Culture Building: Change What the System Teaches

Culture decks are easy to agree with. They are weak evidence.

A company may sincerely value candour while its planning process punishes the person who questions a committed date. It may celebrate ownership while every consequential choice is reopened by a senior leader.

People resolve that contradiction by trusting the system. They learn which concerns are safe to raise, which trade-offs earn status, and which rules bend for powerful colleagues.

For this article, culture is the set of expectations people infer about what is safe, respected, and career-building in recurring situations.

That is an editorial definition, not a claim that culture has one accepted academic meaning.

It gives a product leader something practical to design. Do not begin with an adjective such as “bold” or “customer-centric”. Begin with a moment in which the organisation currently teaches the wrong response.

Treat a cultural claim as a prediction

“We challenge assumptions” should predict behaviour. When fresh evidence contradicts an approved roadmap, someone can name the conflict and the decision owner examines it.

The team’s standing should not fall merely because the news is inconvenient.

If the claim cannot predict a choice under pressure, it cannot guide an intervention. It may still be an aspiration, but employees cannot use it to decide what to do.

Write a cultural claim with four parts:

• Situation: the recurring moment in which judgement matters;
• Expected move: the observable behaviour the organisation wants;
• Boundary: what the principle does not permit or require;
• Organisational response: what leaders and systems will do when the behaviour is costly.

For example: “When evidence threatens a committed plan, show the conflict and its decision consequence.

“Do not delay delivery by reopening settled choices without material evidence. The decision owner will review the claim without penalising the messenger for surfacing it.”

The statement is not elegant. It is useful because people can test whether it is true.

Find the moment where the norm is made

Broad culture surveys can locate a concern, but averages hide the situation producing it.

“People do not feel empowered” might describe unclear authority, leaders reversing decisions, missing information, or consequences for a failed choice.

Trace one recurring episode instead. Use recent planning decisions, escalations, incident reviews, research playback, promotion discussions, or disagreements over quality. Reconstruct what happened before interpreting motive.

Ask:

• What did people believe they were expected to do?
• Which action was rewarded, ignored, or made expensive?
• Who could break the apparent rule without consequence?
• What information was available at the point of choice?
• Which work routine carried the behaviour into the next cycle?
• What would a reasonable person lose by acting differently?

The last question is the most revealing. If raising a risk leads to more work, less status, and no change in the decision, silence is not a mindset problem. It is a rational response to the environment.

Do not diagnose culture from leaders’ intentions. Compare the formal account with artefacts and decisions: meeting records, allocation changes, cancelled work, recognition, escalations, and the examples people tell new colleagues.

Explain why the current behaviour makes sense

Calling a norm “toxic” may express justified anger, but it does not yet explain how the norm persists. A useful diagnosis shows the local logic.

Suppose a product organisation wants teams to stop hiding delivery risk. The current behaviour might be sustained by several mechanisms:

• forecasts are treated as promises rather than current estimates;
• raising uncertainty triggers an executive review with no decision benefit;
• teams are compared on whether dates move, not whether risks are found early;
• programme updates allow only green, amber, or red, with no account of evidence;
• a previous messenger was publicly blamed for a missed commitment.

Each mechanism calls for a different intervention. A workshop on courage cannot repair a governance process that converts candour into punishment.

This is also where power enters the analysis.

Howard-Grenville’s study found that agency, organisational context, and relative power helped explain persistence and change in routines at a high-tech manufacturer.

One field study does not provide a general law for product companies. It supports a more careful question: whose deviation is allowed to redefine the routine, and whose is treated as non-compliance?

Change the reinforcement, not just the message

A culture intervention should alter the conditions people encounter in the chosen moment. Depending on the diagnosis, that may require changes across six levers.

Decision rights

Make authority and escalation conditions explicit. “Teams are empowered” means little if leaders can reopen any decision without new evidence or a named change in constraints.

Information

Give people the evidence required to act on the norm. Customer focus without access to research, support patterns, or commercial context turns into guesswork wearing a value label.

Routine

Put the behaviour inside work that already happens. A planning review might require the strongest disconfirming evidence. A launch review might name the risk owner and the condition for pausing exposure.

Feldman’s qualitative study in a university housing department challenges the view that routines are merely rigid scripts.

It describes participants changing routines as they respond to outcomes from previous iterations.

One setting cannot establish a general intervention rule. It suggests a useful design stance: inspect how people perform and adapt the routine rather than only rewriting its template.

Status and reward

Notice what earns desirable work, visibility, promotion evidence, and access to leaders. Public praise for collaboration cannot compete with career systems that reward individual rescue acts.

Consequence

Respond when behaviour violates a boundary, including when the person is commercially or technically valuable. The response should be fair and proportionate; silent exceptions still teach a rule.

Recovery

Create a path for correcting a mistake without hiding it. A culture that asks for experimentation but treats every failed test as poor judgement will produce safer narratives, not better learning.

Do not change all six levers by default. Choose the smallest set that explains the current norm and is strong enough to make the desired action locally rational.

Make leadership pay a visible cost

Employees watch what a leader gives up, not only what the leader endorses.

A leader who invites dissent when no commitment is at risk has sent a cheap signal. A leader who delays a favoured initiative because disconfirming evidence survives review demonstrates that the principle has decision weight.

The cost does not need to be dramatic. It might be accepting a revised forecast, protecting time for corrective work, giving credit to a team that stopped an unsafe release, or refusing to bypass a decision owner.

Avoid staging vulnerability as theatre. A leader describing a harmless past mistake does not create safety if challenging the leader’s current assumption still carries a penalty.

Keep challenge and accountability together

Psychological safety is often diluted into “be nice” or “make everyone comfortable”.

Edmondson’s field study of 51 manufacturing teams defined it as a shared belief that a team is safe for interpersonal risk-taking.

In that setting, psychological safety was associated with learning behaviour. The study is bounded to one company and does not prove that a particular workshop will improve performance.

For a culture intervention, safety should have a behavioural object. Can a researcher present unwelcome findings? Can an engineer stop a rollout?

Can a PM admit that the causal story is weak? Can a junior colleague question an influential person’s premise?

Safety does not remove standards. The researcher still needs evidence, the engineer needs a credible risk, and the PM must distinguish uncertainty from avoidance.

The organisation protects good-faith challenge while retaining scrutiny of the claim.

That combination prevents two common failures: using “candour” to excuse careless aggression, and using “safety” to avoid a difficult evaluation of the work.

Respect subcultures and genuine tensions

A product organisation contains different professional obligations. Security may value control, sales may value responsiveness, design may value coherence, and operations may value recoverability.

These are not necessarily cultural defects. They are sources of information about the system.

Do not force superficial agreement around a single behaviour. Define where a shared norm is necessary and where local variation is legitimate.

“Move quickly” might mean a reversible interface test in one team and a carefully staged migration in another.

When two values genuinely conflict, publish the decision rule rather than pretending both always win. A culture becomes more credible when it names the trade-off it will make and who has authority to make it.

Pilot the norm inside real work

Choose a routine where the problematic situation occurs often enough to observe and where failure will not expose the whole organisation to unnecessary risk.

Record the current sequence, introduce the changed reinforcement, and observe how participants adapt it. Do not require perfect compliance on the first pass; deviations may reveal missing information or an unrealistic boundary.

Zellmer-Bruhn’s field study surveyed 90 teams in three pharmaceutical and medical-products firms.

It found positive associations among reported interruptions, knowledge-transfer effort, and the acquisition of new work routines.

Exploratory results varied by interruption type; unusually disruptive events were negatively associated with routine acquisition.

It would be a mistake to manufacture disruption from those results. Use an existing break to ask whether the team sought, acquired, or failed to acquire a different routine.

Run a short review after each relevant episode:

• What action did the situation invite?
• What did the person actually do?
• How did peers, leaders, and the process respond?
• Which cost or ambiguity remained?
• What will change before the next occurrence?

This is culture work in operational form: not a campaign, but repeated correction of what the system teaches.

Measure whether the prediction became more true

Do not report culture change from participation, values recall, or favourable comments alone. Those show exposure to the intervention, not a changed norm.

Use an evidence portfolio tied to the original situation:

Evidence	What it can show	What it cannot show alone
Scenario questions	What people believe is expected in a specific situation	What they will do when status is at risk
Decision and meeting records	Whether the desired move appeared and affected the process	Private concerns that never surfaced
Allocation and recognition choices	Whether the organisation reinforced the behaviour materially	Why an individual choice changed
Confidential interviews	How the norm is interpreted and where it feels unsafe	Prevalence across the organisation
Repeated routine observation	Whether the pattern survives beyond a launch moment	Long-term business impact

Segment the evidence by role, level, location, and other relevant contexts where sample size and privacy permit. An average can improve while the people with least power experience no change.

Set reopening conditions. If the desired behaviour appears but decisions become slower without better challenge, narrow the trigger. If challenge rises but retaliation follows informally, the consequence and support mechanisms need repair.

A fictional culture intervention

Consider an explicitly fictional software company that says teams should surface commercial pressure early.

In practice, product managers wait until launch reviews because an earlier escalation usually produces more status meetings but no decision.

The leader chooses one situation: a requested customer commitment that conflicts with current product evidence. The desired move is a written trade-off with affected users, opportunity cost, uncertainty, and a decision request.

The organisation changes three levers. Account executives and PMs gain a named decision owner; the weekly commercial review replaces status colour with a choice and consequence; leaders record what was stopped or changed after a challenge.

The first review examines whether concerns arrived earlier, received a decision, and affected allocation. It also checks privately whether raising the conflict created retaliation. No improvement or business outcome is supplied here.

The example illustrates a culture claim becoming a testable operating prediction. It does not claim that one routine can transform an organisation.

Keep a culture intervention record

For each intervention, preserve:

• the recurring situation and observed norm;
• evidence for the diagnosis and important counterexamples;
• the desired move, boundary, and promised organisational response;
• the reinforcing mechanisms selected;
• affected groups and power differences;
• the leader commitments that give the claim weight;
• safety, consequence, and recovery paths;
• observations, interpretations, and revisions;
• the owner and conditions for stopping or expanding the intervention.

Culture building becomes credible when people can compare the promise with what the organisation did at the costly moment.

Sources

• Howard-Grenville: The Persistence of Flexible Organizational Routines (participant observation in one high-tech manufacturer; evidence about agency, context, power, and routine persistence rather than a universal culture model)
• Feldman: Organizational Routines as a Source of Continuous Change (qualitative study in one university housing department; basis for a performative model, not a universal intervention rule)
• Edmondson: Psychological Safety and Learning Behavior in Work Teams (multi-method field study of 51 teams in one manufacturer; association and process evidence, not a general intervention guarantee)
• Zellmer-Bruhn: Interruptive Events and Team Knowledge Acquisition (survey-based field study of 90 teams in three related firms; associations varied by interruption type and do not support manufacturing disruption)

Read next

• Team Engagement: Diagnose the Work Before Motivating People
• Change Leadership: Design the Transition, Not the Announcement
• Team Alignment: Make Local Decisions Coherent