← People analytics

People analytics · models & decisions

Logic models for HR interventions

Separate resources, activities and deliverables from the changes an initiative is meant to produce.

Programme-evaluation tradition; ADM HR application using UK government guidance · 2026 application · Original editorial explanation
Logic models for HR interventions — simplified model sketchInputs · Activities · Outputs · Outcomes · Impact. Editorial interpretation, after Programme-evaluation tradition; ADM HR application using UK government guidance (2026 application).Inputs01Activities02Outputs03Outcomes04Impact05ADM HR · ORIGINAL INTERPRETATIONProgramme-evaluation tradition; ADM HR application using UK government guidance · 2026 application
Original simplified sketch after Programme-evaluation tradition; ADM HR application using UK government guidance (2026 application). Read the explanation for assumptions and limitations. Download SVG ↗

The core idea

A logic model sets out how an intervention is expected to work. Inputs are resources; activities are what is done; outputs are immediate deliverables; outcomes are changes in behaviour, capability or conditions; impact concerns wider or longer-term effects. Definitions can vary between evaluation traditions, so state how you use them. Add assumptions and external influences rather than drawing an apparently inevitable chain. A logic model supports evaluation design but does not itself demonstrate that a programme caused its outcomes.

Source and attribution [1]

Using it in practice

Develop the model with people who deliver and experience the initiative. Start from the intended change and work backward to what would need to happen. Identify a small number of measures and a feasible way to collect them. Check that each activity has an owner and that the assumed resources exist. Use the model to decide what to investigate if the programme does not work as expected.

An example, not a reported case

Worked example · illustrative

An invented onboarding programme uses supervisor time and training materials as inputs. Activities include supported practice and check-ins. Outputs include completed sessions. Outcomes include demonstrated competence and clearer escalation. The hoped-for impact is more reliable service and sustainable retention. If attendance is high but competence does not improve, the evaluation examines practice quality rather than declaring success because every session ran.

What to watch for

Outputs and outcomes are often confused. A completed form is not necessarily changed behaviour. Linear sketches can hide feedback loops and conditions outside HR’s control. Do not force every initiative into a financial impact measure. Keep the model proportionate and update it when evidence shows that an assumption was wrong.

Make assumptions visible

Write a short sentence between each pair of stages: for example, staff can attend practice during paid time, supervisors provide usable feedback, and the task occurs often enough to retain the skill. Identify assumptions with weak evidence and decide how to test them.

Add process and outcome questions

Process questions ask whether the intended activity happened and reached the intended people. Outcome questions ask what changed. An intervention can fail because its theory is weak or because delivery was incomplete. Gathering both kinds of evidence helps distinguish these possibilities.

Choose a small measurement set

For each measure, record the definition, source, timing, owner and burden on participants. Include an unintended outcome, such as workload displaced to colleagues. Avoid collecting a large survey simply because it is available. Each item should help answer an evaluation question.

Use the model during delivery

Review the model at a planned checkpoint. Record changes to the programme and why they occurred. If the intervention changes substantially, be explicit about what version the evaluation covers. The diagram should remain a working explanation rather than a frozen bid document.

Write outputs and outcomes in different language

Outputs describe what was delivered, such as sessions completed or managers reached. Outcomes describe what changed, such as a more consistent feedback practice. If both columns contain training counts, the model has not yet explained the intended change. Use verbs that make the distinction visible. Then ask how the activity is expected to produce the outcome and what conditions must hold. Those assumptions guide both implementation and evaluation.

Add the people who must do something differently

For an illustrative onboarding programme, HR may create materials, managers may schedule practice, buddies may provide support and new starters may ask questions. Identify the behaviour required from each group. A programme can fail because one participant lacks time or authority even when every planned document is produced. Include these dependencies and the resources they require, rather than drawing a straight line that makes change appear automatic.

Choose evidence before delivery starts

Decide how you will know the programme reached people, was used as intended and produced a meaningful result. For onboarding, this might include access readiness, demonstrated task competence and new-starter experience. Agree the follow-up period and a comparison where feasible. Do not wait until the end and select whichever available metric improved. Record what would count as an unintended harm, such as overloading experienced staff who provide support.

Revise the explanation when evidence disagrees

If attendance is high but behaviour is unchanged, examine delivery quality, relevance and opportunity to apply learning. If behaviour changes but the final outcome does not, investigate the later link or allow an appropriate observation period. A logic model is a working explanation, not a document to defend. Updating it can be a sign of learning, provided the changes and reasons are recorded rather than retrospectively rewriting the original success criteria.

Make the assumptions between boxes explicit

For a supervisor-development programme, training sessions are outputs; improved feedback behaviour is an outcome; better team performance is a further hoped-for effect. Write what must be true for each transition: managers need practice, time and permission to use a different approach, while employees need a chance to respond. Identify external influences such as staffing or workload that could interrupt the chain. The diagram should reveal uncertainty rather than make a weak explanation look complete.

Use the model to decide what evidence to collect

Collect information at the point where the explanation is most uncertain. If managers attend but do not change their conversations, another satisfaction survey is unlikely to resolve the problem. Observe a suitable sample of practice or ask employees about specific behaviours, using appropriate consent and confidentiality. Agree when the evidence would justify continuing, adapting or stopping the intervention. A logic model organises the evaluation; it does not establish that the proposed causal links are true.

Take it into your next conversation

Three useful questions.

  1. Which decision will this analysis change?
  2. Are the population, denominator, time period and assumptions explicit?
  3. What alternative explanation or unintended effect must we check?

Related terms

Go to the evidence

Sources & attribution

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] UK government: Creating a logic model for an intervention ↗

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.

Explore People Analytics →

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 2026-09-22. Editorial approach

Keep exploring

A little more context.

People analytics · practical playbook01

From a business question to a measured outcome

A complete worked route through casual-worker retention, from definitions to a fair pilot and an honest conclusion.

ADM HR editorial workflow6 min
People analytics · technique02

Cohort analysis: compare like starting points

A cohort is a group defined by a shared starting event or period, such as joining in a particular quarter.

Statistical and evaluation practice5 min