Evidence — Organizational Development

Read a weak rating as a warning about what a model can be used for, not as a reason to throw it away. Several weak models are excellent shared vocabulary and terrible measurement instruments; the label and its justification say which. Where a better-evidenced alternative exists, the topic links to it.

The 16 questionnaires are not listed here. Each one operationalizes a model and carries that model's rating, so counting them again would make the field look better than it is.

A split rating — Weak as a model, moderate in components — is filed here under the value it leads with, and the full label is shown beside it. The source repository's own table counts split labels as a bucket of their own instead, so its columns are smaller than these and the two are not in conflict.

5 topics

Strong

Meta-analytic or large replicated experimental support.

Topic	Rating as written	Why
Goal setting theory	Strong	Several hundred studies and multiple meta-analyses across tasks, settings, and cultures, with effects that hold in field research and not only in the laboratory. Among the best-supported findings in the whole field. The boundary conditions are equally well established and are routinely ignored in practice: the effect depends on commitment, feedback, and task complexity, and it reverses in identifiable circumstances described below.
Job characteristics model	Strong	Meta-analytically supported across decades and many samples: the five characteristics relate reliably to satisfaction, motivation, and performance, and the mediating psychological states have held up reasonably well. Two caveats are real. Most measurement is self-report, so perception and objective job design are entangled. And growth need strength as a moderator has produced inconsistent results, so do not assume enrichment helps everyone equally — ask.
Psychological safety	Strong	The best-evidenced construct in this collection. A large literature spanning three decades links it to learning behavior, error reporting, information sharing, innovation, and performance, with meta-analytic support and effects that hold across sectors including health care and software. Three honest caveats: most measurement is self-report; the relationship with performance is strongest where work is interdependent and uncertain, which is precisely this guide's audience; and the measures have multiplied faster than the construct has been pinned down, with recent systematic review work finding no consensus on what belongs inside psychological safety and what merely correlates with it. Use Edmondson's original seven items rather than a vendor's variant, and do not compare your score against one produced by a different instrument.
Rogers diffusion of innovations	Strong	One of the best-evidenced models in this collection, synthesized from hundreds of empirical diffusion studies across agriculture, medicine, and communications, and extended by thousands since. The five attributes that predict adoption rate — relative advantage, compatibility, complexity, trialability, observability — are the most robust part and explain a substantial share of variance in adoption. The adopter categories and their percentages are a statistical convention rather than a law, and the categories are retrospective rather than stable traits.
Self-determination theory	Strong	One of the most extensively tested theories in psychology, with several decades of experimental and field research across cultures and domains, including work, health care, education, and sport. The undermining effect of expected contingent rewards on intrinsic motivation is meta-analytically robust and replicated, with the well-established boundary conditions noted in this file. The strongest live debate is over how universal the autonomy need is across cultures, where the evidence still favors universality.

8 topics

Good

Substantial empirical support, with some methodological limits.

Topic	Rating as written	Why
After action review	Good	One of the better-evidenced interventions in this collection. Meta-analyses of structured team debriefs find substantial and consistent improvements in team performance, on the order of a 20 to 25 percent improvement over teams that do not debrief, across military, medical, and aviation settings. The effect depends on the conditions described here — blameless framing, facilitation, participation regardless of rank, and follow-through — and an AAR that degrades into a status report should not be expected to produce it.
Burnout	Good	Three decades of research with a validated instrument and consistent findings across occupations and countries; the three-dimensional structure replicates well. Two genuine controversies remain. The boundary between burnout and depression is contested and imperfectly separated by current measures. And there is no agreed clinical cut-off score, so prevalence figures vary enormously between studies depending on the threshold chosen — treat any specific "X% of clinicians are burned out" claim with suspicion.
Conway's law	Good	Unusually well supported for an organizational claim. The mirroring hypothesis — that product architecture reflects organizational communication structure — has been tested empirically across commercial and open source codebases and generally confirmed. What is less established is the prescriptive half: that deliberately reorganizing to fit a target architecture reliably produces it.
Crew resource management	Good Good in aviation, moderate to good in health care	Among the better-evidenced interventions here. Aviation adoption coincided with sustained accident reduction, though confounded by many simultaneous safety improvements. In health care, systematic reviews of team training find consistent improvements in teamwork behaviors and knowledge, and improvements in clinical outcomes in several large studies, with substantial heterogeneity. Programs that work share features: multidisciplinary, simulation-based, with leadership participation and sustained reinforcement. One-off e-learning does not work.
DORA metrics	Good	Large-sample, multi-year, repeatedly replicated across annual cohorts, and the finding that speed and stability move together rather than trading off has held up across many years of data. The evidence is correlational and originally survey-based with self-selected respondents, so the causal chain from culture to delivery to organizational performance is argued rather than experimentally established. Strong by the standards of this field; not a randomized trial.
Organizational network analysis	Good Good as a method	Social network analysis is a mature, validated research methodology, and several of its organizational findings are robust: brokerage positions confer information advantage, weak ties carry novel information, and network position predicts performance and influence better than formal rank does. The weakness is practical rather than theoretical — response rates below roughly 70% make a map unreliable, and reported ties differ from actual ties.
Survey feedback method	Good	One of the few OD interventions with early controlled evidence. The Michigan studies found that units where managers discussed results with their own teams improved while units receiving the same report without discussion did not, and later meta-analyses of OD interventions placed survey feedback among the more effective approaches. The corollary is equally well attested: collecting data without feeding it back produces no improvement and measurably reduces trust.
Westrum organizational culture typology	Good	Unusually strong for a culture model. Developed from safety-critical industry case studies, then operationalized as a six-item scale and validated in large-sample software delivery research, where generative culture statistically predicts both delivery performance and organizational performance. The evidence is correlational and self-report, and the typology flattens what is really a continuum, but the construct-to-outcome chain is better established here than almost anywhere else in this collection.

12 topics

Moderate

Real support for the underlying principle; the model itself untested.

Topic	Rating as written	Why
Action research	Moderate	The participation mechanism is well supported: people implement what they helped design, and participation in decision-making is meta-analytically associated with acceptance of change. The full cycle has rarely been tested against controls, because its logic makes controlled comparison awkward. Strong on mechanism, weak on outcome trials.
Competing values framework	Moderate Moderate as a vocabulary, mixed as an instrument	The four types are a useful and widely recognized vocabulary, and the framework has a large psychometric literature by the standards of culture research. The instrument is weaker than its popularity suggests. The standard OCAI is ipsative — respondents divide 100 points across the four types — which forces the scores to trade off against one another, manufactures negative correlations between scales whatever the underlying culture, and makes conventional factor analysis and between-unit comparison unsound. That is the same objection this collection makes to the TKI; see [thomas-kilmann-conflict-modes](../thomas-kilmann-conflict-modes/). Likert adaptations behave better psychometrically, and validation work using them has found the diagonal types positively correlated where the framework predicts opposition, which is a direct challenge to the "competing" premise rather than a technical quibble. Predictive validity is mixed: culture type correlates with some effectiveness measures, and the framework makes no causal claim that a profile produces performance in a given context. It measures orientation, not Schein's third level.
Double-loop learning	Moderate	The espoused theory versus theory-in-use gap is well documented and easily demonstrated, and it remains one of the most useful observations in the field. Model I and Model II are clinically derived from extensive observation rather than experimentally validated, and Argyris's own finding — that people trained in Model II revert under stress — has not been systematically tested. The ladder of inference is a facilitation heuristic, not a cognitive model.
Galbraith star model	Moderate	Rooted in Galbraith's information processing theory of organization design, which has real empirical grounding: coordination need rises with task uncertainty and interdependence, and structures that fail to match it underperform. The five-point framework itself is a design heuristic that has never been tested as a whole. The reward-alignment claim is supported indirectly by the large literature on incentive distortion.
Google Project Aristotle	Moderate	A large, well-resourced internal study of 180 teams, but correlational, largely self-report, conducted in a single company, and never peer reviewed. Its null results may reflect restricted range, since every employee studied was already selected for high ability. Its value is convergent: the top finding independently reproduces the external [psychological-safety](../psychological-safety/) literature, which is strong. Do not cite it as a universal law.
High reliability organizations	Moderate	A rich and credible case study base from aircraft carriers, air traffic control, and nuclear operations, and the mindful organizing construct has been operationalized into validated survey scales with associations to safety outcomes in health care. Causal outcome evidence remains limited, and the founding organizations are unusual in resources and staffing, which complicates transfer. Normalization of deviance is well documented across multiple major accident investigations.
Just culture	Moderate	Well grounded in safety science and human factors research, which robustly supports the core premise that most errors are system-produced and that punishing error suppresses reporting without reducing error. Direct outcome evidence for just culture programs specifically is limited and confounded, since they are rarely implemented alone. The three-behavior classification is a decision aid with fuzzy boundaries, and can be applied post-hoc to justify a decision already made.
Nadler-Tushman congruence model	Moderate	The specific model has not been validated as an instrument, but its central principle is grounded in contingency theory, which has substantial support: there is no single best organizational form, and effectiveness depends on fit between the organization and its task and environment. Person-organization fit research supports the People pairings specifically. Stronger than most diagnostics here, and still not a measurement tool.
Schein model of organizational culture	Moderate	The three-level structure is a widely accepted conceptual framework rather than a measurement model, and the inference of underlying assumptions from artifacts is interpretive by design. The leader-mechanism claims — that what leaders attend to, how they respond to crises, and who they promote shape culture more than formal statements do — have partial support from leadership and socialization research. Conceptually strong, operationally soft.
Sociotechnical systems	Moderate Moderate to good	The original coal mining studies were case-based, but the core claim has accumulated wide support: technology implementations that ignore social arrangements underperform, and this is one of the most consistent findings in the information systems and health informatics literature. The work design principles overlap with the strongly supported [job-characteristics-model](../job-characteristics-model/).
SPACE framework	Moderate Moderate as a principle, weak as an instrument	There is no validated SPACE instrument and no prescribed metric set, so it cannot be benchmarked the way [dora-metrics](../dora-metrics/) can. The claim it rests on is well supported: single-metric productivity measurement produces distortion, and activity counts do not track outcomes. Its authors' central warning against individual-level measurement is consistent with the wider evidence on metric gaming.
Transformational leadership	Moderate	Real and repeatedly observed effects on satisfaction, commitment, and rated performance, supported by many meta-analyses. But the construct is under serious and, in the view of many researchers, unresolved criticism: the four dimensions do not separate cleanly in factor analysis; the standard instrument confounds leader behavior with follower liking and with outcomes, so it partly predicts effectiveness by definition; and most studies are cross-sectional and self-report, inviting common-method bias. Treat it as a useful vocabulary for describing leader behavior, not as a validated causal model, and be sceptical of any claimed effect size.

6 topics

Mixed

Genuinely contested, or components differ sharply in support.

Topic	Rating as written	Why
Burke-Litwin Causal Model	Mixed	More empirically ambitious than most diagnostics here, and only partly validated. The transformational versus transactional distinction is the best-supported element: it corresponds to a real and repeatedly observed failure pattern, in which structural or procedural fixes are applied to problems rooted in leadership, strategy, or culture. The full twelve-factor causal structure — including the specific direction and strength of each path — has not been confirmed, and some published tests find the hypothesized causal ordering only partially supported. Treat the tiers as sound and the individual arrows as hypotheses.
Employee engagement surveys	Mixed	The academic work engagement construct, measured by the UWES, has sound psychometrics and meta-analytic associations with performance and wellbeing. The practitioner engagement bundle is conceptually muddy and overlaps heavily with older, better-defined constructs. Engagement-to-performance correlations are real but modest, and causality runs in both directions. eNPS as a single item is noisy and heavily influenced by recent events; it is the weakest instrument in common use.
Herzberg two-factor theory	Mixed Mixed, tending weak	The specific two-factor structure is probably a method artifact: it was derived from the critical incident technique, in which people attribute good outcomes to themselves and bad outcomes to circumstances, and studies using other methods often fail to replicate the clean separation. The sample was engineers and accountants in 1950s Pittsburgh. What survives is the practical distinction between removing pain and creating meaning, which is better grounded by [self-determination-theory](../self-determination-theory/) and [job-characteristics-model](../job-characteristics-model/).
Plan-Do-Study-Act and the Model for Improvement	Mixed	The underlying components are sound: statistical process control is well founded, and explicit prediction before observation is basic scientific method. The applied literature is poor — systematic reviews of published health care PDSA work find that most reports omit predictions, run a single cycle, use no time-series data, and do not meet the method's own criteria. The method is better than most of its practice.
Switch framework	Mixed	A popular synthesis rather than original research, and the quality of the underlying work varies by tactic. Well supported: shaping the environment, defaults and choice architecture, action triggers or implementation intentions, and social proof. Weakly supported: the Rider and Elephant dichotomy, which is a simplification of dual-process theory that the theory itself no longer cleanly supports. Bright spots derives from solution-focused practice and has modest evidence.
Systems thinking	Mixed	The formal mechanics — stocks, flows, delays, oscillation from delayed feedback — are demonstrable and mathematically well founded. The system archetypes are descriptive heuristics with no validation; they are pattern names, not findings. Causal loop diagrams are easy to draw and hard to falsify, so treat any specific diagram as a hypothesis the group holds, not as evidence.

1 topic

Contested

Influential and seriously criticized on method.

Topic	Rating as written	Why
Hofstede cultural dimensions	Contested	Enormously influential and seriously criticized. The main methodological objections stand: a single company, a single era, a mostly male professional sample, dimensions derived by factor analysis on limited items, and nations treated as though they were cultures. The decisive practical objection is the ecological fallacy — within-country variance exceeds between-country variance on most dimensions, so applying a national score to an individual is invalid. Use it to generate questions about cross-border friction; never to characterize a person.

30 topics

Weak

Practitioner-derived, with little or no independent validation.

Topic	Rating as written	Why
ADKAR change management model	Weak	Commercially developed, with essentially no independent validation. The sequential barrier logic is plausible and clinically useful, and the specific distinction between Knowledge and Ability is well supported by the training transfer literature, which finds repeatedly that classroom learning does not survive contact with real conditions without practice and environmental support. The five-element structure itself is unproven.
Appreciative inquiry	Weak Weak to mixed	The published evidence is largely case studies written by advocates, with few controlled comparisons against problem-focused approaches. Partial indirect support comes from solution-focused practice, from research on positive affect broadening cognition, and from the fact that studying successes generates more actionable specifics than studying failures. The suppression-of-the-negative criticism is substantive and should be stated whenever the method is used.
Beckhard-Harris change equation	Weak Weak, and deliberately non-quantitative	The multiplicative form is a metaphor; nothing here is measurable in a rigorous sense, and the terms are not independent. Its value is diagnostic: the claim that a zero on any of dissatisfaction, vision, or first steps blocks change regardless of the others corresponds to three recognizable and distinct failure patterns. Use the structure, ignore the algebra.
Bridges transition model	Weak	Not validated as a staged theory, and the phases have no established durations or boundaries. The change-versus-transition distinction is conceptually useful and consistent with the job insecurity and organizational grief literatures, which document real performance and wellbeing effects from loss during restructuring. The specific three-phase structure is a clinical framework, not a finding.
Cynefin framework	Weak Weak empirically, high conceptual utility	Not validated, and probably not validatable in its current form: domain assignment is a judgment, and two competent people can disagree. What it does well is name a real and costly mistake — managing a complex problem as though it were merely complicated — that is otherwise hard to articulate. Use it to structure an argument about method, not as a classification instrument.
Drexler-Sibbet team performance model	Weak	No independent empirical validation; a practitioner framework developed for facilitation work. Its practical advantage over [tuckman](../tuckman-stages-of-group-development/) is structural rather than evidential: naming both the resolved and unresolved state for each stage makes it usable as a diagnostic. The seven questions are more useful than the seven labels.
Entry and contracting	Weak Weak as research, strong as practitioner consensus	No controlled studies show that explicit contracting improves outcomes. It is nonetheless endorsed near-unanimously across the practitioner literature, and the failure mode it prevents — a diagnosis that implicates a sponsor who never agreed to be examined — is well documented in case reports. Use it because the downside of skipping it is severe and the cost is ten minutes.
Force field analysis	Weak Weak as a measurement tool, moderate as a principle	The ratings are subjective and the arithmetic is a conversation device. The underlying principle is better supported: behavioral research on friction, defaults, and choice architecture consistently finds that removing barriers changes behavior more reliably and more cheaply than increasing persuasion or incentive. Trust the principle; do not trust the numbers.
Future search	Weak	Case-based, largely documented by its developers, with no controlled trials. Two design choices have independent support: getting the whole system in the room reflects the boundary-spanning finding that cross-unit contact reduces attribution error between groups, and the public admission step draws on well-established reciprocity and intergroup contact research. The three-day format and the five-task sequence are convention.
Immunity to change	Weak	Limited independent empirical validation; the supporting evidence is largely case material produced by the developers and their practitioners. The adult development theory underneath it is contested. The method's practical defense is different and does not depend on the theory: it converts vague complaints into specific, falsifiable claims about what actually happens to people here, which can then be tested directly. Requires skilled facilitation and genuine safety, and is intrusive when done badly.
Improvement kata and coaching kata	Weak	Derived from observation of Toyota, with little independent validation outside Toyota-lineage settings and no controlled comparison against other improvement approaches. The deliberate practice premise it rests on is strongly supported in skill acquisition research. The chief practical risk is documented rather than theoretical: most organizations abandon the daily coaching cycle within months, so the method's effect and its adherence are hard to separate.
Kotter 8-step change model	Weak	Derived from practitioner observation of roughly 100 change efforts, with no control group, no published method, and no subsequent test of the model as a whole. Individual steps have independent support — short-term wins map onto goal setting and progress research, and removing barriers maps onto force field logic — but the sequence, the count, and the causal claims are assertions. Use it as a completeness checklist.
Leavitt diamond model	Weak	A conceptual checklist from 1965, never validated. Its claim — that a deliberate change to one variable produces unplanned changes in the other three — is supported anecdotally and by the sociotechnical literature, but the model offers no way to predict which changes or how large. Use it as a five-minute prompt, not as a diagnostic.
Lencioni five dysfunctions of a team	Weak	A business fable, not research: no validation, no sample, no measurement. It remains popular because executives recognize themselves in it, which is a fact about resonance rather than about accuracy. Its first level overlaps heavily with [psychological-safety](../psychological-safety/), which covers similar ground with far better evidence, and the commitment principle — people commit to decisions they disagree with if they were genuinely heard — is consistent with procedural justice research.
Lewin change model	Weak	The three-step formulation was substantially simplified after Lewin's death, and has never been tested as a model. Later scholarship disputes how much of it is his. The unfreeze concept is supported indirectly — disconfirmation plus safety produces learning where disconfirmation alone produces defensiveness — and the institutionalization requirement is supported by the reliability with which unreinforced changes decay. Foundational, not evidential.
Maslow hierarchy of needs	Weak	The need categories have reasonable support; the strict ordering does not. People pursue esteem and self-actualization while lower needs are unmet, routinely and across cultures, and cross-cultural work finds the ordering and relative importance vary. The familiar pyramid does not appear in Maslow's own writing. Alderfer's ERG formulation, with no strict prepotency and an explicit frustration-regression mechanism, fits the data better. One practical implication survives: active safety needs block higher-order interventions.
Maturity Models	Weak	Maturity models are descriptive conventions, not tested theories. Progression through levels has rarely been shown to predict outcomes, the level definitions are usually authored by advocates or vendors, and the assumption that a single linear path runs from worse to better is an assertion rather than a finding. Evidence for CMMI specifically is contested and largely drawn from studies with selection problems. The rebuttals listed in this file are worth taking seriously. What maturity models do well is give a group shared vocabulary for a capability conversation; they fail when treated as a scorecard tied to funding or used to compare units with different constraints.
McKinsey 7S Framework	Weak	Practitioner-derived from consulting experience in the late 1970s, and never validated as a measurement model or tested for predictive value. There is no established way to score the seven elements, no evidence about how much each matters, and no cause-and-effect ordering among them. Its durable contribution is the hard-versus-soft distinction and the reminder that changing one element requires adjusting the others — a claim that is plausible, widely observed, and not formally demonstrated. Note also that it omits the external environment entirely, which [burke-litwin-causal-model](../burke-litwin-causal-model/) treats as the ultimate trigger for change.
McKinsey influence model	Weak Weak as a model, moderate in components	Consultancy-derived with no independent validation, and the survey evidence cited for it is proprietary. Each of the four conditions maps onto a better-supported literature: role modeling onto social learning theory, formal mechanisms onto the incentive distortion literature, skills onto deliberate practice and training transfer, and conviction onto internalized motivation in [self-determination-theory](../self-determination-theory/). Its practical value is as a completeness audit, which does not require the model to be true.
Open space technology	Weak	Practice-based, with no controlled evidence and no accepted outcome measure. Its four principles and the law of two feet are design conventions, not findings. The self-selection mechanism has a defensible rationale — every session contains only people who chose to be there — but this has not been tested against facilitated alternatives. Its documented failure mode is well attested: good day, no change, when convergence to action is skipped.
Process consultation	Weak Weak as research, strong as practitioner consensus	No controlled evidence that the process consultation stance outperforms expert or doctor-patient consulting. It is endorsed near-universally in the OD literature, and its central claim — that solutions the client owns survive longer than solutions delivered to them — is consistent with the participation and internalized motivation literatures. Treat it as a well-argued professional stance rather than a demonstrated method.
Satir change model	Weak Weak as a model, well documented as a phenomenon	The five-stage structure has no formal validation. The phenomenon it describes — a substantial temporary performance drop after a competence-disrupting change — is repeatedly documented, most clearly in electronic health record implementations, where clinician productivity declines measurably for weeks to months post go-live. Trust the dip; treat the stage names as vocabulary.
Situational leadership model	Weak	Several empirical tests have failed to confirm the style-to-development-level matching prescriptions, the underlying constructs are loosely defined and inconsistently measured, and the model has been revised repeatedly without new evidence. Its real value is as a conversation structure that gives a manager and a report shared language for asking for more support or less supervision — a genuine service that does not require the four-quadrant theory to be true. Better-evidenced neighbors are goal setting theory and leader-member exchange.
Team topologies	Weak Weak as a whole, good in its components	The four-type, three-mode scheme has not been tested. Its two foundations are much better supported: [conways-law](../conways-law/) has direct empirical backing, and cognitive load theory is well established in educational psychology, though its transfer to team-level capacity is an extrapolation. Treat the vocabulary as useful and the taxonomy as unproven.
The learning organization	Weak	Aspirational, influential, and largely unfalsifiable as stated. There is little evidence that the five disciplines produce measurable outcomes, and no accepted measure of whether an organization is one. Individual operational mechanisms described here are better supported — structured debriefs in particular, see [after-action-review](../after-action-review/). The learning disabilities list is more useful in practice than the disciplines, because each names a recognizable and checkable pattern.
Thomas-Kilmann conflict modes	Weak Weak instrument, useful taxonomy	The five-mode scheme is widely used and intuitively sound, but the TKI is a forced-choice ipsative instrument, which limits its psychometric properties and makes between-person comparison inappropriate. There is little evidence that mode preference is stable or predictive. The better-supported complement is principled negotiation's interests-versus-positions distinction, which has substantial support in negotiation research.
Tuckman stages of group development	Weak	A 1965 literature review of roughly 50 studies, mostly of therapy and training groups rather than work teams. Later research does not find that teams reliably pass through these stages in this order; Gersick's punctuated equilibrium work suggests a quite different pattern, with an abrupt transition near the temporal midpoint. The model also assumes stable membership, which most modern teams do not have. One message survives: early conflict in a new team is normal.
Viable system model	Weak	Internally coherent and derived rigorously from cybernetics, with almost no independent empirical validation as a predictive or diagnostic model. Ashby's law of requisite variety is a formal result and is sound; the mapping from it to five organizational systems is an argument, not a finding. Case reports of VSM diagnosis are largely written by advocates.
Weisbord six-box model	Weak	Never validated as a measurement model, and never tested for predictive value. It is a facilitation device: six prompts that reliably generate a useful conversation in ninety minutes. The one component with independent support is the formal/informal distinction, which mirrors the well-documented espoused-theory versus theory-in-use gap. See [double-loop-learning](../double-loop-learning/).
World café	Weak	No controlled evidence. The small-group-then-rotate structure has a plausible mechanism, since small groups reliably increase the proportion of people who speak, and that specific effect is supported by group dynamics research. Everything else — the number of rounds, the table hosts, the tablecloths — is convention. The value is almost entirely in question design and harvesting, neither of which the method itself guarantees.

3 topics

Very weak

Contradicted, or transferred from an unrelated domain.

Topic	Rating as written	Why
Discredited instruments	Very weak Very weak to refuted	Each entry below states its own status. The common pattern: an intuitively appealing typology, an appealing self-recognition experience, commercial incentives to sustain it, and either no predictive validity or a directly failed test of the mechanism it claims.
Kübler-Ross change curve	Very weak	The stages were not empirically validated even for grief, Kübler-Ross herself denied they were a linear sequence, and bereavement research has since largely rejected stage models. The transfer from terminal illness to organizational change is an analogy with no supporting study. It also carries an active harm: labelling a colleague as being 'in denial' converts a substantive objection into a psychological symptom. Prefer [bridges-transition-model](../bridges-transition-model/) or [satir-change-model](../satir-change-model/).
SWOT analysis	Very weak	There is no evidence that SWOT improves decision quality, and some reason to think it harms it: it produces unranked lists, mixes magnitudes, invites consensus platitudes, and encourages internal factors to be labelled strengths with no comparison point. Its survival is due to familiarity and low cost. If you use it, the TOWS pairing step is the only part that produces a decision.

3 topics

Not an empirical model

A taxonomy or checklist, to be judged on usefulness.

Topic	Rating as written	Why
Interventions catalog	Not an empirical model Not an empirical model — a taxonomy	The four-family classification is an organizing convention. Evidence varies enormously between the interventions inside it: work design and goal setting are strongly supported, survey feedback and structured debriefs are well supported, and most team-building and values programs are not. Check the individual file before assuming a family's reputation transfers to a specific intervention.
PESTLE analysis	Not an empirical model Not an empirical model — a checklist	The six categories are an arbitrary but reasonably exhaustive partition, with no validation and no theory behind the specific cuts. The broader practice of environmental scanning has modest support in the strategy literature, associated with better adaptation in turbulent environments. The value is in the impact-versus-confidence step, not in the categories.
The organizational development process cycle	Not an empirical model	This is a practice convention, not a tested theory. Its strongest component is the diagnose-before-intervene rule, supported indirectly by the poor track record of intervention-first programs, and by the survey feedback studies described in [survey-feedback-method](../survey-feedback-method/). Treat the eight stages as a checklist, not as a claim about how change necessarily proceeds.

3 topics

Not a model

A method or guidance, to be judged on how well it is run.

Topic	Rating as written	Why
Evidence-based practice in organizational development	Not a model 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.
Measurement and metrics for organizational development	Not a model Not a model — measurement guidance	The components have different pedigrees. Statistical process control and time-series analysis are formally sound. Goodhart's law is an observation with abundant documented instances rather than a theorem. Kirkpatrick's four levels are a widely used convention with known weaknesses, chiefly that the levels are not causally linked as often assumed. The staggered-rollout design advice reflects standard quasi-experimental methodology and is well founded.
Theory of change	Not a model Not a model — an evaluation method	Widely required by funders and evaluators, and there is limited direct evidence that producing one improves program outcomes. Its defensible claim is narrower and does not need a trial: an intervention whose assumptions are written down in advance can be falsified, and one whose assumptions are not written down will be evaluated retrospectively against whatever happened. The discipline is the deliverable.

1 topic

Normative

Value commitments, to be argued rather than measured.

Topic	Rating as written	Why
Ethics of organizational development	Normative Normative, not empirical	These are value commitments, and they should be argued rather than measured. Two of the specific claims here do have empirical support: surveying without acting lowers trust below the pre-survey baseline, and demographic cuts on small groups re-identify respondents. The rest is professional ethics.