Evidence-based practice in HR means making a people decision by combining credible research, organisational data, stakeholder concerns and professional expertise, rather than relying on habit, anecdote or a fashionable intervention. The four sources do not produce an automatic answer. They provide different forms of insight that must be weighed against the decision, the context and the likely consequences.
This matters because people decisions affect cost, capability, fairness, wellbeing and organisational trust. A manager may be convinced that a retention bonus will solve turnover; an employee survey may point to workload; labour-market data may indicate a scarce skill; research may show that money alone has limited value where leadership, work design or career opportunity remain weak. Evidence-based practice makes those competing explanations visible before a decision is made.
The four sources of evidence
CIPD identifies four complementary sources: scientific research, organisational data, stakeholder concerns and professional expertise (CIPD, n.d.). The strength of the approach is not the number of sources cited. It is the disciplined integration of relevant, credible and proportionate evidence.
| Source | What it can contribute | Limitation to manage |
| Scientific research | Tested theories, systematic reviews and evidence about likely relationships or interventions | Findings may come from a different sector, country or workforce |
| Organisational data | Local patterns in turnover, absence, performance, reward, progression or employee experience | Data can be incomplete, biased or mistaken for an explanation rather than a signal |
| Stakeholder concerns | Lived experience, feasibility, competing interests and unintended consequences | Powerful voices can crowd out less visible groups unless participation is designed carefully |
| Professional expertise | Context, practice knowledge, ethical judgement and an understanding of operational constraints | Experience can become overconfidence if it is not tested against data and alternative views |
No source should dominate by default. Research without local data can be abstract. Data without stakeholder insight can hide meaning. Stakeholder views without quality checks can become a contest of preferences. Professional experience without challenge can preserve ineffective practice.
A decision process rather than a reading list
A useful starting point is to define the decision in a way that can be investigated. “Improve engagement” is too broad. “Reduce regretted turnover among newly qualified customer advisers within twelve months, without creating inequitable reward outcomes” is a decision problem. It identifies a population, outcome, timeframe and constraint.
The next step is to identify plausible explanations. If turnover is high, is the main issue pay, workload, career progression, management quality, job fit, flexibility, external demand or a combination? The purpose is not to create a perfect causal model. It is to avoid selecting an intervention before understanding the problem.
| Stage | Practical question | Useful output |
| Define | What decision must be made, for whom and by when? | Clear problem statement and success measures |
| Gather | What does each evidence source contribute? | Short evidence map, not an unfiltered data dump |
| Appraise | How credible, current and applicable is the evidence? | Strengths, gaps and assumptions |
| Integrate | Where do sources agree or conflict? | Options and trade-offs |
| Act and learn | What will be implemented, measured and reviewed? | Decision record, safeguards and review points |
Appraising research and organisational data
A source is not useful simply because it is published or numerical. Research should be assessed for method, setting, sample, currency and relevance. A study of voluntary turnover in a different labour market may still offer insight, but it should not be treated as proof that the same intervention will work locally. Systematic reviews and high-quality professional guidance may be particularly valuable because they consider a wider evidence base.
Organisational data needs similar discipline. A dashboard showing increased turnover may be accurate, but it cannot by itself explain why people leave. Break data down by role, length of service, manager, location, demographic group and reason for leaving where ethical and statistically meaningful. Check definitions: does “turnover” include fixed-term contracts, internal moves or planned restructuring? Is the data timely? Are there missing records? These questions are central to responsible people analytics, not technical extras.
Workplace application: Northstar Logistics
Consider fictional Northstar Logistics, where voluntary turnover among depot planners has risen from 14% to 24% in a year. A senior leader proposes a retention payment. An evidence-based approach does not reject the idea automatically; it tests it.
| Evidence source | Northstar finding | Decision implication |
| Research | Retention is associated with a combination of fair reward, development, leadership and job quality rather than one lever alone | Test a portfolio of measures rather than assuming a payment is sufficient |
| Organisational data | Exits are concentrated in the first 18 months and under three managers; overtime is highest in those teams | Investigate induction, workload and manager practices |
| Stakeholder concerns | Planners report weak progression visibility and last-minute rota changes | Co-design career and scheduling improvements with affected employees |
| Professional expertise | Operations leaders identify a genuine external shortage in planning capability | Target labour-market action while avoiding a blanket, costly response |
Northstar may still use a targeted payment for scarce roles, but it should be paired with manager support, transparent progression criteria and workload review. Success should be measured through regretted turnover, time to competence, overtime, internal progression and employee experience—not retention alone.
✅ Get Your Assignment Done by Experts
Evidence, ethics and inclusion
Evidence-based practice is not value-free. A proposal can be statistically effective yet ethically problematic. For example, an algorithm may identify employees at high risk of leaving, but acting on that insight could create privacy concerns, reinforce bias or lead managers to overlook people not represented well in the data. Professional judgement includes asking what is fair, lawful, proportionate and explainable.
Inclusion also affects evidence quality. If survey participation is low among night-shift or remote workers, treating the headline result as representative may produce a decision that favours the most visible group. Stakeholder input should therefore be designed for access, psychological safety and confidentiality. See ethical decision-making in people practice for a broader discussion of values, evidence and accountability.
Where evidence-based practice goes wrong
The most frequent weakness is confusing data availability with decision relevance. A people team may report many metrics because they are easy to obtain, while failing to collect the evidence needed for the actual decision. Another weakness is confirmation bias: seeking evidence that supports a preferred intervention. A decision log that records alternatives, assumptions and disconfirming evidence can help.
| Weak approach | Why it weakens the decision | Better alternative |
| Copying another employer’s initiative | Context, workforce and constraints may differ | Identify the mechanism, then test local applicability |
| Treating a survey score as a cause | Scores are signals, not explanations | Combine patterns with qualitative insight and operational data |
| Selecting only supportive research | Creates false certainty | Look for limitations, mixed findings and counter-evidence |
| Ignoring feasibility | A good idea may not be deliverable | Consider capability, cost, governance and implementation from the outset |
| Reporting only favourable results | Prevents learning and erodes trust | Review outcomes openly and adjust action when evidence changes |
Connecting evidence to action
Evidence-based practice should lead to a proportionate decision, not permanent analysis. Set out options, expected benefits, risks, resources, owners and measures. A small pilot may be preferable where uncertainty is high. A decision can then be reviewed against agreed indicators and stakeholder experience. This connects naturally with strategic business proposals and the use of balanced measures rather than one headline outcome.
Implementation governance: making evidence use routine
Evidence-based practice becomes credible when it is built into normal decision routines rather than reserved for major projects. Northstar could require a short evidence note for material people proposals: the decision being sought; the four sources considered; the principal assumptions; risks to inclusion or privacy; the intended measures; and a date for review. This does not turn management into bureaucracy. It creates a proportionate record of why an intervention was chosen and what would cause leaders to reconsider it.
Senior leaders have a particular role in modelling this discipline. When they ask only for a preferred option, a people team may produce evidence to defend it. When they ask what evidence would change the recommendation, which voices are missing and what trade-offs remain, they create better conditions for challenge. The same applies to pilots. A pilot needs an explicit hypothesis, a comparison where feasible, participant safeguards and an agreed interpretation of success; otherwise it becomes a small-scale version of the original assumption.
Counter-evidence deserves specific attention. If exit interviews suggest pay is the dominant issue but workforce data shows departures rising after manager changes and stakeholder discussions identify unsustainable workload, leaders should not simply average the signals. They should investigate the mechanism: pay may matter in a tight market, while workload and management determine whether employees decide to act on external opportunities. This is the difference between collecting evidence and reasoning with it.
Frequently asked questions
What are the four sources of evidence in CIPD practice?
Is evidence-based practice only for large organisations with HR analytics teams?
Can research evidence override employee views?
How does evidence-based practice link to people analytics?
References
CIPD (n.d.) Evidence-based practice for effective decision-making. Available at: https://www.cipd.org/en/knowledge/factsheets/evidence-based-practice-factsheet/ (Accessed: 24 August 2026).
CIPD (n.d.) Evidence-based: Profession Map. Available at: https://www.cipd.org/en/the-people-profession/the-profession-map/explore-the-profession-map/professional-values-purpose/evidence-based/ (Accessed: 24 August 2026).
Barends, E., Rousseau, D.M. and Briner, R.B. (2014) Evidence-based management: The basic principles. Amsterdam: Center for Evidence-Based Management.
Rousseau, D.M. (2012) ‘Envisioning evidence-based management’, in Rousseau, D.M. (ed.) The Oxford handbook of evidence-based management. Oxford: Oxford University Press, pp. 3–24.