Article: Designing for Flow
Reviewed: November 2026 · Method version 2 · How these reviews work
Each factual claim in the article was checked against the source it cites, then challenged by reviewers whose job was to find the strongest evidence against it: replication attempts, meta-analyses, later papers that cite the original work, and research from other fields that study the same mechanism. A separate reviewer weighed both sides, and a script re-checked every quoted source. Where a claim was overstated or misattributed, the article was corrected. The research below supports the article as it now reads.
There is some evidence that it can. When marketing researchers studied flow among early web users, they found more of it in task-oriented activities, such as searching for information, than in browsing for fun.
The 2003 finding is confirmed from the authors’ working paper: among 588 early web users who described a flow episode, flow was reported more often in goal-directed use, such as information search and building web sites, than in experiential browsing, though it occurred in both. The study did not test administrative forms. Experience-sampling research on work points the same way (Csikszentmihalyi and LeFevre 1989), while flow at work is predicted by planning, problem solving and evaluation rather than routine activities (Nielsen and Cleal 2010).
Each person has an average level of challenge and skill, and flow is most likely when both are above that average.
A 2015 meta-analysis found the challenge-skill balance relates moderately to flow, with weaker links in work and education settings.
Earlier research had suggested recreational use, yet Novak, Hoffman and Duhachek (2003) found more evidence of flow in task-oriented activities than in experiential ones, though it occurred in both.
Confirmed from the authors’ working paper: the finding rests on 588 early web users’ own written descriptions of a flow episode, coded and clustered, so it describes where self-reported flow clustered on the 1998 web rather than a test of particular tasks.
Because the right level of challenge is personal, an interface that serves a first-time user perfectly can be flat for the hundredth visit.
Support for this point comes from the expertise reversal effect in instructional psychology.
Flow is not reserved for the game. It can belong to the settings screen and the invoice form too.
Csikszentmihalyi and LeFevre (1989) found most flow reports came during work, though people were more motivated in leisure.
Gloria Mark’s team found that people compensate for interruptions by working faster, at the price of more stress, frustration, time pressure and effort.
Mark and colleagues also found no drop in quality; the cost was stress, frustration, time pressure and effort.
Non-urgent messages should be held until the task is saved or finished.
Iqbal and Bailey (2008) found that holding notifications until a breakpoint reduced frustration.
Front-loaded feature tours are worse than starting one real task with tips shown beside the control at the moment they are needed.
Game research suggests up-front tutorials matter mostly for complex products; simpler ones are learned by doing.
These edits were made on November 12, 2026 after the review.
Claim reviewed: When marketing researchers went looking for flow on the web, they found more of it in task-oriented activities than in browsing for fun.
Now reads: There is some evidence that it can. When marketing researchers studied flow among early web users, they found more of it in task-oriented activities, such as searching for information, than in browsing for fun.
Why: The 2003 result is now confirmed word for word from the authors’ own eLab working paper (December 2001, marked forthcoming in the Journal of Consumer Psychology, archived by the Wayback Machine): they found more evidence of flow for task-oriented than for experiential activities, with flow in both. The data were 588 open-ended descriptions of flow episodes from 1998 web survey respondents, and the task-oriented cases were goal-directed uses such as searching for information and building web sites, not forms. So the sentence is right, but the lead-in about tax returns, expense reports and checkouts, and ‘the research points the other way’, goes beyond what the study tested; work research finds flow in planning and problem solving rather than routine activities. Now cites Csikszentmihalyi & LeFevre, Optimal experience in work and leisure, JPSP 56(5), 1989.
Claim reviewed: Each person has an average level of challenge and skill, and flow arrives when both are above that average.
Now reads: Each person has an average level of challenge and skill, and flow is most likely when both are above that average.
Why: The talk says exactly this. But in the experience-sampling work, ‘both above your own average’ is partly how flow was defined and coded, and a meta-analysis finds the link between challenge-skill balance and flow is moderate, and weaker in work and education settings. ‘Flow arrives when’ reads as a law; ‘flow is most likely when’ is closer to the data. Now cites Fong, Zaleski & Leach, The challenge-skill balance and antecedents of flow: a meta-analytic investigation, J Positive Psychology, 2015.
Claim reviewed: Most flow research before 2008 was correlational.
Now reads: Much of the flow research before 2008 was correlational.
Why: Keller and Bless frame their study as ‘departing from correlational research’, but they do not say ‘most’ research was correlational. Some earlier experimental work existed, so ‘most’ is a fair characterization rather than a sourced figure.
Claim reviewed: Keller and Bless built an experimental task in which demands either matched a person’s skills or did not, and showed that the match itself causes flow.
Now reads: Johannes Keller and Herbert Bless built an experimental task in which demands either matched a person’s skills or did not, and found that the match itself produced more flow in the lab.
Why: The paper does claim to document a causal effect of skills-demands fit on flow. It is two lab studies with one paradigm and self-reported flow, so ‘the match itself causes flow’ is right in direction but stated more broadly than the evidence. A meta-analysis supports a moderate link between balance and flow.
Claim reviewed: Flow belongs to the settings screen and the invoice form as much as the game.
Now reads: Flow is not reserved for the game. It can belong to the settings screen and the invoice form too.
Why: Experience-sampling research found most flow reports come from work, which supports the spirit of the line. But no study here tested settings screens or invoice forms, people are more motivated in leisure, and challenge-skill balance predicts flow less well in work settings. ‘As much as the game’ claims more than the evidence shows. Now cites Csikszentmihalyi & LeFevre, Optimal experience in work and leisure, JPSP 56(5), 1989.
Claim reviewed: Field-level validation as each field is completed, rather than errors listed after submit, gives feedback in seconds while attention is still on the field.
Now reads: After: Every error appears beside the field it belongs to, never in a list at the top of the page, and nothing is flagged while the person is still typing. Drafts save automatically. If the form runs to several steps, a short, honest count of them sets expectations, but a bar that crawls through the early steps makes people quit sooner, so show progress only when it is accurate and encouraging. (Do’s and Don’ts: replace ‘Give feedback at the field or the action, within seconds.’ with ‘Put each error beside the field it belongs to, and answer every action within seconds.’; replace ‘Show how much of a task is left.’ with ‘Give an honest count of steps, and show progress only when it is accurate and encouraging.’; replace ‘Saving all validation for the submit button.’ with ‘Listing errors at the top of the page, away from the fields they belong to.’)
Why: The only controlled experiments on validation timing (Bargas-Avila and colleagues, 2007) found that errors shown at the moment a field was left led to significantly more errors, because people in completion mode ignored them, and the same group’s open-access 2010 guideline says to show all errors after the form is sent, embedded next to the fields. Practitioner tests (Wroblewski 2009, Nielsen Norman Group) favor checking each field once it is finished and warn against checking while people type. Both sides agree on placement: errors beside the field beat a list at the top of the page (Seckler et al. 2012, n = 303). So the article’s ‘each field confirms itself’ rule is not settled, and the dependable fix is where errors appear and not interrupting typing. Now cites Seckler, Tuch, Opwis & Bargas-Avila, User-friendly locations of error messages in web forms: Put them on the right side of the erroneous input field, Interacting with Computers 24(3), 2012.