The IDEO/JWT Creative Professionals Study
What it is
115 creative professionals at IDEO and JWT — eleven offices on six continents — sat for interviews about their work. Each described a project that succeeded, one that failed, and one still running: 350 projects in all, for more than 130 client companies. The transcripts were coded line by line. This site shows two readings of that corpus.
The interviews were conducted while the author was on the Harvard engineering faculty; the analysis was re-run in full in 2026.
The conditions wheel
Each mark is one of the 252 projects with a clear outcome. Its colour is the outcome the interviewee assigned — success or failure. Its length is the number of decision-makers they described. Projects are grouped by office. Click a mark to read the project in the interviewee's words, with the conditions they named: client relationship, brief stability, risk tolerance, team mood, team size. Click a tag to light up every project that shares it.
The patterns wheel
Three raters — two people and one AI model — marked every passage where one of 31 named cognitive biases appeared: 2,353 passages in all. Each spike is one bias. The part pointing outward counts passages in success stories; the part pointing inward counts passages in failure stories. When the inward part dwarfs the outward, that bias leans toward failure; when the outward dominates, it leans toward success. The coloured outer ring groups biases that tend to co-occur into six clusters. Click a spike for the definition, the counts, where in a project narrative it tends to appear, and an example passage.
How to read it honestly
These are retrospective accounts, coded. A bias that leans toward failure appeared more often when people described failures; that is a pattern in the telling, not by itself a cause. Outcomes are the interviewees' own verdicts. Two firms, eleven offices, one era of practice. Interviewees are anonymous; clients and projects appear as the interviewees named them.
Findings as first published
Research on collective intelligence (Woolley et al., 2010) showed that groups have a measurable shared intelligence that predicts performance — driven by social dynamics, not individual brilliance. The study asks what happens when that collective intelligence breaks down in real projects.
Clustering the 31 biases revealed six functional patterns — not textbook categories, but how biases co-occur in practice. These thinking patterns predicted project outcomes independently of client quality, brief stability, and team structure (p = 0.016): bad conditions are associated with failure, and so are these patterns, separately. They unfold in a recognisable cascade, with a tipping point roughly 39 % of the way through the project narrative, after which recovery becomes rare.
On the conditions side: creative project success is not produced by creative acts alone but emerges from a relational infrastructure of trust, structural clarity, and organizational containment. When the conditions are right, the creativity tends to take care of itself.
Who
Research, writing, and the current visualizations: Beth Altringer Eagle, Director of the Design Initiative at Dartmouth. An earlier version of this site (2021) was designed with Federica Fragapane and Alex Piacentini. The research and visual information design has expanded substantially since.
Cite
Questions and data requests: via ames.studio. Updated August 2026.