Promoting a Positive Health and Safety Culture

A complete climate-survey package

A defensible package contains an assessment brief, population and sampling frame, respondent information, accessible questionnaire, response and coding rules, confidential administration arrangements, coverage-monitoring plan, analysis template, feedback commitment and repeat-measurement plan. A census invites everybody, but its respondents can still be unrepresentative. A sample can be defensible only when the selection logic and missing groups are visible.

1. State the decision the survey must inform and the period it represents.
2. Choose cultural dimensions and define what evidence would count for each.
3. Identify eligible employees, contractors, agency workers, shifts, sites, functions and access needs.
4. Choose a census or reasoned sample; define subgroup reporting and small-number protections.
5. Select an established instrument or design an organisation-specific questionnaire. Do not casually alter a validated instrument’s wording or scoring.
6. Draft respondent information, questions, scales and limited demographics.
7. Cognitively pilot with people similar to intended respondents: ask what each item means, how they chose an answer and whether the options fit.
8. Revise ambiguous, leading, double-barrelled, absolute or burdensome items.
9. Explain anonymity or confidentiality honestly, provide protected time and accessible completion routes, then launch consistently.
10. Monitor group coverage without pressuring individuals or inspecting responses prematurely.
11. Apply predefined validation, missing-data, scoring and exclusion rules; retain a decision log.
12. Analyse item distributions and safe subgroups, triangulate with other sources, return findings and action, then reassess after implementation.

What actual questionnaire questions can look like

Important: these are original DB HSE illustrative questions, not a validated instrument. Agreement scales suit statements; frequency scales suit repeated experience; confidence scales suit perceived ability or security. “Not applicable” or “not experienced” must be coded separately from a neutral opinion.

A carefully protected open question may ask: “What is one recent organisational decision or response that strengthened or weakened safe work?” Comments provide context; they are not automatically numerical scores.

A usable semi-structured interview guide

Interviews are normally confidential, not anonymous, because the interviewer knows who participated. The opening should explain purpose, voluntary participation where applicable, recording or notes, information use, privacy limits, withdrawal arrangements and the route for an immediate serious-risk disclosure. Use an interviewer who is competent, neutral and preferably outside the participant’s direct reporting line.

1. Opening: “Please describe your role and a typical working period, without naming colleagues.”
2. Pressure: “Describe a recent occasion when delivery pressure and a safety requirement competed.”
3. Voice: “What happened the last time you or a colleague raised a hazard or near miss?”
4. Control: “What makes stopping or delaying work easy or difficult here?”
5. Leadership: “How do managers respond when a required control is unavailable?”
6. Usability: “Can you describe a procedure that is easy or difficult to use during real work?”
7. Learning: “How are workers and contractors told what happened after raising a concern?”
8. Subculture: “What differs between sites, shifts, roles or employment groups?”
9. Strength: “Which practice strengthens safe work and should be protected?”
10. Closing: “What single organisational change would make the greatest improvement, and is there anything important I did not ask?”

Neutral probes include “Can you give a recent example?”, “What happened next?”, “Who was involved in that decision?”, “Is that typical or unusual?”, “What evidence might help verify it?” and “Have you experienced an example that contradicts this?” Produce a coverage record, de-identified authorised notes or transcripts, codebook, theme matrix, contrary cases, causal hypotheses, evidence gaps and limitations. A theme’s frequency in a small purposive sample does not establish workforce prevalence.

Common errors: starting with questions before defining the decision; promising anonymity that cannot be delivered; using managers’ preferred interviewees; changing questions midstream without a record; treating a favourable average as proof of culture; and collecting sensitive evidence without a response, retention or feedback plan.

Quantitative workflow and record

1. State the question, population, variable and period.
2. Check definitions, dates, duplicates, impossible values and missing answers; log every exclusion or correction.
3. Record numerator and denominator before calculating percentages.
4. Show response distributions, not only a combined mean.
5. Compare like instruments, definitions, exposure bases and time periods.
6. Examine safe subgroups without creating re-identification risk.
7. For event comparisons, use a justified exposure base and show the formula.
8. Test sensitivity to missing data, changed definitions, small counts and reasonable alternative assumptions.
9. Ask whether the difference is practically important for safety, not merely numerically different.
10. Triangulate and state what the numbers cannot explain.

Favourable proportion = favourable responses ÷ valid responses × 100. Response rate = usable responses ÷ eligible invitations × 100. Exposure rate = events ÷ exposure units × chosen standard base. State the base and never compare rates built from different event definitions.

Qualitative workflow and record

1. Define the cultural evaluation question and assemble the authorised, de-identified evidence corpus.
2. Read the complete evidence set before selecting memorable extracts.
3. Build a codebook from the declared criteria while allowing relevant new codes to emerge.
4. Define inclusion and exclusion boundaries for each code; code evidence segments, not people.
5. Review a sample with another competent analyst where proportionate and resolve material differences transparently.
6. Group related codes into themes and test their internal coherence.
7. Compare sites, shifts, roles and employment groups where safe.
8. Search deliberately for disconfirming cases and alternative explanations.
9. Connect themes to cultural criteria, operating conditions, risk and decisions.
10. Record analyst influence, coverage, limitations and confidence.

Common errors: percentages without denominators; percentage change confused with percentage-point difference; unlike periods compared; averages hiding subcultures; response rate treated as representation; event counts used without exposure; dramatic quotes cherry-picked; theme frequency treated as prevalence; contradictory evidence ignored; and correlation converted into causation.

Report against criteria and measures of a positive culture

Here, assessment criteria means the declared organisational culture dimensions, not OTHM AC 4.1–4.5. A measure may be a leading or lagging indicator and may be quantitative or qualitative. No universal “positive culture score” proves success. Define what the desired condition means, select balanced indicators and state interpretation rules before reviewing the outcome.

A criterion can be an established strength, partial or inconsistent, a priority weakness or have insufficient evidence. Confidence is a separate judgement. Low confidence does not prevent immediate protection when a credible serious-risk signal is present; it limits the broader cultural claim.

Define the improvement need before selecting the solution

Use this structure: current condition + required condition + affected group + consequence + plausible organisational cause + strength to preserve + evidence confidence. “Buy training” is a proposed measure, not an improvement need.

1. Separate urgent physical or legal-control needs from discretionary ranking.
2. Name the organisational need or approved objective rather than writing “supports business goals.”
3. Generate credible alternatives with affected people and eliminate options that leave unacceptable risk inadequately controlled.
4. Specify each option sufficiently to calculate whole-life cost and non-cash resource demand.
5. Compare causal fit, benefit, evidence, urgency, equity, feasibility, dependencies and unintended consequences.
6. Test whether reasonable changes to assumptions reverse the decision.
7. Record why one recommendation should proceed first, what must be preserved and when the judgement will be reviewed.

Common errors: traffic-light ratings with no rule; zero injuries treated as positive culture; evidence confidence mixed with performance; weakness converted instantly into training; every low score ranked equally; cheap options favoured over adequate control; existing staff time treated as free; and benefits claimed without an owner or measure.

Strategic planning and timelines

1. Verify the AC 4.2 findings and group AC 4.3 measures into coherent workstreams.
2. Describe current state, measurable future state and the causal steps—the theory of change—between them.
3. Write outcome objectives and preserve finding-to-measure traceability.
4. Break each workstream into deliverables; identify predecessors, dependencies, resource demand and affected groups.
5. Estimate duration with people who will deliver and experience the change.
6. Sequence 0–30 days, 31–90 days, 3–6 months and 6–12 months; distinguish immediate protection from sustainable redesign.
7. Set milestones, consultation points, assurance reviews, buffers and scale/adapt/pause/stop gates.
8. Baseline the authorised schedule, report status honestly and reforecast only through change control.

A timeline record should contain work package, deliverable, owner, start, finish, duration, predecessor, resource demand, consultation point, milestone, evidence, decision gate, buffer, status and revised forecast. Dates without dependencies or capacity are a calendar, not an implementation strategy.

Communication must be two-way and test understanding

Expectations are reciprocal operating standards

An expectation should state actor + observable behaviour or control + situation + response time + decision boundary + evidence + escalation + reciprocal organisational commitment. It should be feasible under real workload and applied fairly. Test understanding through examples, non-examples and teach-back.

All Northstar time commitments above are fictional teaching choices, not universal standards. They demonstrate specificity. The same expectation drives system capacity, named accountability, contractor/night-shift communication, milestones, governance review and a budget line.

Common errors: a training calendar presented as strategy; arbitrary dates; work scheduled beyond capacity; posters or email treated as communication; “sent” confused with “understood”; vague “safety first” slogans; zero-reporting targets that suppress voice; rules applied only to workers; and scaling before a pilot gate.

Resource allocation schedule

Resources include named people or full-time-equivalent capacity, protected worker and contractor time, backfill, competence and facilitation, equipment, technology, data/privacy support, procurement, translation and accessibility, independent assurance, direct finance, internal opportunity cost and contingency. “Existing staff” does not mean “no resource.”

Roles, decision rights and governance

For every deliverable and material decision, name one accountable role, the responsible people doing the work, those consulted while influence remains possible and those informed. Confirm authority, competence, time, deputy and escalation; a RACI letter without capacity is decorative. Operational leaders should own operational culture change; the H&S function advises, enables and assures rather than absorbing all accountability.

Governance should specify sponsor and steering forum, terms of reference, representative membership, quorum, decision rights, cadence, evidence pack, risk thresholds, escalation, independent assurance, conflicts, change control, decision/action/risk logs and benefit reviews. At each gate, leaders should have authority to scale, adapt, pause or stop. Record rationale and material dissent—not only attendance.

How to build an itemised budget

1. Fix currency and price year, scope and planning horizon.
2. Take quantities and timing from the work and resource breakdown structures.
3. Estimate quantity × supported unit rate; cite the price basis and record confidence.
4. Classify one-time or capital, direct implementation, recurring operating, internal time, backfill, technology, accessibility, evaluation and assurance costs.
5. Record opportunity cost separately where capacity is diverted; avoid double counting internal time and cash invoices.
6. Profile costs by phase and cost owner, then add evidence-based contingency for identified uncertainty.
7. State tax, inflation, approval limit and excluded costs; obtain Finance validation.
8. Compare approved budget, committed, actual and forecast; explain variance and reforecast through change control.

Common errors: assuming H&S can absorb the work; funding data collection without response capacity; assigning two or no accountable owners; holding governance meetings with no decisions; dashboards showing activity but not effect; treating a funding ceiling as the estimate; omitting recurring or internal costs; double-counting contingency; and presenting hoped-for savings as guaranteed cash.