Design Thinking Explained: The Five Stages With Examples

The five stages of design thinking were never meant as a fixed sequence. The 2010 process guide from Stanford’s Hasso Plattner Institute of Design, the d.school, sets out the five stages in a line only for simplicity and says design challenges can be tackled by using them in various orders.¹ Design thinking is an approach to solving problems for people: learn how they really live and work, frame the real problem, generate many ideas, then build and test rough versions with them before committing. The five stages are:

1. Empathize: watch and talk with the people you are designing for.
2. Define: state one problem worth solving, from their point of view.
3. Ideate: generate many possible solutions before judging any.
4. Prototype: build quick, cheap versions that each answer a question.
5. Test: put those versions in people’s hands and learn from what happens.

It is one structured route to what psychologists call creative work, ideas that are both new and useful, as our guide to how psychologists define and measure creativity explains. Each stage below comes with the study that tests the idea behind it and one running example: a small company whose IT help-request form nobody uses.

Where design thinking came from, and what it is

Design thinking as businesses use it today grew largely out of product design teaching at Stanford University. A 2021 historical review by Jan Auernhammer and Bernard Roth traces it to a joint product design program founded there in 1959, to the design firm IDEO, which two of the program’s longtime teachers, David Kelley and Bill Moggridge, co-founded with Mike Nuttall in 1991, and to the d.school, which the SAP co-founder Hasso Plattner funded in 2005.²

There is still no generally accepted definition. A 2019 systematic reviewsystematic review: A review that fixes its question and its rules for including studies in advance, then searches out every study that fits and weighs them together. Some systematic reviews pool the results into a meta-analysis; others describe what the studies found without combining the numbers.Full entry in the glossary by Pietro Micheli and colleagues found writers treating design thinking as a trait of individuals, a set of tools, an attribute of organizations or a culture. The process models the literature cites most do share a shape, though: explore the problem, generate alternatives, then prototype and test in repeated cycles.³

The order stays loose because each stage produces information that can send you back. The d.school guide notes that testing sometimes reveals not only a weak solution but a problem that was framed wrongly, which means returning to the define stage.¹

1. 1Empathizeobserve and talk with people
2. 2Defineone user, one need, one insight
3. 3Ideatego wide, judge later
4. 4Prototypecheap versions that answer a question
5. 5Testwatch people use them, then loop back

Then repeat from “Empathize”

In the help-form project, for instance, a test might show that people never open the form at all, which sends the team back to watching how they ask for help. The lesson: treat the stages as five kinds of work a project needs, not five boxes to tick in order. Going back to redefine the problem after a test is the method working, not failing.

Further reading

• 

  Sketching User Experiences: Getting the Design Right and the Right Design

  by Bill Buxton

  Buxton, a co-author of the test-many-designs study cited here, on sketching several rough alternatives before committing to one.

• 

  The Design of Everyday Things

  by Don Norman

  Norman's classic on designing for how people really behave, with a chapter on design thinking and the observe, ideate, prototype, test loop.

As an Amazon Associate WiserHours earns from qualifying purchases.

Stage 1, empathize: watch how people really do the task

The empathize stage means learning first-hand how the people you are designing for do things, and why. The d.school guide recommends observing people in their own setting as well as interviewing them, because some of the most powerful realizations come from a gap between what someone says and what they do, or from a work-around they would not think to mention.¹

Why not just ask, or trust what you already know? Because your own habits are a poor guide to other people’s. In four studies published in 1977, Lee Ross and colleagues found that people tend to see their own responses as relatively common, a bias they named the false consensus effect.⁴ Applied to design, the risk is plain: a team that finds its own tool easy may assume its users do too. False consensus is one of many predictable errors in judgment; the guide to how cognitive biases shape everyday judgment explains which of them hold up when retested.

Take the running example. Ask the IT team why the help form goes unused and you may hear that people are lazy about forms. Sit with a few colleagues while they report a real problem and you may see something else: the form asks for an asset tag nobody can find, so they give up and message the IT person directly. That work-around is the finding.

Stage 2, define: turn what you saw into one problem worth solving

The define stage turns scattered observations into one specific problem statement. The d.school guide calls it a point of view, a sentence that combines a user, a need and an insight, and argues that a narrowly focused statement tends to produce more and better ideas than a broad one.¹

Research on problem finding, the skill of discovering and framing problems, is consistent with giving this stage real time, with a caution. A 2020 meta-analysismeta-analysis: A study that combines the results of earlier studies on the same question into one overall estimate. Pooling makes the estimate more precise, but it cannot repair the studies it pools: a meta-analysis of surveys is still survey evidence.Full entry in the glossary of 40 studies by Ahmed Abdulla and colleagues reported that higher problem-finding scores went with higher scores on creativity measures, but the link was modest and varied widely between studies.⁵ Because it is a correlationcorrelation: A measure of how closely two things move together, running from minus one, where one rises as the other falls, through zero, meaning no link, to plus one. It says how strong the relationship is, not what causes it, and it is not a percentage.Full entry in the glossary, it cannot tell you whether better framing causes better ideas.

For the help form, “Improve the IT form” is a weak statement. A point of view is sharper: “Staff reporting a broken laptop need to ask for help in under a minute, because hunting for an asset tag makes them give up.” The second version tells you which ideas to reject.

What to do: write the point of view as one sentence with a person, a need and a “because”. If it could apply to any team in any company, narrow it. The guide suggests turning it into “How might we…?” questions to start the next stage.¹

Stage 3, ideate: go wide before you judge

Ideation is the stage for generating many possible solutions and deferring judgment until later. The d.school guide says the aim is the broadest range of possibilities rather than the right idea, and suggests carrying two or three ideas into prototyping instead of settling on the one most of the team can agree on.¹

The main risk at this stage is fixation, getting stuck on the first solution you see. In a 1991 series of four experiments, David Jansson and Steven Smith gave engineering students a design problem and showed half of them an example solution. Those who saw the example copied its features, including its flaws. Asked to design a spill-proof coffee cup with no straw or mouthpiece, students shown a leaky example with a straw and a mouthpiece put a mouthpiece in 39 percent of their designs, against 10 percent for students who saw no example. Just over a dozen professional engineers showed the same pattern on a different task.⁶ The groups were small, so read this as a demonstration rather than a measurement.

In the help-form project, if someone opens the session by showing the ticketing tool a rival company uses, expect many of the ideas to look like it.

The practical fix: have everyone write ideas alone before anyone shows an example, a competitor’s product or last year’s solution. Then pick ideas with more than one test, such as the guide’s suggestions of the most likely to delight, the rational choice and the most unexpected, and keep more than one alive.¹

Stage 4, prototype: build several rough versions at once

A prototype is anything a user can interact with, from a wall of sticky notes to a role-play or a storyboard, built to answer a question. The d.school guide advises starting with low-resolution versions that cost minutes and cents, and letting go of each one before you become attached to it.¹

A Stanford experiment tested how many prototypes to make before asking for feedback.

The study

Limited evidence

Five practice ads, two feedback schedules: a Stanford experiment, 2010

Stanford researchers randomly assigned novice designers to make five prototype web ads and then a final ad, with equal time and the same kind of written critique. One group got feedback after each prototype; the other made three, got feedback on all three, then made two more. In a real online campaign, the parallel group’s final ads drew about 12 percent more clicks per view and were rated higher by the client and ad professionals. Their prototypes were more varied, and they reported a larger gain in confidence.⁷

Why would timing matter when the amount of work and feedback was the same? The authors offer three explanations: comparing versions side by side helped people learn what works, it kept them exploring instead of polishing one idea, and the serial group read the critique as negative and so gained no confidence in ad design. Nearly half of that group described the feedback as negative; nobody in the parallel group did.⁷

The caveat: this is one small study, funded by the Hasso Plattner Design Thinking Research Program, and in an earlier technical report the same team saw no clear gap between the two schedules on a simple mechanical task, building a vessel to protect a dropped egg.⁸ The benefit may depend on the kind of task.

1. One at a time (serial): feedback after each of five prototypes, then a final design
2. Several at once (parallel): three prototypes, feedback on all three, two more, feedback again, then a final design

For the help form, three prototypes might be a paper form without the asset-tag field, a chat command that files the request, and a sticker on each laptop with a code that fills in the details. Each can be mocked up in an afternoon.

Stage 5, test: show people options, then watch

The test stage puts prototypes in people’s hands to learn about both the solution and the user. The d.school guide advises showing rather than explaining, watching how people use and misuse what you give them, and bringing several prototypes so that people have something to compare.¹

Comparison changes what people tell you. In a 2006 experiment at the University of Toronto, Maryam Tohidi and colleagues had students and recent graduates test paper prototypes of a home climate-control screen, either one design alone or three different designs. Those who saw one design rated it higher and were more reluctant to criticize it than those who saw the same design among three.⁹ The authors cite earlier usability testers who saw participants hold back criticism and guessed at reasons, such as not wanting to hurt the tester’s feelings.

Even with three designs to compare, users were poor at suggesting improvements: testing, the authors concluded, is a way to find problems, not to get solutions.⁹

always prototype as if you know you’re right, but test as if you know you’re wrong

For the help form, put each version in front of a colleague who has a real problem to report, say nothing, and watch where they hesitate. When you run the test, ask people to complete a task rather than rate a design, show more than one version, and treat complaints as data and suggested fixes as clues. The loop of predicting, trying small and learning will feel familiar if you have used the plan-do-study-act cycle from process improvement.

How much evidence stands behind design thinking

The evidence for design thinking as a whole is thinner than its popularity suggests, while several of its individual practices have experimental support. Micheli and colleagues’ review found that most articles on design thinking discussed examples or reported single case studies, and that very few of the studies were quantitative.³

A later systematic review by Selina Mayer and Martin Schwemmle, published online in 2024, included 69 articles. Its authors describe the explanations of how design thinking produces its effects as scattered and inadequately grounded in existing research.¹⁰ The studies in this guide test the parts, not the package, mostly with small groups of students on short tasks.

For you, that means treating design thinking as a set of habits with some support behind them, not a program with proven results. If someone proposes rolling it out across a company, ask which outcome will be measured, and try it first on one contained problem, such as a form nobody uses.

The bottom line

Design thinking is five kinds of work rather than five steps in a fixed order: see how people really do the task, name one problem, go wide, build rough versions and watch people use them. The practices with experimental support in this guide all resist committing too early: writing ideas before looking at examples, making several prototypes before asking for feedback, and testing alternatives side by side. Evidence that the whole method makes organizations more innovative is still thin, so run it as a small experiment of your own.