The new Magenta Book: Improving evaluation in Government

HM Treasury (HMT) and the Evaluation Task Force^[1] (ETF) have published a major update to the Magenta Book^[2], the UK Government’s central guidance on evaluation. In a blog to accompany the launch, the Civil Service^[3] described it as a major shift in government evaluation standards, with a new focus on learning and adapting in real time, rather than simply attempting to assess the outcomes of a policy or intervention towards the end.

For organisations such as the Greater London Authority (GLA), which place evidence at the centre of policymaking, the Magenta Book is considered the key guidance document for analysts and officers on how to effectively evaluate projects and programmes. In addition to the new ‘test and learn approach’, this version contains many practical tips and advice on how to improve the quality of evaluation, making it more accessible to a wider range of practitioners.

Why the Magenta Book matters

The Magenta Book is the UK Government’s authoritative guidance on evaluation. It provides a framework for understanding whether policies, programmes and projects are working, for whom they are working, and whether they represent good value for money. The guidance emphasises that evaluation serves two fundamental purposes: learning and accountability. Evaluation helps improve interventions during delivery, as well as providing evidence to justify public spending and inform future decisions.

At a time of tight public finances, the need for robust evidence on impact and value for money becomes all the more important. Improvements to the Magenta Book, which help to guide practitioners in the pursuit of this evidence, are therefore welcome.

‘Test and learn’ approaches

An important message in the new guidance is the stronger emphasis on learning during implementation rather than evaluation only after a programme has concluded.

While the previous version referred to evaluation being both formative (during delivery) and summative (at the end), the updated Magenta Book gives greater prominence to “Test and Learn” approaches. A dedicated annex has been added to the guidance, positioning Test and Learn as a way of working that uses social research and evaluation methods to adapt and optimise interventions before large-scale implementation. Planning evaluation early can allow interventions to be designed in ways that maximise learning opportunities and enable evidence to be used in real time.

For local and city government, this shift is particularly valuable. Many policies are complex, rapidly changing and characterised by uncertainty. In these circumstances, programme managers benefit from timely evidence on what is and is not working rather than waiting years for a final evaluation report. The stronger focus on experimentation, piloting and iterative improvement therefore represents one of the most useful innovations in the revised guidance.

Greater emphasis on value for money

GLA Economists read with particular interest the new sections on value for money (VfM) evaluation. The redrafted guidance provides expanded methodologies for assessing how effectively public resources are being used. This draws on both HMT Green Book and National Audit Office approaches to assessing VfM.

Consistent with the Green Book supplementary guidance on VfM^[4], the Magenta Book reminds practitioners that VfM is a judgement that can be informed by different types of analysis and both quantitative and qualitative evidence. The Green Book Review^[5] made clear that practitioners should not place undue emphasis on the Benefit Cost Ratio and should make a rounded assessment of VfM, including other factors such as whether strategic objectives were met.

Many public sector interventions generate social outcomes that cannot always be easily monetised. This is particularly relevant for London. Many programmes seek to improve community cohesion, partnership working, or long-term economic opportunity. While these outcomes are important, they are not always straightforward to monetise and build into a VfM evaluation. The new guidance broadens the range of approaches available to evaluators when assessing costs and benefits.

Better integration with business cases

One of the less prominent but helpful additions concerns a strengthening of the link between evaluation and business cases. It is surprisingly common for evaluations to make very little, if any, reference to the original business case for a project or policy. There are various reasons for this, a very practical one being that evaluations often occur many years after the original business case was developed and are typically undertaken by different contractors or departments. However, testing whether the original assumptions in a business case were valid should be a key part of the policy-making cycle. The guidance urges evaluators to look back at the original business case and to feed findings from evaluations into future appraisals.

There is also new content explaining how evaluation and benefits management complement one another. It distinguishes between benefits management's focus on project delivery and the tracking of expected benefits that were in the business case, from evaluation's role in systematically assessing outcomes and whether they can be attributed to the intervention. There can sometimes be some confusion between these activities. The new guidance clarifies these distinctions and encourages a more integrated approach.

A stronger commitment to transparency

Transparency is another area that receives greater attention. The revised Magenta Book introduces new principles promoting research transparency, including recommendations on pre-registering evaluation protocols and sharing analytical code through open-source platforms such as GitHub.

These developments align with broader trends across social science and public sector research. Greater transparency can help strengthen confidence in evaluation findings by making methods, assumptions and analytical processes more visible. It also supports learning across organisations by allowing practitioners to build on previous work. This should ultimately contribute to higher-quality evaluations and more effective knowledge sharing across government.

Addressing artificial intelligence and emerging methods

The guidance includes a new section on the responsible and ethical use of artificial intelligence in social research and evaluation. As AI becomes increasingly important in data collection, analysis and evidence synthesis, the Magenta Book now provides a framework for considering both the opportunities and risks associated with these technologies.

Rather than treating AI as a specialised technical topic, the guidance integrates it into broader discussions about quality, ethics and methodological rigour. This is an important development because many organisations are currently experimenting with AI-enabled tools while simultaneously seeking to maintain transparency, accuracy and public trust.

The inclusion of AI-related guidance demonstrates the Magenta Book’s intention to remain relevant as evaluation methods continue to evolve.

Recognising place-based policymaking

Place-based policymaking has moved further up the political agenda since the appointment of the new Prime Minister. The Treasury is in the process of finalising guidance on place-based business cases so the new section on place-based analysis was perhaps unsurprising. This acknowledges that many public policies operate differently across places and that local context often matters significantly for outcomes and effectiveness.

While the updated guidance provides valuable advice for evaluators working in local and regional government, it highlights many of the challenges for robust evaluation of place-based initiatives. Not least the issue of quantifying spillover or displacement effects to neighbouring areas and isolating the effects of a specific policy or project through a robust counterfactual.

More accessible and practical guidance

While many of the headline changes are methodological, one of the most significant achievements of the update may be its accessibility. The revised Magenta Book explicitly aims to serve a broad audience that includes policymakers, analysts, delivery professionals and finance specialists. The foreword describes the guidance as a shared roadmap across professions rather than a document intended solely for evaluation specialists.

The guidance explains not only what good evaluation looks like but also how evaluation can be planned, managed and used effectively. In GLA Economics’ experience, these practical tips around process are just as important as the more technical methodological guidance. Good evaluation depends not only on methodological expertise but also on organisational culture, programme design and effective collaboration between analysts and decision-makers.

Summary

The 2026 update to the Magenta Book represents an important evolution in the UK Government’s approach to evidence and evaluation. Its stronger focus on learning, experimentation, value for money, transparency, place-based policy and emerging technologies reflects developments in evaluation practice over the past decade.

From the perspective of GLA Economics, these changes are welcome. The revised guidance is more accessible, more practical and better aligned with the realities of contemporary evaluation. It encourages evaluation to be embedded early in programme design and delivery rather than treated as an afterthought. It also provides useful tools and approaches that can help raise the quality of evaluation across government.

As London continues to address complex economic and social challenges, high-quality evaluation will remain essential for understanding what works, improving policy effectiveness and ensuring good stewardship of public resources.