Process models: the diagram, the decision and the promise of a method

A process model promises how a decision will be made before anyone knows what the decision is. Three familiar models organise a design project as a sequence of stages: the Double Diamond, the engineering design cycle and the design science loop. They do not name the same stages (Design Council, n.d.; Pahl and Beitz, 1996; March and Smith, 1995).

This unit, the first of the ten, argues that a process model earns its place only when it changes a decision, and the nine units that follow take one part of that claim in turn. The Double Diamond and its relatives specify phases. They do not specify the inferential step that turns an observation into a decision. They align a team, and they do not constrain what it concludes. Three literatures pull apart here: what a method is for, the order in which designers work, and the disagreement between the diagrams.

A method is an artefact with a purpose, not a description of nature

Cross (1982) argued that design is a third area of education alongside the sciences and the humanities. Cross (2001) separated scientific design, design science and a science of design. He defended the last as a discipline of domain-independent theory and research. Its axiom is that forms of knowledge peculiar to the designer exist. A method is therefore a local instrument inside that discipline, and not the discipline itself (Cross, 2001).

March and Smith (1995) make the point checkable. Their framework separates building and evaluating from theorising and justifying, and names four research outputs: constructs, models, methods and instantiations. A process model is therefore a method, assessed as an artefact rather than as a description of what happens. The question whether the Double Diamond is true is a category error; the useful question is what it changes (March and Smith, 1995).

The Double Diamond names phases, and the diagrams do not agree among themselves

The Design Council's own page sets out four phases. Discover means understanding the problem rather than assuming it, by speaking to the people affected, including the older passenger who never answers a web form. Define means restating the challenge from that insight. Develop means seeking different answers with inspiration from elsewhere and co-design with a range of people. Deliver means testing solutions at small scale and improving the ones that work. The page describes the diagram as a way to describe the steps of any design and innovation project, irrespective of the methods and tools used (Design Council, n.d.).

A diagram that holds irrespective of the methods used cannot tell a team how to turn an observation into a decision. In 2003 the Council had no standard way to describe the process. Its then Director of Design and Innovation, Richard Eisermann, asked his team how design process should be described. The team codified an existing idea, renamed the steps and popularised it. Codifying, renaming and popularising are coordination work. The diagram's persistence is a fact about culture, and it is not yet evidence.

Pahl and Beitz (1996) show what such a model looks like when it does carry the inference. The book is a prescriptive textbook, not a study of designers, so it shows what a method can contain and not what it achieves.

Dubberly (2005) collected over one hundred descriptions of design and development processes. They range from the four-word mnemonic define, design, develop, deploy to Archer's nine-phase, 229-step systematic method for designers. He presents them together to provoke debate, and a collection ranging from 229 steps to four is not evidence for one process.

Wynn and Clarkson (2018) turn that observation into a framework. Their review relates the main families of process models to one another, so the first question is what a process model is for.

Working design is co-evolution, not sequence

Dorst and Cross (2001) tested the sequence against what designers do. They ran protocol studies of nine experienced industrial designers. The brief asked those designers to design a litter disposal system for a new Netherlands train. The authors analysed the protocols against a model in which the problem and solution spaces evolve together. They confirmed that model, and they added two spaces, the default and the surprise (Dorst and Cross, 2001). The mechanism is visible in one detail of the brief. Newspaper waste appeared on several information sheets at once, as a share of the bin contents, as litter on the luggage racks, as a complaint from the cleaners and as a recycling ambition for the railway. All nine designers combined those separate items into the same core idea, keeping newspapers apart from other litter. Each designer took it as an original insight and let it change what they took the problem to be. The same idea had also occurred to all twenty-one students in the earlier study. The move is therefore real, and its originality is not. A designer whose solution idea changes what the problem is has not failed to complete Define.

Parsons and Shukla (2026) give the finding a current base. They followed eleven expert visualisation designers through design challenges, diaries and interviews. They report framing as a continuous activity, not a preliminary step: their designers reframed the problem as solution attempts and tool feedback arrived. The inference the diagrams omit is visible in current practice, not only in nine designers from 2001.

Stompff, Smulders and Henze (2016) give the missing mechanism a name. Their in-depth video protocol study followed one experienced multidisciplinary design team. It found that reframing runs in two stages. Sensemaking reconstructs the frames already in use. Future framing then designs a frame for what comes next. Surprises incite reframing, and it takes considerable time and effort in a team (Stompff, Smulders and Henze, 2016). A phase diagram has no slot for this, so a team that reframes reads as back at the start of the first diamond.

Contested question: does a method constrain inference or align people?

The literature disagrees about what a process model is for. The first position holds that a method constrains inference. On this view a method is validated by whether it produces the support it claims. Its strongest statement is Blessing and Chakrabarti's DRM (2009). Its stages are clarification, a descriptive study of design as it is, a prescriptive study that develops design support, and a second descriptive study that evaluates it. A prescriptive method is therefore a claim, and it is validated against observation before and after the support arrives (Blessing and Chakrabarti, 2009).

The second position holds that a method aligns people, its value being the shared frame it produces. Badke-Schaub, Neumann, Lauche and Mohammed (2007) distinguish five types of mental model for design teams: task, process, team, competence and context. Their review nevertheless states the rival hypothesis. If shared understanding predicts collaborative performance rather than process, a diagram earns its place by aligning five people's frames without changing any individual inference.

What would settle it is a study that held a design problem constant, gave two teams models of different prescription, and measured the decisions reached and the shared understanding developed. Dorst and Cross (2001) studied nine designers working individually. Stompff and colleagues (2016) studied one team. Badke-Schaub and colleagues (2007) report the absence of team-level design studies. Wynn and Clarkson (2018) measure no model in use.

A second disagreement sits underneath the first. Rittel and Webber (1973) argued that policy problems are wicked rather than tame. Such problems cannot be definitively described, and they have no solutions in the sense of definitive and objective answers. Simon (1973) answered that the boundary between well-structured and ill-structured problems is vague, fluid and not susceptible to formalisation. He argued that a problem appears ill-structured to a solver with a very large but finite memory. Settling that would need a case where a better representation makes a wicked problem behave like a tame one.

The two positions rest on different evidence. The constraining side rests on DRM, a prescriptive methodology whose stages propose how design support should be validated rather than reporting one. The aligning side rests on a review that states the rival hypothesis and records a gap: no team-level design study tests it. Neither has a controlled comparison of models in use. Until one appears, the designer should treat the diagram as a coordination device and keep the inference visible, recording the options live before each gate. The strongest objection to the unit's thesis is that a model may earn its place by aligning a team even where no individual inference changes, so a test for a changed decision alone would report a real benefit as a null.

Where the argument is heading

Named work points two ways. Wynn and Clarkson (2018) treat the model literature as a portfolio indexed by purpose. Dorst and Cross (2001), Stompff and colleagues (2016) and Parsons and Shukla (2026) locate the interesting activity in what the diagrams omit. Simon (1973) makes the question conditional: if ill-structuredness is relative to a solver's memory and search, better tools shrink it, and the diagram's status shrinks with it. What the field lacks is a study that runs the DRM sequence on a process model.

The sources this unit rests on

• Cross, N. (1982). Designerly ways of knowing. Design Studies. doi:10.1016/0142-694X(82)90040-0 Cross argues that design is a third area of education beside science and the humanities, with criteria of its own. It is a position piece, not a study of practice, and it sets the unit's opening boundary: design knowledge is a kind, not applied science.
• Cross, N. (2001). Designerly ways of knowing: design discipline versus design science. Design Issues. doi:10.1162/074793601750357196 Cross separates scientific design, design science and a science of design, and defends the last as a discipline of domain-independent theory. It is a clarification by argument rather than evidence, and it draws the unit's line between a method and a discipline. Conceptual clarification.
• Rittel, H. W. J. and Webber, M. M. (1973). Dilemmas in a general theory of planning. Policy Sciences. doi:10.1007/BF01405730 Rittel and Webber argue that planning problems are wicked: they cannot be definitively formulated, have no stopping rule and admit no test of a correct solution. It is a conceptual essay naming ten properties, and it supplies the unit's objection that a phased method cannot be a checklist.
• Simon, H. A. (1973). The structure of ill structured problems. Artificial Intelligence. doi:10.1016/0004-3702(73)90011-8 Simon argues that the line between well-structured and ill-structured problems is vague and not formalisable, and that a problem looks ill-structured to a solver with large but finite memory. It is a conceptual analysis, and it gives the unit its strongest counter without a case.
• Dorst, K. and Cross, N. (2001). Creativity in the design process: co-evolution of problem-solution. Design Studies. doi:10.1016/S0142-694X(01)00009-6 Dorst and Cross ran protocol studies of nine experienced industrial designers on a litter-disposal brief and found problem and solution co-evolve, the problem being reformulated in response to solution ideas. It is the unit's empirical reason to doubt a closed Define phase. Nine designers is a small base.
• Wynn, D. C. and Clarkson, P. J. (2018). Process models in design and development. Research in Engineering Design. doi:10.1007/s00163-017-0262-7 Wynn and Clarkson review the design and development process-model literature and organise the families by purpose, context and limits. It is a review measuring no model in use, and it shows the models describe different things, so a diagram is a purpose-specific instrument.
• Blessing, L. T. M. and Chakrabarti, A. (2009). DRM, a Design Research Methodology. Springer. doi:10.1007/978-1-84882-587-1 Blessing and Chakrabarti set out DRM: clarification, a descriptive study of design as it is, a prescriptive study that develops support, and a study that evaluates it. It is a methodology, not a study of practice, and it supplies the unit's test for a method claim. Only the overview was read.
• March, S. T. and Smith, G. F. (1995). Design and natural science research on information technology. Decision Support Systems. doi:10.1016/0167-9236(94)00041-2 March and Smith separate design research from natural science, dividing building and evaluating from theorising and justifying, and name four outputs: constructs, models, methods and instantiations. It is a conceptual taxonomy, and it lets the unit assess a process model as an artefact rather than a description.
• Pahl, G. and Beitz, W. (1996). Engineering Design: A Systematic Approach. Springer. doi:10.1007/978-1-4471-3581-4 Pahl and Beitz present systematic engineering design as a staged procedure from task clarification to detail design, with prescribed methods at each stage. It is a prescriptive textbook, not a study, and it shows what a process model contains when it carries the inference.
• Badke-Schaub, P., Neumann, A., Lauche, K. and Mohammed, S. (2007). Mental models in design teams: a valid approach to performance in design collaboration? CoDesign. doi:10.1080/15710880601170768 Badke-Schaub and colleagues review the mental-model literature and argue that shared team mental models, not the process followed, may predict collaborative design performance. It is a review stating the rival hypothesis and measuring nothing, and it supplies the unit's second position.
• Design Council (n.d.). The Double Diamond. Design Council, resource page. https://www.designcouncil.org.uk/our-resources/the-double-diamond/ The Design Council's page sets out the Double Diamond's four phases and records that its team codified, renamed and popularised an earlier idea in 2003. It is a first-party page and the artefact the unit critiques, and it offers no evaluation of the method.
• Dubberly, H. (2005). How do you design? Dubberly Design Office. https://www.dubberly.com/articles/how-do-you-design.html Dubberly assembles over a hundred process descriptions, from a four-word mnemonic to a 229-step method, and presents them together to provoke debate. It is a collection rather than a study, and it shows the diagrams disagree, so one diagram is a choice among many.
• Stompff, G., Smulders, F. and Henze, L. (2016). Surprises are the benefits: reframing in multidisciplinary design teams. Design Studies. doi:10.1016/j.destud.2016.09.004 Stompff, Smulders and Henze video-analysed one experienced multidisciplinary design team and found reframing runs in two stages, sensemaking then future framing, incited by surprises. It names the mechanism a phase model omits. One team makes the finding suggestive, not general.
• Parsons, P. C. and Shukla, P. C. (2026). Beyond problem solving: framing and problem-solution co-evolution in data visualization design. IEEE Transactions on Visualization and Computer Graphics. doi:10.1109/TVCG.2025.3633866 Parsons and Shukla followed eleven expert designers through design challenges, diaries and interviews and found framing continuous rather than a preliminary step. It gives the unit a current base for co-evolution. Eleven experts, one domain.