The core idea
A value chain makes the proposed route from HR activity to organisational benefit explicit. Our simplified route is investment and activity → employee outcomes → operational performance → business outcomes. Guest’s HRM–performance work provides an important research foundation, but this four-box diagram is an editorial simplification, not a reproduction of his full model. Counting training hours or vacancies filled describes activity; it does not establish that capability, service or financial outcomes improved because of that activity.
Source and attribution [1]Using it in practice
Choose an initiative and write a testable statement for each link. Identify how long a change could reasonably take and who controls the next step. Ask whether managers provide an opportunity to use the new capability. Select evidence for the mechanism as well as the final result. Discuss alternative explanations with finance and operations before attributing a commercial improvement to HR.
An example, not a reported case
Worked example · illustrative
A fictional security firm funds improved onboarding. It expects quicker competence, then fewer avoidable handover errors, then more dependable service and stronger client retention. HR checks practical competence rather than attendance alone. Operations tracks errors using a consistent definition. Client outcomes are reviewed over a longer period, with changes in contract mix and staffing recorded. The chain is a hypothesis to test, not a calculation that turns every training participant into a fixed amount of profit.
What to watch for
Reverse causation is possible: successful organisations may have more resources to invest in HR. Several initiatives and external conditions can change at once. A plausible sequence of arrows is not causal evidence. Avoid double-counting the same benefit under engagement, productivity and retention. Some valuable outcomes should be reported directly rather than forced into money.
Build evidence for each link
For investment, record the actual resources used. For employee outcomes, define a capability or experience that could change. For operational outcomes, choose a relevant quality or service measure. For business outcomes, agree a time horizon and accounting definition. Note where data is missing instead of replacing it with an attractive assumption.
Distinguish activity from outcome
Delivering 200 training hours is an output. Demonstrating a new skill is a nearer outcome. Fewer errors may be a later outcome. Ask what would make each link fail: poor practice opportunities, an unusable system or a workload that prevents the intended behaviour. Those failure points should shape the intervention.
Consider the counterfactual
What would likely have happened without the initiative? A comparison group, phased introduction or carefully justified time-series design may help. If those are not feasible, report a contribution argument with its limits. Do not relabel a before-and-after change as a proven causal effect.
Use the chain with AMO and logic models
AMO helps ask whether ability, motivation and opportunity support the mechanism. A logic model expands the resources, activities and assumptions. Together they can improve programme design, but combining frameworks does not strengthen evidence unless the actual links are observed and tested.
Write the mechanism, not just the arrows
Take one proposed investment and explain why each link should follow. 'Manager training improves retention' skips the mechanism. A more useful hypothesis is that practice and feedback improve the quality of one-to-ones, which helps employees resolve work problems earlier, which may reduce avoidable exits. Each link can fail independently. Training may be well delivered but managers may have no time to use it. Record the assumptions beside the chain so the diagram becomes something that can be investigated.
Choose an indicator at each link
Separate delivery, behaviour, employee experience and organisational outcome. For the one-to-one example, attendance is a delivery measure; sampled conversation quality is a behaviour measure; perceived access to useful support is an experience measure; retention is a later outcome. Decide when each could reasonably change and who will collect it. Do not assume a rise in satisfaction proves a financial return. Equally, an unchanged annual turnover rate after four weeks is not a sensible reason to dismiss an early behavioural improvement.
Investigate a broken link
Suppose managers complete training and report more frequent conversations, but employees say problems remain unresolved. Examine the content and authority available to managers before commissioning more training. They may understand the issue but lack control over schedules or staffing. The relevant intervention might then involve operations rather than L&D. Treat the chain as a diagnostic tool: identify where evidence ceases to support the proposed explanation, and revise that part instead of defending the original business case.
Make a proportionate contribution claim
Compare outcomes with a credible baseline or comparison group and record other changes, such as pay increases, a new contract or labour-market shifts. Where strong causal identification is unavailable, describe evidence consistent with a contribution rather than claiming proof. Present the cost, plausible benefit range and uncertainty together. This makes the evaluation more useful to a decision-maker: they can judge whether to continue, adapt or stop the intervention without being asked to accept a diagram as evidence of causation.
Use the chain to challenge an investment case
Write the proposed mechanism before selecting a metric. For example, better induction may improve task confidence and access to help, which may reduce avoidable errors. Check each link: attendance is not competence, and fewer errors may reflect quieter demand. Identify a comparison and a possible unintended effect, such as rushing new starters to independence. Report which links have evidence and which remain assumptions.
Specify what would disprove the proposed explanation
For an onboarding investment, set out the expected sequence and a competing explanation. Better role clarity may reduce mistakes, but new equipment or lower demand might do the same. Record when each change occurs and look for evidence at intermediate points. If confidence improves but errors do not, investigate whether employees can use their knowledge or whether the original diagnosis was wrong. This is more informative than reporting a favourable financial result and attributing all of it to HR.
Present the chain as a decision, not a certainty
Khan and Millner emphasise connecting analysis to the business question and the next action. Present which links have evidence, what remains uncertain and the smallest useful test. For example, propose a supported onboarding pilot at comparable sites rather than immediate organisation-wide rollout. Include employee experience and workload as safeguards. Templeman’s business-partner perspective adds the need to understand operational constraints: a theoretically sound intervention will fail if managers cannot release time or if success depends on an unavailable resource.
Take it into your next conversation
Three useful questions.
- Which decision will this analysis change?
- Are the population, denominator, time period and assumptions explicit?
- What alternative explanation or unintended effect must we check?
Related terms
Go to the evidence
Sources & attribution
Further reading: Michael Armstrong with Stephen Taylor (2023), Armstrong’s Handbook of Human Resource Management Practice, 16th edition, chapters 5. Edition & reading notes ↗
Further reading: Nadeem Khan and Dave Millner (2023), Introduction to People Analytics: A practical guide to data-driven HR, 2nd edition, PDF pp. 227, 244 and 294–295. Edition & reading notes ↗
[1] Guest (1997), Human resource management and performance: a review and research agenda ↗
The core idea is an original summary of the cited work. Application notes, examples and sketches are our interpretations, not quotations or reproductions of the authors’ figures. Publisher records may require access to read the full original work.
Expanded on 20 September 2026 for the People Analytics collection; original URL retained. Application guidance expanded on 26 September 2026. These practical suggestions and invented examples are editorial explanations; further-reading sources are not the original authors of the historical models unless expressly identified. Original overview review dates are retained.
Published 2026-09-20 · Reviewed 2026-09-25. Editorial approach