Double Diamond
Separate finding the problem from solving it, and do each in two movements.
At a glance
The problem it solves
Most requests arrive already shaped as a solution, whether that's a mobile app, a restructure of support, or a new onboarding flow. Work usually starts there, because starting there is what everyone is being paid to do, and because questioning the brief looks like reluctance.
What follows is usually competent, and that's the uncomfortable part. Teams ship a decent piece of work, on time, answering a problem nobody checked, and the failure stays invisible for months before it shows up as a thing nobody uses.
The opposite failure is just as common and gets less criticism because it looks like rigour. A team that does go and look never stops looking, so research runs for a quarter, each finding is interesting, nothing converges, and the eventual decision gets made by whoever is least tired in the last meeting.
Both failures are the same missing thing. There's no rhythm, no point at which the group agrees it has stopped opening things up and started closing them down, so opening and closing happen simultaneously, in the same conversations, where closing wins.
The idea
Two diamonds, each one a widening followed by a narrowing, with the first diamond about the problem and the second about the solution.
Two diamonds side by side, joined by a small labelled band at the waist. Each widens and then narrows, the width standing for how many options are in play. The four halves read left to right as Discover, Define, Develop and Deliver, so the first diamond covers the problem and the second covers the solution. The band between them is the written problem statement.
Discover widens, which means going and looking at the situation rather than at the brief. Define narrows, which means committing to one written statement of the problem you're actually solving. Develop widens again, generating genuinely different ways of answering that statement. Deliver narrows, so you choose one, build it and ship it.
The four names are the least important part, and two rules do the work.
Never widen and narrow at the same time. A half-formed idea loses to a fully formed objection, so a room that's generating and judging at once converges on whatever was said early and confidently. That's why the shape is drawn as movements rather than as a list of stages.
The waist is a deliverable. It's a sentence somebody wrote down and other people agreed to, rather than a transition or a milestone, and everything in the second diamond is judged against it. If you can't write it, you're still in the first diamond, whatever the plan says.
The Design Council drew this in 2005, from a study of how design teams in large organisations already worked. What makes it more than a picture of a process is the claim buried in the symmetry, which is that the problem deserves as much structured attention as the solution. Most organisations give it a paragraph at the top of a document.
How to use it
Say which half you're in, out loud, at the start of each session. Most of the value arrives from that one sentence, because it gives everyone in the room permission to behave differently, and it gives anyone the standing to say "we're supposed to be widening" without it sounding like an objection to the idea on the table.
In Discover, look at what people do rather than at what they say they do, and talk to the people with the problem rather than about them. The purpose isn't to collect requirements but to find out whether the brief describes anything real.
In Define, produce one sentence, written, shared, argued over and agreed. This is the point at which the team is allowed, and really required, to say the original request was the wrong problem, and the sentence is what makes that sayable rather than insubordinate. If the statement is the brief restated, the first diamond didn't happen.
In Develop, generate more than one genuinely different answer, because the reliable tell of a counterfeit second diamond is a single option presented with three variations in it.
Timebox each half, and put the tighter box on the narrowing halves, because divergence is self-justifying, there's always another interview and no interview ever proves you have enough, so Discover will eat the project unless Define has a date and an owner.
Where it breaks down
The evidence is a description rather than a finding. It came out of observing what design teams appeared to do, and it's a good description. There's no controlled evidence that following it produces better outcomes than not following it, and there's unlikely ever to be any. It's a shared vocabulary and a meeting discipline, and it's regularly presented as though it were a validated method, which it isn't.
Real work loops and the diagram has no arrow for that. Teams reach the second diamond and find the problem statement was wrong, which is a success and the process working, but the shape says a single pass, so teams either force on with a statement they no longer believe or conclude they did the first diamond badly.
The symmetry is fiction. Two equal diamonds imply two equal efforts, when in real work one usually dominates. A well-understood problem with an instrumented cause needs an afternoon in the first diamond, and a genuinely novel problem can make the second nearly trivial once the first is done properly. The picture gives no help in telling which kind you have, and process templates built on it routinely mandate equal time in both.
For wicked problems the central claim is simply false. Rittel and Webber's argument is that for problems with interacting causes and no stopping rule you can't finish defining before you begin solving, because attempted solutions are how the problem becomes visible. That covers most organisational and social problems, which is to say most of the interesting ones. The rhythm still helps, the sequencing doesn't, and the honest adaptation of holding the statement loosely and expecting to rewrite it is the one thing the diagram can't express.
It's a commercial artefact as well as a conceptual one. The most common real-world use of the first diamond is a consultancy justifying a paid discovery phase, and the shape is unusually convenient for that, because it makes a long, low-output period look like rigour rather than delay. Being sold something isn't an argument against it, though it does explain why the model is far more popular than its evidence.
Discovery theatre passes for discovery. The workshops run, the walls fill with notes, the photographs are taken, and the problem statement is the one the sponsor arrived with. Widening is easy to perform and hard to actually do, and nothing in the model detects the difference.
It tells you nothing about what to do inside each half. All four names are containers, and the difficulty of the work sits entirely in the parts the diagram doesn't cover, which is how to frame a problem and how to generate an option that's genuinely different rather than adjacent.
In one line
Widen then narrow, twice, once to find the problem worth solving and once to solve it, and never do both at the same time.
References
- Primary
The Double Diamond (opens in a new tab)
The originator's own account, linked rather than restated. It's worth reading first-hand for how modest the original claim is, because it was drawn to describe how design teams were already working rather than to prescribe a process for everybody else.
- Further
The core of 'design thinking' and its application (opens in a new tab)
Goes underneath the diagram to the reasoning move design actually makes, which is working from a desired outcome back to a way of framing the problem. This is the content the four phase names leave out, and it's the part that's hard.
- Critique
Dilemmas in a General Theory of Planning (opens in a new tab)
Where wicked problems were named, and the sharpest objection available to this model. For problems of that kind you can't finish defining before you start solving, because attempted solutions are how the problem becomes visible at all.