Topic
Evidence-based practice in organizational development
Not a model Not a model — a method for weighing evidence
Evidence-based management, as formulated by Denise Rousseau, Rob Briner, and Eric Barends, means making decisions by conscientiously combining four sources of evidence, none of which is sufficient alone.
Evidence: Not a model — a method for weighing evidence. The underlying claim, that combining research, organizational data, stakeholder input, and checked judgment beats intuition alone, is supported by decision research on clinical and managerial judgment, where structured methods routinely outperform unaided expert prediction. The six-step process itself is a procedural convention borrowed from evidence-based medicine.
Use when: choosing between competing interventions, or challenging a popular practice that a sponsor has already decided on.
Do not use when: nothing. This is a standing discipline, not a technique.
The four sources
- Scientific evidence: findings from peer-reviewed research, ideally meta-analyses rather than single studies.
- Organizational evidence: your own internal data — turnover, incidents, cycle time, survey trends, financials.
- Stakeholder evidence: the values, concerns, and knowledge of the people affected.
- Practitioner expertise: your judgment and experience, treated as evidence to be checked rather than as authority.
The six steps
- Ask: turn the problem into an answerable question.
- Acquire: search systematically for evidence in all four sources.
- Appraise: judge quality and relevance, not just existence.
- Aggregate: weigh the sources together.
- Apply: decide and act.
- Assess: evaluate the outcome, and feed it back.
Evidence quality across popular OD models
Practitioners are rarely told which of their tools are supported. A rough guide:
| Model | Evidence strength | Note |
|---|---|---|
| Psychological safety | Strong | Large literature; robust link to learning behavior and error reporting |
| Self-determination theory | Strong | Decades of experimental support |
| Job characteristics model | Strong | Meta-analytically supported |
| Goal setting theory | Strong | Among the best-supported findings in the field |
| Transformational leadership | Moderate | Real effects, contested measurement |
| Kotter 8-step | Weak | Practitioner-derived, rarely tested as a whole |
| ADKAR | Weak | Useful checklist; little independent validation |
| Tuckman stages | Weak | Memorable; groups do not reliably proceed in order |
| Kübler-Ross applied to organizations | Very weak | Derived from bereavement; not validated for change |
| Maturity models | Weak | Descriptive; level progression rarely predicts outcomes |
| Hofstede dimensions | Contested | Influential; heavily criticized on method and level of analysis |
| Learning styles | Refuted | Do not build interventions on this |
| MBTI as a selection or development tool | Weak | Low test-retest reliability; avoid for consequential decisions |
Weak evidence does not mean useless. A weakly supported model can still be a good communication device — a shared vocabulary that helps a group talk about a real problem. The error is treating it as a causal claim and predicting outcomes from it.
Practical habits
- Ask "what would change my mind?" before launching an intervention.
- Prefer effect sizes to anecdotes, and meta-analyses to single studies.
- Distinguish "no evidence of effect" from "evidence of no effect".
- Beware the consultant's citation chain: a claim repeated in ten blog posts that all cite each other is one claim, not ten.
- Beware benchmark envy: "Google does it" is not evidence that it will work at a 200-bed district hospital.
Examples by audience
Health care: clinicians already practice evidence-based medicine and will hold management claims to the same standard. Use that; it is an advantage. Bring the effect size.
Software: engineers respond to instrumented internal evidence far more than to external frameworks. Show the pipeline data.
Executive: the highest-value use is challenging an expensive committed decision — a reorganization, a return-to-office mandate, a values relaunch — with the question "what evidence would tell us this worked?"
See also
- measurement-and-metrics
- discredited-instruments — the tools this discipline rules out, and what to use instead.
- action-research
- Barends & Rousseau, Evidence-Based Management.
- https://cebma.org