Designing for the whole population, not the average one
Commercial design optimises for the profitable majority. Public service cannot. This changes the research strategy in a specific way: you design from the edges inward.
Recruit for extremes deliberately. The citizen with no smartphone, no permanent address, limited literacy, a disability, a language other than the service’s default, or an irregular income is not an edge case to be handled later. They are the design constraint. A service that works for a citizen with three of those conditions works for everyone; the reverse is not true.
Treat channel choice as a right, not a cost centre. Digital-by-default is a reasonable strategy; digital-only is an exclusion policy. The measure that matters is whether an assisted route exists and whether people know it exists.
Test comprehension, not preference. Asking whether citizens like a form tells you very little. Asking them to complete it, unaided, while you observe, tells you everything. Comprehension failures cluster in exactly the populations least likely to complain.
This is the same principle we apply in accessibility-led commercial work — see our note on sustainable and inclusive UI/UX practices — but in a public context the ethical weight is different, because there is no competitor for the excluded citizen to switch to.
Metrics for public service excellence
If you report satisfaction alone, you will optimise for the experience of people already being served well. A defensible measurement set has five layers.
| Layer | Metric | What it tells you |
|---|---|---|
| Reach | Uptake as a share of the eligible population, not of applicants | Whether the service is finding the people it exists for |
| Equity | Completion rates disaggregated by district, income band, language, disability and channel | Whether the service works for everyone or only for the confident |
| Effort | Time to resolution; number of contacts required; documents requested more than once | The burden the service places on citizens |
| Effectiveness | First-time completion rate; rate of avoidable rejections; appeals overturned | Whether the process itself is generating failure |
| Trust and cost | Citizen trust measures; cost to serve per successful outcome | Whether improvement is affordable and whether it registers |
The two most diagnostic numbers on this list are avoidable rejections and documents requested more than once. Both are entirely within the department’s control, both are usually uncounted, and both are direct measures of design failure rather than citizen failure.
Six failure modes in public service design projects
- The portal reflex. Every problem receives a digital front end; the underlying process, policy and data-sharing constraints go untouched. Countermeasure: require every proposed digital solution to be accompanied by the process or policy change that makes it work.
- Consultation mistaken for co-design. A survey is issued, a public comment window opens, a decision already made is confirmed. Countermeasure: co-design means citizens influence the options, not just react to one.
- Piloting with the willing. The trial recruits digitally confident, motivated volunteers and reports excellent results. Countermeasure: specify recruitment quotas for the hardest-to-serve segments before the pilot starts.
- Procurement locking the design. A full specification is tendered before discovery, so the contract forbids learning. Countermeasure: separate a small discovery phase from build procurement, and write outcome-based rather than output-based requirements.
- Evidence with no decision pathway. Research is completed, dashboards are built, nothing changes — the failure the OECD names directly. Countermeasure: before research begins, name the decision it will inform, the person who owns that decision, and the date it will be taken.
- Political cycle mismatch. A three-year transformation is announced eighteen months before an electoral cycle ends. Countermeasure: sequence for visible citizen-facing improvements within twelve months, with structural change carried underneath them.
A 180-day sequence for a single service
Public bodies rarely have the option of a single-team pilot in the way a product organisation does. This sequence is scoped to one service within one department.
| Phase | Focus | Concrete steps |
|---|---|---|
| Days 1–45 — See the real service | Establish the current-state truth | Map the end-to-end journey from the citizen’s trigger, not the department’s intake. Interview 15–20 citizens including refused and abandoned applicants. Interview frontline staff. Count avoidable rejections and duplicate document requests. |
| Days 46–90 — Frame and prioritise | Choose the problem worth solving | Write one problem statement naming user, need and insight. Disaggregate completion data by segment to find who is worst served. Agree one measurable outcome and name the decision owner. |
| Days 91–135 — Prototype cheaply | Test before procuring | Rewrite the key letter or form. Script an alternative counter conversation. Run in one office or district for four weeks. Measure completion, not satisfaction. |
| Days 136–180 — Institutionalise | Convert a pilot into policy | Document before-and-after numbers. Take the evidence to the named decision owner. Fix the reporting metric to uptake-among-eligible. Write the discovery step into the next procurement. |
The output of 180 days is not a transformed department. It is one service that measurably works better for the people it was serving worst, and a documented internal precedent — which in public sector governance is worth considerably more than an external benchmark.
Frequently asked questions
What is design thinking in the public sector?
It is the application of human-centred design to policy and service delivery: understanding citizens’ actual behaviour and constraints through direct research, framing the problem from their outcome rather than the department’s process, generating multiple options across policy, process and technology, and testing cheaply before committing. The OECD describes this under the broader heading of human-centred public services and has been building recommendations and good-practice principles around public service design and delivery in the digital age.
How is citizen-centric government different from e-government?
E-government digitises existing processes. Citizen-centric government reorganises services around citizens’ life events and outcomes, which frequently means changing the process, the data-sharing arrangements and sometimes the policy — with digital as one instrument among several. A digitised bad process is a faster bad process.
Can design thinking work within public procurement rules?
Yes, but the sequencing has to change. The workable pattern is to procure a small, separately scoped discovery phase before specifying a build, and to write build requirements in outcome terms so that evidence gathered during delivery can change the solution. Tendering a full specification before discovery contractually prohibits learning.
How do you measure success in a public service redesign?
Uptake among the eligible population, completion rates disaggregated by segment, effort imposed on citizens, avoidable rejections, and cost per successful outcome. Satisfaction scores are the weakest measure available, because they systematically sample the people who successfully reached the end.
Who should be involved from within the department?
Frontline staff are the highest-value and most frequently omitted participants — they hold the informal knowledge of where the process actually breaks. Beyond them: the policy owner, the operations lead, a legal or compliance representative early rather than late, and someone with authority over the budget. Legal involvement at the framing stage prevents most late-stage collapses.
Does this apply to state and municipal government, or only national?
It applies particularly well at municipal level, where the distance between decision and citizen is shortest and prototypes can be run in a single ward or office. Many of the strongest documented cases in the OECD’s collection are sub-national precisely for this reason.
What about services delivered through intermediaries or agents?
Treat the intermediary as a user in their own right with their own incentives, and map the citizen’s journey through them rather than around them. Services delivered via agents frequently show excellent departmental metrics and poor citizen outcomes, because the measured transaction is the agent’s, not the citizen’s.
Where to start
If you take one thing from this guide, take the disaggregation. Publish your service’s completion rate broken down by district, by channel and by income band, and look at the worst-performing cell. That single view usually reveals more about where a service is failing than any amount of additional research, and it is almost always available in data the department already holds.
Start with the citizens your service is currently serving worst. Design from that edge inward, and the middle takes care of itself.
Humane Design works with public institutions and large regulated organisations to build citizen-centric and customer-centric capability that survives governance, procurement and scale. Explore our design thinking consulting, service design approach and systems thinking learning programmes. You may also find our writing on design thinking for social good and our case studies relevant.
