The Biggest Shake-up to Government Evaluation in Six Years
In May 2026, HM Treasury published its first major revision of the Magenta Book since 2020. Led by the cross-departmental Evaluation Task Force, the update represents the most significant shift in government evaluation standards in six years, moving the emphasis from retrospective measurement towards real-time learning during delivery. For any organisation involved in public programmes, this raises the bar for how evaluation is planned, conducted, and utilised.
What is the Magenta Book?
The Magenta Book is the UK Government’s central guidance on evaluating policies, programmes, and projects. It complements HM Treasury’s Green Book (which covers appraisal and business cases) to form a continuous cycle. The Green Book determines whether an intervention should proceed; the Magenta Book governs how effective an approved intervention is.
Once a policy or project is underway, the Magenta Book ensures that there is a robust approach to evaluating whether it works and why; who the target audience is; at what cost; and what lessons can be learnt for the future. The guidance applies to all government bodies, including private organisations that provide services to the government.
What has Changed: The Six Big Shifts
The 2026 update introduces or substantially expands six areas [1]:
Value for Money (VfM): The guidance broadens VfM assessment beyond traditional cost-benefit analysis, incorporating methods such as cost-utility analysis, social return on investment (SROI), the 4Es framework (Economy, Efficiency, Effectiveness & Equity), and rubric-based approaches. The core message is that VfM must reflect social and non-monetisable outcomes, not just financial returns.
Test and Learn: A new annex formalises iterative, small-scale testing with built-in evaluation feedback loops. Teams are encouraged to trial interventions in real-world settings, combining quantitative and qualitative evidence to refine before scaling. The Cabinet Office's £100m Test, Learn and Grow programme, which pilots new public service approaches across ten locations in England before scaling what works, is a live example of this principle in practice [2].
Benefits Management and Evaluation: The update distinguishes between benefits management (tracking whether planned benefits are being realised) and evaluation (assessing whether the intervention caused those outcomes, and why). Both should be planned from the get-go and done simultaneously instead of sequentially.
Transparency and Open Science: Pre-registration of evaluation plans is now encouraged, alongside sharing methodologies, code, and data where appropriate. In practice, this involves setting out evaluation questions, methods, and intended outcomes pre-delivery to reduce bias and strengthen accountability.
Artificial Intelligence: New guidance addresses the ethical use of AI in evaluation, emphasising human oversight, bias mitigation, transparency, and strict controls on the use of personal or sensitive data.
Place-Based Evaluation: For interventions targeting specific regions or communities, the guidance introduces expectations around place-based theories of change, geographic spillovers, and cross-project evidence synthesis.
Why Should You Care?
These changes carry practical implications for not just government organisations but also private sector organisations that are engaged to support the delivery of public programmes and initiatives.
Most fundamentally, evaluation can no longer bean afterthought. The guidance is explicit that evaluation plans, data collection strategies, and baselines should be built into programme design from the outset rather than bolted on in-life. Alongside this, value for money expectations have broadened, so organisations will increasingly need to evidence social and qualitative impacts as well as financial performance.
Transparency is also becoming a baseline requirement, with methods, assumptions, and findings expected to be documented to a standard that supports external review. At the same time, iterative delivery is now formally endorsed. The Test and Learn framework provides a structured basis for piloting, adapting, and scaling, which is an approach already familiar in transformation and service design but now codified in government practice. With the Green Book and Magenta Book more closely aligned, the appraisal-evaluation chain is tightening, so consistency between business case assumptions and evaluation criteria will need to be maintained more explicitly through delivery.
The Bottom Line
For organisations looking to respond effectively to the change, five priorities stand out.
Plan evaluation and benefits tracking from day one: Define success measures, data requirements, and attribution methods before delivery begins.
Adopt “Test and Learn” where uncertainty is high: Where outcomes are uncertain, pilot interventions on a small scale, evaluate the results, adapt the approach in response, and only then scale up what demonstrably works.
Capture the full picture of value: Go beyond financial metrics to include social, qualitative, and place-based outcomes.
Maintain methodological transparency: Document and, where appropriate, share evaluation approaches to support credibility and learning.
Use AI as a tool, not a substitute: Ensure human oversight, ethical governance, and clear communication of limitations.
The 2026 Magenta Book update marks a clear step forward in the government’s approach to evaluation. For organisations working in or alongside the public sector, staying aligned with this framework is both a compliance requirement and an opportunity to deliver better, more accountable outcomes.
Deecon Consulting has extensive experience in building evaluation frameworks and theories of change into programme design from the start, ensuring the right data, baselines and success measures are in place. We help ensure that plans meet government standards and translate into practical, accountable improvement.
Written by Afeef Alam
Edited by Kate Randall