Design is not about choosing a visual style or following a fashionable diagram. It is about making informed decisions on what a solution must achieve, for whom, under which constraints, and with what consequences.

We often hear “human-centered design,” “service design,” “participatory design,” and “systems thinking” presented as if we had to choose just one. Not necessarily. An app may be designed with its users, incorporate accessibility criteria, coordinate a human service behind the screen, and respect engineering constraints. Each perspective addresses a different question. The job is to recognize which are needed and how to verify that they genuinely shaped the outcome.

First, separate approach, principle, process, method, and technique

These terms are often mixed, leading to unproductive debates:

Level What it guides Example
Approach or perspective The angle from which a problem is examined Human-centered design, service design
Principle A decision rule to uphold Include people excluded by an interface
Process or framework How exploration and development are organized Discover → Define → Develop → Deliver
Method How information is gathered or turned into decisions Contextual research, co-design, service blueprint
Technique or tool A specific activity or representation Sketch, stakeholder map, constraints table

Boundaries between these levels are not identical across design traditions. The distinction is practical: using a method does not prove a principle was upheld, and drawing a process does not prove that meaningful research occurred.

For example, the Design Council's Double Diamond describes four broad activities—Discover, Define, Develop, Deliver—and the movement between exploring possibilities and making focused decisions; it does not prescribe a mandatory technique or replace evaluation of solutions [R03]. Nor does it demand a strictly linear real-world process.

1. Human-centered design: understand the people who use or receive the solution

Human-centered design (HCD) begins by understanding the needs, capabilities, contexts, and experiences of those affected. For interactive systems, ISO 9241-210:2019 describes human-centered design principles and activities throughout the lifecycle [R01]. IDEO.org presents inspiration, ideation, and implementation as phases involving research and prototyping [R02]. These references do not imply that every project must purchase a standard or copy a single process.

Applying HCD means asking: What is the person trying to accomplish? What do they already know? What errors are predictable? What barriers arise under real conditions of use? Which aspects of the experience can we observe rather than guess?

Example: a restaurant owner wants visibility into bookings, but the manager needs to change them quickly during service. If we design only around the owner's opinion, we may deliver an attractive dashboard that adds work for the person using it every day.

Especially useful when: success depends on the behavior, comprehension, safety, trust, or satisfaction of specific people. What it does not prove by itself: profitability, technical feasibility, or that a small sample represents an entire population.

2. Participatory design and co-design: involve people in creating alternatives

In human-centered design we might research and observe; in co-design we invite participants to help define, propose, or evaluate alternatives. These are not opposing ideas: co-design can be practiced within a human-centered project.

Restaurant managers could sketch how booking changes ought to work, compare options, and explain conflicts in their daily activities. That gives access to situated knowledge a technical team may miss. But participation does not mean that every suggestion must be implemented, nor does it automatically transfer final decision-making authority.

Before a session, clarify which decisions remain open, who will participate, what background information they need, how disagreements will be handled, and who can select a direction. Avoid gathering people solely to endorse a solution already chosen. Where power differs—for example, employees and their managers—create conditions where concerns can be voiced without pressure.

Especially useful when: several people share a process, work practices are changing, community services are involved, or practical knowledge is distributed. Limit: participation does not remove recruitment bias or replace later tests.

3. Inclusive design and accessibility: notice who is being left out

A solution may work for its creators while excluding others. Microsoft describes three inclusive-design principles: recognize exclusion, learn from diversity, and solve for specific needs in ways that may help more people [R05]. This complements, but does not replace, checking applicable accessibility requirements.

Consider permanent and temporary abilities as well as changing situations: low vision, limited mobility, a hand injury, background noise, unreliable connectivity, limited available time, or an unfamiliar language. Do not assume all these situations share a single remedy.

Example: if a booking status relies only on color, some people may not distinguish it reliably. Adding clear text and supporting keyboard use can address different barriers. For an in-person service, an alternative help channel may matter as much as the interface.

When to use it: from the start, and especially explicitly whenever a design decision could deny access. Limit: claiming that a design is “inclusive” does not replace testing with relevant people and conditions; nor does a guideline confer certification.

4. Service design: connect experience to the operations that make it possible

Service design looks at how people, processes, channels, systems, and resources create an end-to-end experience. A customer journey map represents the user's point of view; a service blueprint adds visible actions, internal activities, support processes, and their relationships [R04]. The artifacts complement each other and do not substitute for observed data.

Example: a diner asks to change a booking through a page. The interface says “Request received.” A rule must exist behind that message: Is the new time confirmed automatically or must someone review it? Who responds if capacity is exhausted? A clear screen does not repair an undefined operational responsibility.

To apply service design, map the customer journey and, below each step, note who performs the activity, what system enables it, what evidence remains, and what happens if it fails. Highlight handoffs between people or systems. Points where nobody clearly owns a result often hide more risk than a visually imperfect screen.

Especially useful when: services span channels, people, or providers; outcomes depend on delivery as much as on contact. Limit: a blueprint documents or proposes a system; it does not prove the system works until tested.

Before creating a service blueprint, map the customer journey using evidence-based stages and friction points. Then connect that experience to internal work without confusing the two maps.

5. Systems thinking: understand dependencies and side effects

A solution can improve one task while making the overall system worse. Systems thinking examines how components affect each other, where feedback loops appear, and what delays or unintended effects may result. It is not synonymous with drawing an enormous organization chart.

Example: sending more booking notifications appears to make more information available. However, if messages overwhelm staff, they may stop paying attention; confirmation quality decreases and more incidents follow. The intervention fails not because of a button but because of the relationship among notification volume, attention, and capacity.

Start with three layers: elements (users, system, staff), relationships (what triggers what), and consequences (what might happen when a variable rises or falls). Mark possible delays, bottlenecks, and external dependencies. Treat the proposed mechanisms as hypotheses to test; a plausible diagram does not establish causation.

Especially useful when: problems recur, multiple roles interact, resources are shared, or a change may shift costs from one part to another. Limit: do not turn analysis into an infinite network that prevents choosing the next test or change.

6. Engineering- and constraint-oriented design: make the proposal feasible and safe

Human experience does not override physical laws, technical limits, or safety conditions. Solutions must consider materials, loads, response times, availability, production, data protection, interoperability, and maintenance as applicable.

Example: a food container may feel comfortable during an observation session but still require specific material and use-condition tests before delivery. A platform may be intuitive with ten users and fail under simultaneous requests. A good initial experience is no justification for overlooking those risks.

This perspective starts by identifying non-negotiable constraints, negotiable ones, and assumptions still requiring evaluation. It then supports comparing alternatives and explaining trade-offs: greater durability may increase weight or cost; automating a task may reduce human effort while increasing dependency on an integration.

Essential when: failure could cause meaningful harm, information loss, operational breakdown, or inability to manufacture or deliver. Limit: meeting technical conditions does not prove that people understand, need, or value the outcome.

The selected perspectives must become testable conditions: the requirements and acceptance criteria guide helps express what the solution must meet without reducing the experience to a feature list.

7. Use a selection matrix, not a competition between schools

Dominant situation Priority perspective Helpful complement Evidence to seek
Users do not understand a flow Human-centered Inclusive and accessible Observed tasks and specific barriers
Service fails across channels Service design Systems thinking Handoffs, incidents, and actual timing
A new procedure changes several roles' work Participatory Human-centered Alternatives built with participants and tests
Physical product faces material constraints Engineering and constraints Human-centered Technical performance and contextual use
Local change causes unexpected effects Systems thinking Service design Relationships, delays, and measured outcomes

This table is a decision aid developed for this Atlas; it does not attribute its categories to a particular standard. All of these perspectives may coexist within one project. What matters is choosing which should guide the next decision and what evidence would show whether that choice was reasonable.

8. Select approaches in seven steps

Step 1. Define the outcome and the pending decision. “Reduce confirmation errors” is a better starting point than “build a modern app.” If you still need to choose a solution, do not close exploration before comparing alternatives.

Step 2. Identify those who experience and sustain the outcome. Distinguish buyer, direct user, beneficiary, operator, and critical third parties. Mark whose perspective is still missing.

Step 3. Describe context and constraints. Physical setting, connectivity, resources, time, safety, accessibility, obligations, and capacity. Distinguish known conditions from assumptions.

Step 4. Choose the two or three perspectives most likely to prevent a bad decision. If the main risk lies in channel handoffs, prioritize service design. If users cannot understand controls, increase user research and inclusive review. This is not a ban on other perspectives.

Step 5. Turn each approach into concrete activities. Instead of “we'll do human-centered design,” write “we'll observe managers changing reservations.” Instead of “we'll use systems thinking,” write “we'll map how notification volume affects workload and confirmation.”

Step 6. Define evidence and ownership before acting. Who will participate, which task will be observed, what decision depends on the result, and which conditions would cause reconsideration? The Atlas's Pillar 3 experiment protocols cover formal study design.

Step 7. Revisit tensions and adapt. If an accessible flow requires a different technical implementation, or automation increases an operator's workload, document the trade-off, the decision, and the new test required. Approaches do not replace judgment or verification.

When the approaches reveal competing needs or trade-offs, use the comparison of solution alternatives before committing to one design. Choosing how to work does not replace deciding what to build.

Complete exercise: choosing approaches for a reservation service

Hypothetical case. Three small restaurants want fewer lost or double-confirmed bookings. The initial proposal is a digital schedule, but we do not yet know whether it is the right solution.

Decision A: what must we learn? Two managers are observed in representative work conditions: how they receive requests by phone, message, and in person; how they prevent conflicts; and what happens when times change. The human-centered perspective reveals actual work, without generalizing population rates from two cases.

Decision B: who should participate? An owner, managers from different shifts, and a customer-service employee compare options: a shared manual procedure, a digital schedule, or assisted confirmation. The co-design workshop yields options and objections; it does not automatically approve any of them.

Decision C: who may be excluded? The team checks whether the flow is understandable without relying only on color, works with a keyboard, uses clear language, and handles weak connectivity. The observations point toward further testing; they are not a claim of certified accessibility.

Decision D: how does the service work behind the scenes? A blueprint is developed: request → availability check → provisional booking → confirmation → modification → arrival. Each transition identifies who updates the record, who responds to the customer, and what happens if messaging fails.

Decision E: are there side effects? The team states a systems hypothesis: if every change sends three alerts, available attention may decline. It proposes recording notification volume, time spent, and confirmation mistakes before automating at scale.

Decision F: what can be delivered safely? Minimum technical and operational conditions are set: prevent duplicate bookings in the test system, protect relevant data, have a procedure for connectivity outages, and identify who owns exceptions. The requirements, quality, and testing articles will develop those criteria in detail.

Result: the decision is not “we'll use six methodologies.” It is a plan with justified perspectives, concrete activities, expected evidence, and unresolved decisions. The team may find that improving a shared procedure first is more valuable than immediately building an entire platform.

Common mistakes and limitations

Calling a decision “human-centered” when it relies only on internal opinions; asking for participation when nothing can change; confusing accessibility with attractive colors; treating a blueprint as proof of capacity; drawing systems relationships without observing them; positioning technical requirements as enemies of experience; choosing a framework for its reputation rather than the unanswered question.

No approach guarantees success on its own. These sources have different scopes: ISO 9241-210 concerns interactive systems, service-design guides arise from particular contexts, and inclusive-design resources offer general principles. Specific accessibility, safety, or data-protection obligations must be checked for the actual activity and jurisdiction.

What you should have when you finish

A design-approach worksheet that records: desired outcome, affected people and roles, contexts and constraints, selected perspectives and rationale, concrete activities, participants, expected evidence, decisions to be made, and unresolved problems. If it contains only the names of methodologies, it does not yet explain how you will design the solution.

References