Methodology
The GROWTH Model® is a data model for learning and development, built around the learner. It groups the outcomes of learning into five categories — behaviour change, skills change, culture change, human network growth and performance change — and standardises how those outcomes are captured across topics, modalities and providers.
In brief
Kirkpatrick, LTEM, Phillips ROI and the Success Case Method are frameworks: ways of thinking about evidence, applied one programme at a time through bespoke analysis. The GROWTH Model® is a data model. It defines what is collected, how it is structured and how it relates, so learning outcome data is standing, standardised and always on. Because the data is standardised, analysis can be automated, outcomes can be compared across every programme an organisation runs, and the same data set can power content personalisation and academic modelling. Implemented in The GROWTH Platform™, it surfaces correlation continuously and supports controlled trials to demonstrate causation where it exists.
The established approaches to learning measurement are frameworks. A framework is a way of thinking about evidence: it tells you what categories of evidence exist and roughly what counts as strong or weak. It does not tell you what data to collect, what shape to store it in, or how one programme's results relate to another's.
The practical consequence is that every evaluation is a bespoke project. Someone decides what to ask, designs an instrument, gathers data, analyses it by hand and writes it up. That work is repeated for the next programme, usually with different questions, so the two results cannot be compared. Most organisations can only afford to do it for a small number of flagship programmes, and by the time the analysis lands the decision has often already been made.
The GROWTH Model® takes a different approach. It is a data model: a defined structure of entities, attributes and relationships, centred on the learner. Because the structure is fixed and the data is standardised, outcome data accumulates continuously as learning happens, and analysis runs against the standing data set whenever it is needed. It is always on. There is no evaluation to commission, because the measurement is not an event.
This is why the GROWTH Model® is a replacement for the legacy methodologies rather than an addition to them. It answers the questions those frameworks pose, but it also answers the question they leave open: what data should be collected in the first place. It tells learning designers explicitly what to capture at the design stage, and it makes the resulting analysis automatable.
Every outcome the model captures belongs to one of five categories. Fixing the categories is what makes a leadership programme and a compliance module comparable on the same terms.
Observable change in what people actually do in the work context after an intervention: the questions a salesperson asks, the way a manager runs a one-to-one, the method an engineer uses to diagnose a fault. Behaviour change is captured as structured, comparable data rather than as a narrative written up after the event.
Change in the capability a population holds — knowledge, skill, judgement and decision-making — measured against what the intervention set out to build. Skills change is distinct from attendance, completion, confidence or intention, and is recorded per learner so that skills growth can be tracked over time and across programmes.
Change in whether people feel able to learn, apply and experiment: psychological safety, time, support, access to content and manager backing. Culture is captured as a standing measure across employee populations, so movement can be seen by region, function and demographic rather than only inside a single programme evaluation.
Change in who people turn to for help and how knowledge moves through an organisation. Human network data surfaces hidden influence, isolation, flight risk and the informal structures that determine whether new capability spreads or stalls.
Change in measurable execution and organisational results — productivity, quality, conversion, customer outcomes, delivery, safety, retention, revenue. Performance data is drawn from the metrics the organisation already runs on, and joined to the learner record so that learning data and performance data sit in the same structure.
The learner sits at the centre. Interventions attach to learners; outcomes attach to learners; organisational context attaches to the population around them. Everything is recorded in a consistent shape so it can be joined and compared.
INTERVENTION
(topic · modality · provider · design)
|
v
+-----------+
| LEARNER |
+-----------+
|
+---------+---------+-+-------+---------+
| | | | |
BEHAVIOUR SKILLS CULTURE HUMAN PERFORMANCE
CHANGE CHANGE CHANGE NETWORK CHANGE
GROWTH
surrounded by CONTEXT:
opportunity · environment · management ·
technology · incentives · other interventionsEvery construct the model uses is defined explicitly, so that what is being recorded is unambiguous.
The learner is the centre of the data model. Every record — an intervention experienced, a skill assessed, a behaviour observed, a network tie, a performance metric, a culture response — attaches to a person. This is what makes the model a data model rather than an evaluation framework: it is a persistent structure that accumulates, not a study that is designed, run and closed.
Any deliberate activity intended to develop capability: formal training, coaching, leadership development, performance support, digital learning, workflow interventions and blended designs. Interventions are recorded with a consistent shape — topic, modality, provider, design, population — which is what allows one intervention to be compared with another.
Outcome data is captured in the same shape regardless of topic, modality or provider. A leadership programme delivered face-to-face by one supplier and a compliance module delivered digitally by another produce data that can sit side by side in the same table and be compared on the same terms.
The organisational conditions that determine whether capability becomes behaviour and behaviour becomes performance: opportunity, environment, management, technology, incentives and other interventions running in parallel. Context is held as data alongside outcomes, so that a programme which produced capability but no performance change can be explained rather than simply marked as a failure.
Because the data is standing and standardised, relationships between interventions, behaviour, skills, culture, networks and performance can be surfaced continuously and automatically, across the whole portfolio, without commissioning an analysis.
Correlation is where the model starts, not where it stops. The GROWTH Platform™ is designed to support controlled trials — comparison groups, matched cohorts, staged rollouts, difference-in-differences and interrupted time series — so that causal claims can be made where the design supports them, and contribution claims made where it does not.
Learning outcome data is standardised across topics, modalities and providers. That single decision is what turns evaluation output into a usable data asset.
Senior learning leaders can compare outcomes across every programme they run — by provider, by modality, by course design, by topic and by learner demographic — on a single standard, without commissioning an analysis for each one. Spend can be moved towards what demonstrably produces outcomes.
A standardised outcome data set is a clean, structured signal about what works, for whom, in what context. That signal can power content creation and recommendation tools, so what an individual is offered next is shaped by outcome evidence rather than by consumption history alone.
Standardised, longitudinal, learner-level outcome data across many organisations is the raw material research in this field has largely lacked. The same data set that answers an operational question can support formal modelling of how learning affects organisations.
Because the model specifies the five outcome categories and the shape of data in each, learning designers know at the design stage exactly what to collect. Measurement stops being something bolted on afterwards by someone else.
The GROWTH Model® is the data model. The GROWTH Platform™ is the Gallus Insight software that implements it. The platform captures standardised outcome data across the five categories, holds it against the learner record, joins it to organisational performance data and runs the analysis automatically.
Because the data structure is fixed, the analysis does not have to be rebuilt each time. Correlations between interventions, skills, behaviour, culture, networks and performance are surfaced continuously across the whole portfolio — which providers produce behaviour change, which modalities hold up over time, which populations convert capability into performance and which environments prevent them from doing so.
Correlation is where the platform starts, not where it stops. It is also designed to support controlled trials: comparison groups, matched cohorts, staged rollouts, difference-in-differences and interrupted time series. Where an organisation needs to demonstrate causation and the design supports it, the platform is built to run that design and hold the result in the same structure as everything else.
Correlation: “These outcomes move together across the portfolio.”
Contribution: “The available evidence indicates the intervention contributed materially to the outcome.”
Causation: “A controlled design shows the intervention caused the change.”
A standing data set makes the first of these continuous and cheap. Organisational outcomes generally have multiple causes, so the model is explicit about which of the three claims the evidence supports, and reports accordingly.
Designs the model supports for stronger claims include:
Not every intervention warrants a controlled trial. Because the underlying data is already being captured, the additional cost of running one where it matters is the design work, not the measurement.
Each of these contributed something real to the field. The difference is one of kind, not of quality: they are frameworks for analysing a programme, and the GROWTH Model® is a data model for running a portfolio.
Kirkpatrick's four levels gave L&D a shared vocabulary for moving beyond satisfaction, and it remains the most widely understood evaluation framework in the field. It is a classification of evidence, applied per programme by whoever is running the evaluation. The GROWTH Model® defines the data instead: what is collected, in what shape, against which learner. Kirkpatrick-style reporting can be read out of GROWTH data automatically, without designing an evaluation.
LTEM is rigorous about distinguishing stronger and weaker forms of learning evidence, and that rigour informs how skills change is captured in the GROWTH Model®. LTEM stops at the quality of learning evidence; the GROWTH Model® carries the learner record forward into behaviour, culture, network and performance data and holds it over time.
Phillips ROI produces a financial figure through a bespoke isolation and costing study per programme — thorough, but slow and expensive enough that most organisations apply it to a handful of programmes. In the GROWTH Model®, financial return is one output computed from performance-change data already in the data set, available across the portfolio rather than for selected flagships.
Success-case work explains, vividly, why an intervention worked for some people and not others. It relies on someone selecting and writing up cases by hand. The GROWTH Model® identifies outlier cases automatically from standing data; success-case interviewing then remains a good way to explain what the data has already flagged.
Learning analytics describes consumption inside the learning system — logins, completions, time on task — and rarely reaches behaviour, culture, networks or business performance. Activity data is one input to the GROWTH Model®, which extends the structure beyond the learning system to the five outcome categories and the organisation's own performance metrics.
| Approach | Type | Primary purpose | Limitation | Relationship to the GROWTH Model |
|---|---|---|---|---|
| Kirkpatrick | Evaluation framework | Classify evaluation evidence across four levels: reaction, learning, behaviour, results. | Tells you what kinds of evidence exist, not what data to collect or how to store it. Every programme is evaluated as a separate exercise, and results are not comparable between programmes. | The GROWTH Model® replaces the per-programme evaluation exercise with a standing data structure. Kirkpatrick's levels can be read out of GROWTH data, but no separate evaluation design is required. |
| LTEM (Thalheimer) | Evidence taxonomy | Distinguish stronger and weaker forms of learning evidence and avoid weak assumptions about learning. | Concerned with the quality of learning evidence rather than with organisational outcomes, and offers no data structure for holding evidence over time. | LTEM's rigour informs how skills change is captured in the GROWTH Model®; the model then carries that data forward into behaviour, culture, network and performance categories. |
| Phillips ROI | Evaluation methodology | Convert the value of an intervention into a financial return on investment. | Requires a bespoke isolation and costing study per programme. Expensive, slow, and typically applied to a handful of flagship programmes rather than a portfolio. | Financial return is one output the GROWTH Model® can produce from performance-change data, computed from the standing data set rather than commissioned as a study. |
| Brinkerhoff Success Case Method | Qualitative evaluation method | Identify the most and least successful cases to understand what made an intervention work and for whom. | Illustrative by design. Cases are selected and written up manually, and cannot be generalised to a population or repeated at portfolio scale. | The GROWTH Model® identifies outlier cases automatically from standing data; success-case interviewing then explains why those cases behaved as they did. |
| Learning analytics | Activity data | Collect and analyse learning activity and engagement data, often at scale and in real time. | Describes consumption — logins, completions, time on task — inside the learning system. It rarely reaches behaviour, culture, networks or business performance. | Activity data is one input to the GROWTH Model®. The model extends the data structure beyond the learning system to the five outcome categories. |
| The GROWTH Model® | Data model | Define, standardise and hold learning outcome data around the learner across five outcome categories. | Requires consistent data capture across providers and programmes; empirical validation as a distinct data model is still in progress. | Replaces the per-programme evaluation exercise with a standing, always-on data set and automated analysis, implemented in The GROWTH Platform™. |
These are stated as characteristics of the model that can in principle be tested, not as proven advantages.
Illustrative only. The numbers below are invented to show how the data model is applied; they are not the results of a real evaluation.
A sales organisation introduces a consultative selling programme. Under a legacy framework this would become a one-off evaluation project. Under the GROWTH Model® the data is already being captured, so the work is reading it.
The same data then contributes to portfolio comparison against other sales interventions, and the finding about manager support feeds the design of the next programme.
This section distinguishes what is established, what is emerging and what remains to be tested.
The model draws on established approaches to learning evaluation — Kirkpatrick's four levels, the Phillips ROI methodology, the Brinkerhoff Success Case Method and Thalheimer's LTEM — and on established approaches to performance measurement, organisational network analysis and causal inference, including comparison groups, difference-in-differences and triangulation. These foundations are referenced below.
The model is set out in a published paper by Derek Mitchell (2026). Adoption is growing, with a number of organisations recently adopting the model and case studies to follow as those implementations are documented. Evidence supporting the underlying approaches is evidence for those approaches, not evidence validating the GROWTH Model® as a distinct data model.
Documented organisational implementations will be listed here as they are published. No case evidence is presented on this page yet.
Whether standardised, learner-centred outcome data produces materially better learning investment decisions than bespoke per-programme evaluation is a testable question, and the model is designed so that it can be tested. Longitudinal validation across organisations, and formal modelling of the relationships between the five outcome categories, remain areas for further research.
| Version | Date | Major changes | Evidence / reason for change |
|---|---|---|---|
| v1.1 | 2026 | Restated as a learner-centred data model with five outcome categories; relationship to The GROWTH Platform™ and to controlled trials made explicit. | Clarifies the distinction between a data model and the legacy evaluation frameworks. |
| v1.0 | 2026 | Initial published methodology. | First public statement of the framework and its evidence chain. |
Every substantive methodological change updates the version number and explains why the change was made.
The GROWTH Model® was developed by Gallus Insight, a UK learning measurement and L&D analytics company based in Falkirk, Scotland. The original author is Derek Mitchell, founder of Gallus Insight. It was first published in 2026 and is implemented in The GROWTH Platform™.
The GROWTH Model® is a UK registered trademark owned by Gallus Insight. The trademark governs use of the name “The GROWTH Model®” rather than the underlying data model, which is described openly on this page.
Canonical citation: Gallus Insight (2026), The GROWTH Model: A Data Model for Learning Outcomes, version 1.1.
To discuss applying the model to your programmes, contact Gallus Insight or read the definitions and glossary.
Common questions
The GROWTH Model® is a data model for learning and development, built around the learner. It groups the outcomes of learning into five categories — behaviour change, skills change, culture change, human network growth and performance change — and standardises how outcome data is captured across topics, modalities and providers.
A framework tells you how to think about evidence for one programme at a time; the analysis is designed and commissioned each time. A data model defines what is collected, how it is structured and how it relates, so the data set is standing and the analysis can be automated. The GROWTH Model® is always on: outcomes accumulate continuously rather than being reconstructed after each programme.
Yes. Those frameworks classify evidence after the fact; the GROWTH Model® tells learning designers what data to collect up front and holds it in a structure that supports automated analysis. Their levels and outputs can still be read out of GROWTH data, but the bespoke per-programme evaluation exercise is no longer required.
Behaviour change, skills change, culture change, human network growth and performance change. Every learning outcome the model captures belongs to one of these five categories, which is what makes outcomes comparable across very different programmes.
Because the model defines a persistent data structure rather than a study design, outcome data is captured continuously as learning happens. Analysis runs against the standing data set at any time, without commissioning a new evaluation for each programme.
The GROWTH Model® is the data model. The GROWTH Platform™ is the Gallus Insight software that implements it: it captures the standardised outcome data, automates the analysis, surfaces correlations across the portfolio and supports controlled trials where causal evidence is required.
Standardised outcome data across topics, modalities and providers makes three things possible: senior learning leaders can compare outcomes across every programme they run; the data set can power content creation tools for hyper-personalisation; and the data can be used for academic modelling of how learning affects organisations.
It distinguishes the two. The standing data set surfaces correlation continuously and automatically. Where a causal claim is required, the GROWTH Platform™ is designed to support controlled trials — comparison groups, matched cohorts, staged rollouts, difference-in-differences and interrupted time series — so causation can be demonstrated where it exists.
It removes the guesswork about measurement. Because the model specifies the five outcome categories and the shape of the data in each, designers know at the design stage exactly what to collect, and that data will be comparable with everything else in the organisation.
A standardised outcome data set is a clean, structured signal about what works, for whom, in what context. That signal can drive content creation and recommendation tools, so what an individual is offered next is shaped by outcome data rather than by consumption history alone.
Yes. Any deliberate intervention can be recorded in the model, provided the intended capability is defined. Leadership development and workflow learning are recorded in the same structure as formal training, with the longer timescales that behaviour and performance change require.
The GROWTH Model® was developed by Gallus Insight and is a UK registered trademark owned by Gallus Insight. The original author is Derek Mitchell, founder of Gallus Insight. It was first published in 2026.
The model is set out in a published paper by Derek Mitchell and draws on established approaches to learning evaluation, performance measurement and causal inference. Adoption is growing, with a number of organisations recently adopting the model, and case studies will follow as those implementations are documented.
It is an emerging methodology. Its underlying concepts draw on established approaches to learning evaluation, performance measurement and causal inference, while empirical validation of the GROWTH Model® as a distinct data model remains an area for further research.
Stable links to original framework authors and primary methodological sources are used wherever available.
Last updated: 2026-09-09