Action deepens knowledge.

A prototype can behave impeccably in the room where it was designed and become considerably less cooperative after meeting humidity, production speed, human hands and a night shift with targets to hit.

Consider an illustrative packaging-development case. Sofia Benali leads structural packaging for a consumer-goods company preparing a new premium refill pack. On the design bench, the locking tab closes cleanly. Compression tests fall within specification, the board grade is approved and the sample survives the presentation to senior management with the polished composure expected from an object that has never met a loading bay.

Two weeks later, Chinedu Okafor calls from the production floor at 2.16 a.m.

The tab has started jamming.

The failures cluster around stock stored closest to the loading doors, and they appear mainly when the night shift runs the line at full commercial speed. Operators have already developed a workaround involving a slight manual bend before feeding each pack into the machine. It keeps output moving and adds several hundred tiny human corrections to a process designed to require none.

Sofia's original knowledge remains technically grounded. She understood the material, dimensions and locking geometry well enough to create a functioning design. Action has now produced information the workbench could not supply: moisture exposure, machine speed and handling behaviour interact with the design in ways the prototype conditions failed to reveal.

Within the Unified Skills Map, Practical Skills can deepen Knowledge Domains when enacted work exposes understanding to real constraints, consequences and context. The experience becomes developmentally useful when somebody detects the discrepancy, interprets it and revises the knowledge carried into later action. Our framework preserves that condition because repeated experience can also stabilise a weak explanation, a local superstition or an efficient workaround that quietly protects the original mistake.

The Workbench Cannot Reproduce the Whole Job

Professional action generates observations that remain unavailable during abstract preparation. A negotiation reveals which interests become rigid under pressure. A software deployment exposes interactions with legacy systems. A sales call makes a carefully designed value proposition collide with the customer's actual objection, which occasionally has the discourtesy to be absent from the persona deck.

For Sofia, production reveals two kinds of knowledge.

Her Conceptual Knowledge changes as the locking mechanism becomes linked to moisture absorption and line speed. The board can meet specification under controlled conditions while behaving differently after environmental exposure changes its stiffness and dimensions. The original model acquires another condition.

Contextual Knowledge develops through the production system surrounding that mechanism. Sofia learns where board waits before use, how long doors remain open during unloading, which shift runs closest to maximum speed and how operators detect an emerging jam before the machine stops. She also learns why the workaround appeared so quickly: the team is measured heavily on throughput, while minor manual interventions disappear inside the output number.

Workplace-learning research has long described professional development through participation in real activity. Eraut (2004) documented the informal and often unplanned character of much workplace learning, including learning through challenge, consultation, observation and problem solving. Recent evidence reinforces the importance of the surrounding environment. Wijga, Beausaert and Kyndt's (2025) systematic review of 73 studies identifies both individual and contextual antecedents associated with participation in workplace learning, including task characteristics, autonomy, organisational support and learning culture.

Those findings support a conditional account of experience. Work creates access to evidence, yet access varies sharply across roles and organisations. Sofia can learn from the night shift because Chinedu contacts her, the line can be observed and the problem is still visible. A designer who receives only a monthly defect percentage inherits a very different learning opportunity.

An Event Becomes Evidence Through Interpretation

A jam tells the team that performance departed from expectation under particular conditions. It leaves the cause unresolved.

The first explanation sounds reasonable: the locking tab is too weak. Chinedu is less convinced because the same stock ran successfully during the previous afternoon. Mei-Ling Chen, the quality engineer, pulls humidity records and discovers that the affected pallets spent several hours near an open loading door during unusually damp weather. The team now has a competing explanation involving material condition, machine speed and storage exposure.

Research on learning from errors helps clarify the conversion. Zhang and Fiorella (2023) propose an integrated model in which error generation can activate prior knowledge and create opportunities for learning when the learner detects the error, processes corrective information and updates understanding. The error supplies a discrepancy. Development depends on the work performed around that discrepancy.

Narciss and Alemdag's (2025) synthesis reaches a compatible conclusion while showing how many variables remain involved. Across the 11 papers in the special issue they review, learning from errors and failure is shaped by contextual conditions, motivational and emotional factors, processing of the error and forms of instructional support. The authors also identify conceptual and methodological gaps that limit any universal account of failure as a developmental mechanism.

The packaging team therefore has several possible lessons available from one machine stop. Sofia could enlarge the tab. Chinedu could slow the line permanently. Mei-Ling could revise storage controls. Each response contains an implied causal account, and the quality of later knowledge depends on which account survives comparison with the evidence.

A workplace can accumulate experience rapidly while learning very little when consequences are converted immediately into fixes and nobody investigates the relationship that produced them.

Feedback Needs Enough Information to Revise the Model

Production has already given Sofia outcome feedback: some packs jammed under commercial conditions. That signal identifies a performance gap while leaving the mechanism open.

Useful feedback becomes more diagnostic when it connects the observed result with the action, conditions and assumptions that preceded it. Wisniewski, Zierer and Hattie's (2020) meta-analysis found an overall positive effect of feedback on learning alongside substantial heterogeneity. Their synthesis differentiates feedback focused on task, process and self-regulation from feedback directed at the person, with effects varying according to the information carried and the learning context.

Earlier meta-analytic work by Kluger and DeNisi (1996) demonstrated how unreliable the blanket instruction to "give feedback" can become. Feedback interventions improved performance on average, while a substantial proportion produced negative effects. Attention can be redirected away from the task, and information with little diagnostic value can strengthen an unhelpful response.

On the packaging line, a manager saying "the design needs to be tougher" sounds decisive and provides Sofia with almost no mechanism. Chinedu's observation that failures rise after damp storage and at full line speed gives the team something testable. Mei-Ling can then compare board moisture, dimensional change and jam frequency across controlled trials.

The same issue appears in commercial work. Imagine Daniela Ruiz, a sales director reviewing a failed enterprise proposal. The client chose a competitor, which provides a clear outcome and a poor diagnosis. If the internal review concludes that price caused the loss because procurement mentioned cost during the final call, the team may discount the next proposal and learn an expensive fiction. Interviewing the client might reveal that implementation risk, governance and integration support carried greater weight than price.

Action produces consequences that become developmentally useful when feedback connects them to a sufficiently accurate account of what generated them.

Reflection Can Reconstruct What Happened

Sofia returns to the line the following afternoon with Chinedu and Mei-Ling. They reconstruct the sequence from pallet arrival to machine stoppage, then compare the affected stock with material stored farther from the doors. The discussion begins with a practical question about the jam and gradually changes the design team's understanding of the operating environment.

Schön's (1983) account of reflective practice describes professional inquiry arising when a situation resists the practitioner's existing framing. Surprise prompts examination of the assumptions embedded in action, and the practitioner experiments with a revised understanding of the situation.

Contemporary workplace research gives reflection a more differentiated structure. Hommel, Fürstenau and Mulder (2023) propose a conceptual model in which triggers, individual and social processes, outcomes and contextual conditions interact during reflection at work. Their empirical exploration focuses on vocational teachers, so the model offers conceptual guidance for other professions while remaining bounded by its study context.

Structured debriefing provides another relevant evidence base. Tannenbaum and Cerasoli's (2013) meta-analysis found that individual and team debriefs were associated with improved subsequent performance across the studies they examined, with effects varying according to design and context.

For the packaging team, useful reflection stays close to observable events by comparing which pallets failed, the moisture conditions surrounding those failures, the changes that appeared at higher line speed and the adjustments that reduced the jam without creating another defect. The team can then test a revised explanation instead of writing a polished paragraph about "key learnings" that everybody approves before returning to exactly the same process.

Professional reflection loses value when the organisation rewards the appearance of insight more consistently than actual revision. A beautifully formatted lessons-learned document can become the administrative equivalent of placing a small plaque beside the problem.

Workplaces Decide Which Lessons Travel Back

Experience is distributed unevenly across an organisation. The person who designs a process may never see its most informative consequences, while the people who encounter those consequences may lack the authority, language or channel required to change the underlying model.

Chinedu's team has already learned something before Sofia reaches the factory. Operators know that a slight bend reduces jams, and they have learned to hear a change in the machine rhythm before the stoppage occurs. That knowledge is practical and partly tacit. Its value depends on whether the pattern reflects a stable relationship and whether it can reach the people responsible for design and process control.

Cerasoli et al.'s (2018) meta-analysis found informal learning behaviours associated with knowledge, skill, attitudes and performance, while individual and situational factors also predicted engagement in those behaviours. The correlational nature of much of this evidence leaves causal direction open: stronger performers may seek learning opportunities, and supportive environments may influence both learning behaviour and outcomes.

Wijga, Beausaert and Kyndt (2025) similarly identify job and organisational characteristics among the antecedents associated with workplace learning. Access to challenge, autonomy, managerial support and a climate that permits learning can determine whether experience becomes visible enough to examine.

This has direct implications for business systems. A product team can remain ignorant of customer workarounds when support data never reaches product management. Finance can continue using an assumption after operations has discovered why it fails locally. A strategy team can celebrate a successful launch while customer service quietly acquires an advanced education in everything the launch deck forgot.

Practical action generates knowledge opportunities across the organisation, and organisational design influences whether those opportunities travel back to the people who can interpret them.

Delayed Consequences Distort the Lesson

The packaging jam provides unusually generous feedback because the machine stops within seconds. Many professional actions produce consequences across weeks, months or departments.

A pricing decision can improve quarterly revenue while weakening renewal behaviour later. A procurement shortcut may appear efficient until quality failures emerge downstream. A hiring decision can look successful during onboarding and create capability gaps much later. The original decision maker may never see the full consequence.

Delayed feedback complicates causal attribution because intervening events multiply. A favourable result can be credited to the wrong action, while an unfavourable result can be blamed on the most visible event. Organisations therefore create mechanisms such as audit, incident review, customer research and longitudinal performance analysis to reconnect action with later evidence. Those mechanisms inherit their own measurement assumptions and reporting incentives.

Deliberate-practice research illustrates why the quality of information around activity matters. Macnamara, Hambrick and Oswald's (2014) meta-analysis found considerable variation in how much performance variance deliberate practice explained across domains, with smaller contributions in professions and education than in highly structured domains such as games and music. Professional environments often provide noisy outcomes, mixed objectives and limited opportunities to repeat the same task under controlled conditions.

The implication for Knowledge development is precise. Experience supplies observations from enacted work. Their developmental value depends on whether the practitioner can connect those observations to a plausible mechanism, compare cases and revise the representation carried into later action.

The Next Version Carries the Lesson Forward

Sofia's team eventually runs controlled trials at different board-moisture levels and line speeds. The revised design changes the tab tolerance slightly, while Operations changes storage guidance for incoming board. Chinedu helps define an early-warning cue that operators can recognise before the machine reaches a full stoppage, and Mei-Ling adds moisture condition to the trial record so future failures can be interpreted against something stronger than memory.

The next prototype retains almost the same appearance on the design bench, while its specifications now carry knowledge produced by the line itself. Conceptual Knowledge includes an interaction between material condition, geometry and production speed that the original tests failed to capture. Contextual Knowledge includes storage exposure, shift practices, production incentives and the sensory cues operators use during actual work. Those changes arose because Practical Skills placed the original understanding into contact with the operating environment and produced evidence worth investigating.

Our architectural contribution is to make that directional relationship explicit while preserving its conditions. Action can deepen Knowledge Domains through the constraints, consequences and feedback it generates. Experience acquires developmental value through detection, interpretation and revision, with organisational context influencing whether those processes can occur.

By the time the refill pack returns to full production, the line is running at commercial speed without the manual bend operators invented at 2.16 a.m. The most valuable output from the earlier stoppage is therefore carried inside the revised model of how the product behaves when design assumptions meet the conditions of actual work.