This chapter examines how Estonia’s public sector uses impact evaluation, value-for-money and evidence-informed decision-making to drive innovation. It finds that few innovations are currently evaluated, scaled or embedded in core activity, limiting the ability of government to demonstrate the returns achieved from innovation investment. Weak evaluation constrains learning, funding decisions and scale-up, and more broadly undermines the capacity to assess whether innovation spending delivers sustained public value. The chapter shows that evaluation remains ad hoc, that rigorous impact methods are under-used, and that funding timeframes and weak links with academia constrain progress. It recommends that Estonia align evaluation with decision needs, require impact evaluation for maturing innovations, build evaluation planning into innovation projects, strengthen specialist and system-wide evaluation capacity, as well as improve data infrastructure and external partnerships to support more robust, portfolio-level assessment of value for money.
Strengthening the Innovative Capacity of the Public Sector of Estonia
6. Impact and evidence-informed decision-making
Copy link to 6. Impact and evidence-informed decision-makingAbstract
The assessment finds that Estonia’s public sector demonstrates strong ideation capacity and a high rate of piloting. This persists even though overall innovation rates are not that high (see Chapter 1). However, comparatively few innovations are scaled or embedded in policy and service delivery. Survey evidence and interviews suggest that evaluation is the bottleneck. This, coupled with the lack of funding across the innovation lifecycle (Chapter 2), creates an “implementation funnel” (described in detail below).
The failure to evaluate public sector innovations poses a strategic vulnerability. It weakens Estonia’s ability to choose, defend and scale the right innovations. Government needs evaluation to distinguish between promising initiatives and attractive but ineffective ones. Without evidence of impact and value for money, administrations cannot allocate budgets effectively, especially when public resources are constrained. External funders, such as the European Competitiveness Fund, are less likely to invest in reforms without evidence that they will succeed or have succeeded. Sporadic or infrequent evaluation also risks evaluation skills atrophying, undermining Estonia’s long-term capacity to rigorously evaluate its innovations.
Evaluation is most effective when tied to need
Copy link to Evaluation is most effective when tied to needThe OECD has identified how evaluation addresses seven needs that decision makers face (OECD, Forthcoming). Decision-makers in Estonia share these needs for evaluation across the innovation lifecycle:
1. Exploring whether an innovative idea could work. Early in the innovation lifecycle, decision makers must assess which ideas are plausible, fit user needs and policy goals, and are likely to fail. This warrants desk research, literature reviews and feasibility studies.
2. Improving design while the innovation is being built. Early and midway through the lifecycle, decision makers must adapt their innovation. This warrants evaluations that produce quick feedback, such as rapid evaluations, user testing and pilots; often through implementation and process evaluation (Australian Centre for Evaluation, 2023[1]).
3. Demonstrating impact and building trust in innovations. Midway and at the end of the lifecycle, decision makers must know whether the innovation has worked or not. What was its impact? What caused the change in outcomes? Is it worth scaling? Worth proceeding to the next stage? This informs whether to cease or continue an innovation. Experimental and quasi-experimental approaches are often the best tools to answer these questions.
4. Adapting or scaling an innovation to a different context. Even when an innovation succeeds in one location, organisation or user group, it may not automatically succeed elsewhere. When, how and for whom does the innovation work? As implementation conditions change, how might its impact change? These questions are most often addressed through implementation and process evaluations that identify factors influencing delivery and effectiveness.
5. Making the case for funding. Impact in and of itself is seldom a sufficient basis for scaling an innovation: it must often achieve impact cost-effectively. Decision makers ask: Were the outcomes worth the investment? How does this innovation’s cost-effectiveness compare with that of other alternatives? These questions are most often answered using VfM frameworks and cost-benefit analysis, which assess the economy, efficiency and effectiveness of an innovation.
6. Prioritising resources. Evaluation clarifies the trade-offs associated with different resource allocations. Is it worth investing in an untested idea or scaling up a proven innovation? What is the expected impact of one portfolio of programmes compared with another?
7. Supporting system-wide efforts to increase public sector innovation capacity. Governments often make broader investments in the innovation capacity of the public sector, such as training, dedicated teams and frameworks (Kaur et al., 2022[2]). These warrant systematic, rather than discrete, evaluation to understand their impact on the system as a whole.
Choosing how to assess an innovation depends on the questions to be answered. No single approach fits every scenario. Evaluation methods range from experimental and quasi-experimental frameworks that measure causal impacts (internal validity) to qualitative, process and theory-based approaches that explain how and why an innovation works. Financial and resource-focused evaluations such as cost-benefit analysis and value-for-money assessments further aid decision makers by monetising outcomes and assessing efficiency to guide resource allocation.
Box 6.1. Different methods to evaluate innovation
Copy link to Box 6.1. Different methods to evaluate innovationDifferent evaluation methods address different needs. Evaluations that estimate whether innovation caused outcomes to improve address questions of internal validity. Evidence whether an innovation may have the same effect in other settings, user groups, or at larger scale address questions of external validity. Why the innovation improved outcomes, or how it might improve them more effectively, are addressed through qualitative evaluations. No single method is best; but rather addresses different questions. Different methods complement one another in mixed-method approaches:
Experimental methods, of which randomised controlled trials (RCTs) are the most widely cited, estimate the effect of an intervention by randomly allocating participants to treatment (e.g. to participate in the innovation) vs to a control group who receive business as usual. Experimental methods can be modified to take the form of a simple-, rolling-, cluster- or step-wedge RCT. All experimental methods use randomisation to minimise selection bias and establish whether an innovation causes a change in outcomes.
Quasi-experimental methods estimate causation where randomisation isn’t feasible. Techniques such as difference-in-differences, regression discontinuity, synthetic control groups and instrumental variables can construct a non-randomised control group if certain assumptions hold.
Cost-benefit analysis (CBA) monetises the costs and benefits of an innovation. This method informs where to allocate resources and whether to expand, continue, or discontinue an innovation relative to other interventions. Its reliability depends on the quality of the underlying impact evidence, the assumptions used to estimate value, as well as how uncertainty and distributional consequences are treated. It is often informed by other methods, such as (quasi-)experimental methods to estimate the benefit of an innovation.
Process and implementation evaluation studies how an innovation is delivered in practice, including its fidelity to design, its operational efficiency, and the experience of staff and users. It doesn’t, on its own, show if the intervention achieved its outcomes, but it provides operational evidence to understand the impact it is having, and identify blockers to better implementation.
Theory-based evaluation examines the underlying logic of an intervention: the presumed causal chain between inputs, activities, outputs and outcomes, as well as the assumptions and conditions on which that chain depends. This approach is valuable in complex policy environments where outcomes are shaped by multiple interacting factors and understanding the mechanisms of change. However, it cannot evaluate the impact of an innovation as it lacks a counter-factual. Theory-based methods also require a well specified theory of change, which some innovations may lack.
Value-for-money (VfM) evaluation assesses the relationship between the resources committed to an innovation and the results it achieved. VfM assesses an innovation’s economy, efficiency and effectiveness. VfM evaluation is often used by audit institutions, central budget authorities and oversight bodies to inform decisions on resource allocation, programme continuation and public reporting. It is informed by (quasi-)experimental methods and CBA.
Tools, such as the decision tree below, can guide evaluators to the right methods for the right circumstance.
Figure 6.1. The decision tree for selecting an evaluation method
Copy link to Figure 6.1. The decision tree for selecting an evaluation methodThe current state of innovation evaluation in Estonia
Copy link to The current state of innovation evaluation in EstoniaThe OECD heard from stakeholders and survey respondents that evaluation of public sector innovation in Estonia tends to be ad hoc and project-driven. Overall, few innovations are evaluated for impact; innovation projects do not consistently report standardised indicators such as coverage or evaluation rates. There was no indication of a system‑wide view of which innovations have been evaluated, how or with what effect. Where evaluation occurs, it is selective and episodic. It is tied to individual projects, often those supported by external or project‑based funding, rather than embedded in day‑to‑day innovation activity across organisations. Evaluation is driven more by project demand than by pre-defined institutional arrangements. As a result, learning remains localised and evidence rarely accumulates across projects or organisational boundaries.
Innovations often do not progress beyond prototyping, never reaching evaluation
A recurring issue is the lack of systematic evaluation and knowledge sharing in Estonia. Public officials were reported to often rely on gut feelings or superficial data to judge innovation success. High staff turnover and a lack of institutional memory impede incorporation of new evidence, with the same studies and analyses being repeated. Contributing factors include short project timelines, risk aversion, an over-reliance on internal quantitative, cross-sectional registry-based data, and a reluctance to conduct rigorous longitudinal evaluations. The lack of longitudinal or causal data was identified as a blind spot which even linking existing public sector data could not address.
Figure 6.2. The latest recent innovation stage reached by innovation projects
Copy link to Figure 6.2. The latest recent innovation stage reached by innovation projects
Source: OECD based on the Public Sector Innovation Capacity Survey (2026).
The bottleneck contributes to the fact that only a minority of innovations are evaluated (Figure 6.2). The innovation lifecycle suffers from a “funnel effect”: while 62% of the reported ‘most significant’ innovations reach implementation, only 33% are ever scaled (meaning this is the de facto write-off rate on innovation investment). Most projects stall at the prototyping, evaluation or decision stages. Furthermore, there is a noted disconnect from academia; research and research-based institutions are rarely cited as sources of innovation ideas (5%), suggesting untapped potential for evidence-informed policy design.
Experimental, cost-benefit and value for money evaluations are under-used
When innovations are evaluated, they are predominantly process-based. The OECD survey found that 40% of all innovation evaluations were process-based according to heads of organisations and 48% according to staff (Figure 6.3). VfM was used in only 35% of evaluations according to heads of organisations and 23% according to staff, but the number of overall evaluations was very low to begin with and should be taken into account in interpreting the results. Theory-based evaluation was also common: 30% according to heads of organisations and 23% according to staff (see Figure 6.3 and Figure 6.4). This is not unusual in public sectors, where process and theory-based evaluation tend to be the global norm. Frontline innovators are often responsible for understanding a process and refining what already exists. Heads of organisations set high-level strategy based on an intervention logic articulated in a theory of change. However, these methods are better suited to evaluations where the impact of the activity is already well established. Theory and process evaluation address only the first two needs for innovation evaluation: 1) exploring whether an innovative idea could work; and 2) improving design while the innovation is being built. However, needs 3 to 7 are under-addressed.
Figure 6.3. Types of evaluations of innovation projects reported by staff
Copy link to Figure 6.3. Types of evaluations of innovation projects reported by staff
Source: OECD based on the Public Sector Innovation Capacity Survey (2026).
Figure 6.4. Types of evaluations of innovation projects reported by heads of organisations
Copy link to Figure 6.4. Types of evaluations of innovation projects reported by heads of organisations
Source: OECD based on the Public Sector Innovation Capacity Survey (2026).
Experimental, quasi-experimental, CBA methods are not used as often as they could be. Cost-benefit analysis was used in 9% of evaluations according to staff and 25% of cases according to heads of organisations. Quasi-experimental methods were used in 25% of evaluations according to heads of organisations and 9% according to staff; experimental methods in even fewer instances. These patterns are not unusual compared with peer governments, but they nonetheless represent a missed opportunity for Estonia to ensure that innovation achieves positive outcomes and that scarce resources are allocated where they have the greatest impact.
This creates a particular gap in Estonia’s innovation ecosystem. Innovations benefit disproportionately from experimental and quasi-experimental methods to establish what impact the innovation had, as well as CBA or VfM evaluation to inform whether the innovation is an effective use of public resources. This lack of rigorous impact data makes it difficult to justify scaling successful pilots or to secure long-term funding.
There is little to no impact evaluation at the portfolio or system level
The lack of impact evaluation is echoed across Estonia’s overall pattern of evaluation: Tõetamm, the national Tree of Truth dashboard, publishes development-plan indicators; ministries run their own monitoring; and external bodies review individual programmes. While some policies are evaluated, no programme is judged as a portfolio, and the system as a whole goes unassessed (see discussion on the need for portfolio level innovation in Annex 6.A). This is also reflected on the institutional level: for example, in its November 2024 annual report on five years of performance-based budgeting, the Riigikontroll (the National Audit Office) found that Estonia’s 2020 State Budget Reform had met none of its three goals, that the budget conversation still runs on traditional cost lines rather than outcomes, and that around 2.5 billion euros, about 13% of the 2023 state budget, went unused without a documented analysis of why (Riigikontroll, 2024[4]).
Even at the project level the depth of evaluation is uneven. Some initiatives are evaluated thoroughly while many are recorded only as activity. What is missing above the project is the aggregating layer: no mechanism assembles the cost and outcome of a set of innovation projects into a portfolio-level judgement, and none aggregates portfolios into a system-level one. A portfolio-level value-for-money assessment would have to be built on a budgeting system that does not yet reason in those terms. As a result, governments are constrained in their ability to make strategic allocation decisions, to distinguish between high- and low-performing investments, and to build a credible evidence base to justify innovation spending over time. This challenge is not unique to Estonia and is observed across OECD member countries. In response, the OECD is working in partnership with the Government Office of Estonia to generate empirical evidence and develop practical approaches to assessing the portfolio-level value-for-money of government innovation investments.
The current state of VfM evaluation practices in Estonia
The OECD analysed Estonia’s government innovation initiatives and their readiness to support value-for-money assessments. Across innovation programmes, evaluation practices focus on early stages, with more limited tracking of outcomes and no consistent approach yet to assessing value for money at the portfolio level. As seen in Table 6.1, the Government Office’s Public Sector Innovation Fund (Chapter 2 Box 2.2) has financed two cohorts of projects and carries an indicator-based assessment, though no monetised outcome; it is also reportedly being wound down, which makes it a closed empirical anchor rather than a future-facing programme. Accelerate Estonia, a regulatory-innovation lab in the Ministry of Economic and Communications, conducts portfolio-level monetised assessment, based on a cost-benefit analysis methodology developed by RAKE and the University of Tartu (Accelerate Estonia, 2025[5]). Innovation Procurement runs through the Riigihangete public-procurement portal but records only tenders above the 30,000-euro publication threshold. The Innovation Development Projects line, which covers the Sprint programme, the Nudgeathon and the Experimentation Lab, records participation without measuring outcomes. Table 6.1 sets out what each programme measures, at what stage, and against what definition of value.
Table 6.1. Review of evaluation practices across Estonia’s government innovation portfolio
Copy link to Table 6.1. Review of evaluation practices across Estonia’s government innovation portfolio|
Estonia’s government innovation portfolio |
Measurement |
Stages |
Value definition |
|---|---|---|---|
|
Government Office Innovation Fund: Central funding programme for early-stage public sector innovation projects. |
|
|
|
|
Accelerate Estonia: Regulatory innovation programme testing market-creating solutions with the private sector. |
|
|
|
|
Innovation Procurement: Public procurement mechanisms to promote and purchase innovative and experimental solutions. |
|
|
|
|
Innovation Sprints (Innosprint): Short-cycle programme supporting rapid ideation and experimentation in public agencies. |
|
|
|
|
Nudgeathon: Behavioural insights programme designing and testing small-scale policy nudges. |
|
|
|
Almost no programme can state the return on investment of its project portfolio. Projects are scrutinised when they apply and receive funding. Reporting during delivery is limited, and structured evaluation at completion is rare. Terminated projects seldom receive an exit review. Outcomes are monetised in only one programme, and even there the analysis omits the programme’s own cost. No programme evaluates projects by maturity stage, so an early-stage pilot and an operational rollout are judged against the same expectations. This penalises the early-stage project for outcomes it cannot yet show and sets too low a standard for mature projects. No baseline measurements are taken at the point of award, so before-and-after comparison is limited. Benefit realisation goes untracked once a project closes. No shared infrastructure aggregates costs and outcomes across programmes. Innovation procurement below the publication threshold also remains invisible to any central record.
Together these gaps mean that even where individual projects report results, the evidence cannot be assembled into a defensible statement about what a programme, or the portfolio as a whole, returned for the money spent. To demonstrate the possibility of value for money evaluation at a programme level, the OECD in partnership with the London School of Economics (LSE) undertook a preliminary analysis of Public Sector Innovation Fund projects – the results of the work are described in Box 6.2 below.
Box 6.2. Measuring systemic and long-term outcomes of investment in public sector innovation
Copy link to Box 6.2. Measuring systemic and long-term outcomes of investment in public sector innovationThe OECD and the London School of Economics and Political Science (LSE) collaborated to address a critical gap in evaluating public sector innovation funds in Estonia. Existing approaches have focused on short-term inputs and processes rather than tracking multidimensional public value and systemic, long-term outcomes. The project developed a practical, replicable fund-level evaluation framework and toolkit for assessing how innovation investments translate into sustained, system-level change and public value.
Framework
The framework organises public sector innovation outcomes into seven dimensions across three domains, creating a logic of systemic transformation. The operational domain encompasses productivity of services (how innovations improve resource-to-service efficiency ratios) and efficacy (how effectively a programme achieves its intended outcomes). The strategic domain includes economic impact (generating wider economic effects beyond the operational domain) and strategic purpose (addressing relevant societal challenges requiring new approaches). The societal domain captures long-term systemic outcomes: culture of innovation (building institutional norms that encourage risk-taking and learning), trust (enhancing citizen perception of government responsiveness), and equality (establishing fairness as a foundation for advancement in human capital and technology).
To make this operational, the project delivered a qualitative self-assessment instrument. The instrument translates the seven dimensions into 30 indicators and 58 assessment questions on a scale from 1 to 4 based on evidence strength. The framework operates on a sequential gate logic: a fund must achieve a minimum average score of 3 (Good) in the operational domain before strategic outcomes can be credibly claimed and must clear the strategic threshold before societal impacts can be evaluated.
Pilot
The framework was empirically piloted on the Estonia Public Sector Innovation Fund. The evaluation analysed 27 innovation projects spanning nine policy areas, representing €28.42 million from applications between 2023 and Spring 2025. The test evaluation was conducted at the fund level, not individual projects, using evidence from application forms, final reports, and government websites.
Results
The pilot yielded an overall portfolio score of 2.5 out of 4, meaning the Fund has not yet reached the operational threshold needed to validate its evidence on long-term societal value. Results varied markedly due to portfolio maturity:
Strategic purpose (3.6): The Fund targets ambitious, systemic challenges and supports projects intended to generate transformative public-sector change, which it viewed as evidence of a clear strategic rationale
Culture of innovation (2.6): Senior leaders in organisations participating in the Fund are committed to innovation, although the limited number of completed projects makes it difficult to assess how consistently their support translates into implementation.
Productivity of services (2.4) and Equality (2.2): The portfolio demonstrated capability-building and investments in digital infrastructure. However, evidence on cost reductions and longer-term outcomes remains limited because most projects are still ongoing and Fund-level evaluation of efficacy and cost effects has not yet been completed.
Trust (2.1): The Fund collects implementation feedback from project managers and partners, but evidence of systematic end-user satisfaction measurement was limited.
Efficacy (2.1): Evidence of efficacy was constrained because most projects remain ongoing and few projects had documented theories of change or structured reporting on intermediate outcomes.
The project identified several challenges in evaluating public sector innovation investments:
Portfolio maturity and timing mismatch: evaluating an active fund means that emerging and ongoing projects naturally score poorly on trailing indicators like cost-efficiency or sustained embedding, creating a bias if assessments are tied to rigid reporting milestones.
The innovation productivity J-curve: early-stage technology adoption frequently causes an initial drop in organisational productivity due to training and process redesign before long-term gains materialise, meaning standard value-for-money metrics systematically undervalue early innovation investments.
Intangibility and attribution of societal outcomes: dimensions like public trust and social impact are highly complex and embedded across an entire governance system, making it methodologically challenging to isolate and attribute shifts in citizen perception to a specific innovation.
Monitoring infrastructure gaps: the pilot revealed a lack of monitoring mechanisms during project implementation, combined with an absence of tracking tools to capture critical data such as user service satisfaction or population-level equality of outcomes.
Capturing real-time dynamic risks: traditional indicators focus heavily on final outcomes, thereby failing to capture real-time process risks, dynamic pivots, and emergent developments that occur naturally during the non-linear implementation of innovative projects.
Source: (Susetyo et al., 2026[12])
Estonia should invest in its value-for-money capability, at a pace that its growing evaluation culture can sustain. A staged sequence is more realistic than a single comprehensive reform: a modest but defensible portfolio-level assessment can be established first, and a more demanding system-level assessment can follow once the foundations are in place. The phases are interdependent. Tagging projects by maturity stage and recording their costs is a precondition for portfolio-level assessment; a functioning portfolio assessment within a single programme is, in turn, the basis for credible comparison across programmes; and comparison across programmes underpins any system-level judgement. A phased approach also reflects institutional conditions, since a government still reorienting its budget process towards outcomes cannot adopt a demanding evaluation standard at once. A standard introduced ahead of the capacity to apply it tends to be treated as a formality rather than used to inform decisions.
Structural barriers to evaluation
Copy link to Structural barriers to evaluationTo build a more innovative public administration, Estonia must evolve its evidence-informed policymaking (EIPM) system from a passive mechanism of compliance into an active engine for policy experimentation and continuous learning. While Estonia has successfully embedded foundational components of an EIPM ecosystem – such as the ministerial Science Adviser Network and the regulatory impact assessment (RIA) framework – these assets remain underutilised as drivers of innovation (Koppel et al., 2025[13]). Currently, the system over-indexes on traditional, legalistic ex-ante assessments, while systematically neglecting rigorous ex-post evaluations and experimental policy designs. An innovative government requires an administrative culture that treats policies not as permanent solutions, but as hypotheses that must be dynamically tested, evaluated, and iterated upon. Some of the key issues identified in this assessment are outlined below.
Funding timeframes are a barrier to evaluation
The reliance on structural funds and short-term project funding, which often dictate the scope and nature of research and innovation initiatives, leads to fragmented efforts and limited long-term impact. One interviewee pointed out that, much like general procurement, research procurement typically kicks off at the end of July – right when most people go on holiday. Consequently, with a deadline in early December, what should be a one-year timeframe is effectively compressed into just two months. This leads to a choice between producing mediocre results or abandoning ambitious goals.
There is a disconnect between academia and public sector innovation
While experts found it difficult to identify the original sources of ideas, the survey results showing that only 5% of innovation ideas originate from academia and only 2% from research resonated (see Figure 6.5). This indicates that the public sector is ignoring a significant engine of evidence-informed decision-making and a source for more radical innovations. Research in general is viewed as "too slow" for the political cycle, while policy is viewed as "too shallow" by researchers.
Figure 6.5. Source of innovation ideas in Estonia
Copy link to Figure 6.5. Source of innovation ideas in Estonia
Source: OECD based on the Public Sector Innovation Capacity Survey (2026).
Furthermore, overly prescriptive project scopes were also reported to hinder effective collaboration. Concerns were raised regarding the current reform plan wherein the Ministry of Education and Research shifts responsibility for procuring scientific knowledge from individual ministries to the Estonian Research Council; while this centralisation may improve grant management, it may impact ministries' abilities to work directly with researchers and define their own needs. Furthermore, while the OECD and European Commission recommended to upgrade the role of evidence-informed policymaking suggesting a coordinating role for the Government Office (Koppel et al., 2025[13]), this was not taken up. Currently also the ministerial science advisors’ role is under review.
At the same time, ministerial science advisers are uniquely positioned to bridge the gap between academic research and public innovation. Rather than acting merely as administrators of R&D procurement or isolated subject-matter experts, their role should be explicitly redefined to encompass knowledge brokerage and innovation facilitation. When properly integrated into the strategic core of line ministries, science advisers can transform raw scientific evidence into actionable public sector experiments. For example, by collaborating directly with the Government Office, science advisers can infuse innovation sprints (see Chapter 1 Box 1.2) and other agile design-thinking processes with robust scientific methodologies, behavioural insights and ethnographic data. This could ensure that public sector innovations are not just novel but are structurally grounded in empirical evidence and subject to measurable evaluation metrics.
The gap between academia and the public sector may make it harder to evaluate innovation. Evaluation skills are often technical or specialised, such as evaluation design, data collection, randomisation, coding, statistical modelling, and translating findings into recommendations. These skills can be fostered in the public sector; however, it is not always cost-effective to do so. Evaluation capacity is ‘use it or lose it’ – teams for whom evaluation is not their primary focus may lack the critical mass of evaluations to retain and develop their specialist evaluation skills. Conversely, evaluation demand may outstrip the capacity of a public sector evaluation team. Absent further investments, evaluation units may not be able to evaluate all the innovation projects that would benefit from evaluation.
There may be a need to build a stronger working relationship with academic researchers who can conduct evaluations. This goes beyond relationship building. The government may need to develop framework agreements or evaluation helpdesks between ministries and academic research groups. The framework should address key enablers of collaboration, such as roles, mutual responsibilities and timeframes. Many evaluations are commissioned at short notice and with compressed timelines, meaning that it is essential to have a framework already in place so that the evaluation window is not consumed by procurement. These frameworks are especially valuable for technical skills such as data collection, randomisation, coding and statistical modelling. These skills may be inefficient to retain exclusively within the public sector. Therefore, the public sector might broaden its pool of evaluators by building connections and agreeing frameworks with academics who can support innovation evaluations.
Where evaluation exists, it tends to be siloed
Innovation evaluation in Estonia tends to exist at the margins of the innovation ecosystem. The public sector does not lack ideas, but it does not routinely evaluate them. Evaluation is not conducted consistently, and the evaluation that is conducted underweights impact evaluation, which is essential for understanding whether an innovation worked and should be scaled up. Evidence of impact tends to be project-specific rather than shared systematically, is based on qualitative insights and depends on individual initiative rather than structural support.
Protect experimentation and impact evaluation politically and narratively
Public servants often reported concerns that experimentation and evaluation in the public sector could be misconstrued as “waste” and engender media criticism (See Chapter 3, safe space media guide). This framing of unsuccessful innovations as failures incentivises public servants to avoid risky innovations or simply not evaluate them – it’s less likely an innovation will be criticised as ineffective if no one evaluates whether it’s effective or not.
Protecting impact evaluation requires a clear political and narrative framing that distinguishes between responsible testing and full-scale implementation. Senior leaders play a critical role in communicating that experimentation is a prudent approach to managing uncertainty and safeguarding public resources, rather than a sign of mismanagement. There may need to be explicit recognition that some initiatives will not deliver expected results.
Narrative protection should also extend to evaluation itself. Evaluations that identify limited or negative impacts should be recognised as contributing to better decision‑making, not as evidence of poor performance. Evaluators may also prioritise evaluations that are less likely to be misconstrued. For example, rather than conducting an RCT, which lends itself to narratives that the control group was ‘missing out’, evaluators could conduct an A/B test in which one group receives business as usual and another participates in an innovative solution. Both groups receive something, and neither is ‘missing out’. Embedding this perspective might give civil servants the confidence to conduct more rigorous evaluations.
Moving forward
Copy link to Moving forwardGovernments, third sector, and private sector that evaluate innovation share three practices Estonia can adapt: institutionalising evaluation, investing in shared data infrastructure, and assessing performance at portfolio level. As the OECD Recommendation of the Council on Public Policy Evaluation notes, institutionalising evaluation makes it a standing function rather than an occasional exercise (OECD, 2022[14]). For example, the United Kingdom’s Evaluation Task Force builds evaluation into spending decisions, and the Education Endowment Foundation commissions an independent evaluator for every project it funds. Investing in shared data infrastructure ranges from central digital government agencies that publish performance data for each product to the common metric taxonomies that allow venture and impact investors to compare funds on a consistent basis. Assessing performance at portfolio level means judging a set of investments together, on the understanding that a few successes offset the many that do not succeed, as research funders and venture-capital investors do routinely. Table 6.2. presents these practices across the government, third and private sectors, and identifies those most transferable to Estonia.
Table 6.2. Best practices on portfolio- and system-level evaluation across sectors
Copy link to Table 6.2. Best practices on portfolio- and system-level evaluation across sectorsThe table below outlines best practices on portfolio- and system-level evaluation across governments, third, and private sector
|
Government |
Private sector |
Third sector |
|
|---|---|---|---|
|
Governance |
UK Evaluation Task Force: a joint Cabinet Office and HM Treasury unit that builds evaluation into spending decisions and maintains the Magenta Book evaluation standards. ARIA, United Kingdom: evaluates its high-risk portfolio on a decadal horizon, with a formal independent review set for its tenth year, deliberately stepping back from short-cycle metrics. SBIR, United States: subject to independent, congressionally-mandated cyclical assessment by the National Academies. |
Venture capital and private equity: the limited-partner and general-partner structure makes independent evaluation and reporting a contractual obligation, supported by standardised investor reporting (for example ILPA standards) and third-party fund administrators and auditors. |
Nesta, United Kingdom: an innovation foundation that, since 2012, has tied its funding decisions to a published Standards of Evidence framework. Education Endowment Foundation, United Kingdom: an endowed What Works centre that commissions an arm's-length independent evaluator for every funded project. |
|
Data infrastructure |
UK GDS: central digital agencies that publish per-product performance data, giving evaluation a continuous live data feed. Innovate UK: evaluation built on linked national statistical-office business microdata, so supported firms can be compared with a counterfactual. |
Standard portfolio-metric taxonomy: common measures (MOIC, TVPI, DPI, RVPI, IRR, PME) and commercial benchmarking databases (for example Cambridge Associates, Preqin) that make funds comparable vintage by vintage. GIIN IRIS+: a shared catalogue of impact metrics, so impact-investing funds report outcomes on a common basis. |
Education Endowment Foundation: the Teaching and Learning Toolkit, a meta-analytic database that turns thousands of trials into comparable impact, cost and evidence-strength ratings. Social Value International: a shared SROI methodology and metric set, so social-value evidence is recorded consistently across organisations. |
|
Evaluation methods |
i.AI, United Kingdom: the government's AI incubator evaluates every tool it builds against six standard metrics, tailoring the method to each development stage and running randomised or quasi-experimental trials at the Beta stage. HM Treasury and the Evaluation Task Force: supplementary Magenta Book guidance on the impact evaluation of AI interventions, setting out theory-of-change, experimental and quasi-experimental methods for digital and AI projects. SBIR, United States: regression-discontinuity evaluation around the Phase II award threshold. Horizon Europe: structured evaluation against five Better Regulation criteria, with macroeconomic modelling of returns. Contribution analysis: the standard method, after Mayne, where outcomes cannot be attributed cleanly to a single programme. |
Public Market Equivalent benchmarking: fund performance measured against a counterfactual public-market investment, after Kaplan and Schoar. Selection-bias and survivorship correction: returns adjusted for the bias from observing mainly completed deals, after Cochrane. Real-options valuation: pivot, expansion and abandonment decisions priced explicitly, after Trigeorgis. Innovation accounting: validated-learning metrics, after Ries, that track adoption and engagement through pilots and minimum viable products. |
Education Endowment Foundation: commissions independent randomised controlled trials and quasi-experimental designs at scale, feeding the findings back into the Toolkit. Social Return on Investment: stakeholder-defined outcomes monetised through financial proxies and expressed as a ratio of value to cost. |
|
Portfolio and funding logic |
OECD strategic portfolio approach: managing a government's innovation initiatives as one coordinated portfolio, so that risk, learning and synergies are assessed across the set rather than project by project. UK National Audit Office: a 2022 good-practice framework for reviewing the delivery and value of portfolios, giving departments a shared portfolio-review standard. SBIR, United States: Phase I to II to III financing, where each go/no-go gate is an explicit assessment point. ARIA and DARPA: high-risk portfolios assessed at portfolio rather than project level, on the expectation that a few breakthroughs offset many write-offs. |
Staged financing with reassessment: capital released round by round, each round an explicit reassessment point. Portfolio-level return assessment: performance judged across the whole portfolio, which is sized for power-law returns where a few outsized winners offset many write-offs. |
Nesta: the Centre for Social Action Innovation Fund assessed a 52-innovation portfolio as a whole, with structured evaluation support moving grantees up the evidence ladder. Strategic-learning grant-making: foundations increasingly run portfolios against a theory of change, with performance metrics and evidence collected for continuous adjustment (for example Hewlett outcome-focused grant-making). |
Source: OECD based on (HM Treasury, 2020[15]; HM Treasury, 2022[16]; National Audit Office, n.d.[17]; U.S. Government Accountability Office, 2012[18]; European Commission, 2021[19]; OECD, 2022[20]; OECD, Forthcoming[21]; Avnimelech and Teubal, 2006[22]; Howell, 2017[23]; National Academies of Sciences, Engineering & Medicine, 2020[24])
In the immediate term, Estonia should make four short-term changes to improve evaluation.
1. Stage-tagging projects with a maturity stage at the point of award and at each reporting period. This is the single change that makes stage-aware assessment possible;
2. Standard post-mortem template for terminated projects to convert closures from lost information into learning;
3. Shared indicator panel that could link Sprint and Nudgeathon participation to the ministry-level measures already published on the Tõetamm dashboard; and
4. Accelerate Estonia, together with its evaluators, should disclose the programme’s own cost, the missing denominator without which a € 136.6 million benefit pool cannot be turned into a return.
None of these steps requires legislation or significant budget, and each could be in place within a year.
Over the medium term, Estonia should build the infrastructure, baselines, a below-threshold procurement register and a common data structure, that lets these records add up to a portfolio and eventually a system-level assessment. Baseline measures taken before a project starts would make before-and-after comparison possible, where today only after-the-fact data exists. A parallel register for below-threshold innovation procurement would close the largest blind spot in the procurement record. A common data structure across the four programmes, with consistent cost and outcome fields, would let an assessment move from one programme to the whole portfolio. Once those foundations hold, a system-level assessment becomes feasible, and a government able to evaluate each of its portfolios credibly can begin to ask whether its whole government innovation effort is worth what it costs. This is the harder, long-term goal which depends on the immediate steps being taken first.
These reforms will address immediate and medium-term gaps, though longer-term challenges will remain. A defensible portfolio assessment still needs agreed methods for several hard problems: attributing shared outcomes to a single programme, shadow-pricing public-value benefits that carry no market price and correcting for the survivorship bias that comes from observing mainly the projects that succeed. The forthcoming OECD work on value-for-money of government innovation investments will address these gaps and provide further guidance.
Build a T-shaped evaluation capacity
As the OECD’s Improving Governance with Policy Evaluation: Lessons From Country Experiences noted, many public organisations face a gap in evaluation capacity, particularly for impact‑focused methods (OECD, 2020[25]). This gap reflects a lack of specialised evaluation skills available to innovation projects, as well as a broader awareness of these methods and their requirements. Many organisation’s needs may be T-shaped: they need a deep specialisation among a small group of evaluators and broad awareness among the rest of their staff.
On the one hand, organisations need specialised skills to conduct evaluations. These include evaluation design, data collection, randomisation, coding, statistical modelling and translating findings into recommendations. Advanced evaluations such as difference-in-differences and synthetic-control methods require econometric training and the ability to scrutinise whether the model’s assumptions hold. It is not feasible for a broad range of civil servants to be specialists in highly technical evaluation methods. These skills are best maintained in small, dedicated teams for which evaluation is the primary focus. These teams are the vertical line in the T, representing a small cohort with deep expertise. They might exist in public sector organisations, academia or the broader innovation ecosystem. What matters is that they have the technical training to conduct evaluations to a high standard and apply these methods regularly enough that their skills do not atrophy. The Government Office should advocate for the creation of a central government evaluation unit. It should monitor the unit’s performance through activity reporting.
On the other hand, the wide range of non-specialists working on innovation need a sound understanding of evaluation. Even if most civil servants do not conduct advanced or technical evaluations themselves, they need to know enough to work with those who do. These individuals are the horizontal line of the T, representing the vast majority with a broad understanding. They might not know how to code, clean data or specify a statistical model. However, it helps if they know which questions each type of evaluation can answer, the preconditions for conducting each evaluation (e.g. that RCTs rely on randomisation) and how their data are collected so that they can vouch for their accuracy. The Government Office should establish broad-based training on the fundamentals of innovation evaluation for any civil servant involved in innovation.
Involve citizens in evaluating innovation
Estonia already consults citizens on draft laws and their impact assessments through the EIS system and is moving toward earlier co-creation (OECD, 2025[26]). However, this engagement is concentrated at the ex-ante, design end of the policy cycle, while ex post evaluation remains comparatively underdeveloped. The OECD Recommendation on Open Government envisages stakeholder participation across all phases of the policy cycle, including monitoring and evaluation (OECD, 2017[27]), Estonia could usefully explore how to carry its participatory strengths through to the evaluation stage, and extend them from laws to public-sector innovation. This might involve trialling lighter-touch approaches first, such as civic monitoring or user feedback on deployed innovations, before considering whether a citizen panel could contribute to the ex-post review of selected flagship initiatives (OECD, 2022[28]). This would establish whether the assessment-and-comment approach that Estonia already uses for regulation could help evaluate innovations as well.
Recommendations
Copy link to RecommendationsMatch evaluation to the innovation lifecycle and decision points
Structure evaluation around different needs for evidence. Government Office and Ministry of Finance should disseminate guidelines of what types of evaluations are appropriate to answer questions of ROI, impact, scaling and ongoing investment in innovations. The public sector has different evidentiary needs at different points in the innovation lifecycle and for different decision‑makers. Early‑stage innovations often require exploratory evidence to assess whether a problem is well understood, whether an intervention is feasible, and whether there is a credible pathway to impact. At later stages, decision‑makers require stronger evidence on ROI effectiveness, efficiency, and distributional effects in order to justify scaling, mainstreaming, or termination. Treating evaluation as a single, uniform exercise risks producing evidence that is misaligned with decision needs and therefore under‑used.
Properly apply theory and process-based evaluations. Currently, when innovations are evaluated, they’re overwhelmingly theory-based or implementation and process evaluations. As part of their expectation setting, Government Office and Ministry of Finance should encourage innovators to limit theory and process-based evaluations to questions of whether an innovative idea could work; and, improving design while the innovation is being built. Government Office and Ministry of Finance (as part of the budgeting process) should push innovators to make greater use of experimental, quasi-experimental, CBA and VfM evaluations to demonstrate impact and build trust in innovations.
Make impact evaluation a requirement for relevant innovation projects (e.g., especially those connected to central government priorities). Stakeholders reported that impact evaluation is often deprioritised in favour of delivery pressures, particularly in project‑based innovation systems. However, OECD assessments consistently find that innovations are more likely to be evaluated for impact when it is a formal requirement tied to funding, governance, or approval processes. As part of their expectation setting, Government Office and Ministry of Finance should make impact evaluation a requirement for innovation projects. This requirement does not imply mandating complex methods for all initiatives but rather establishing a clear expectation that innovations progressing beyond initial testing must demonstrate evidence of impact.
Embed ROI evaluation in AI innovation. Many government AI projects remain stuck in pilot stages. This is not unique to Estonia: only 28% of OECD countries report conducting any financial or non-financial impact measurement of AI use cases in government, and fewer still assess impacts at a broader level (OECD, 2026[29]). Where criteria for AI investment decisions exist, they tend to emphasise trustworthy use principles and risk management more than evidence of realised benefits, making it harder to build a robust case for scaling what works. The OECD heard from stakeholders in Estonia that innovation is increasingly using AI to improve and transform the public sector. As part of their expectation setting, Government Office should publish guidance describing why and how innovators should quantify the benefit of their AI investment in terms of return on investment (ROI), as ROI is essential to show whether AI creates enough value to justify further funding, support scale-up, and strengthen confidence in what works.
Build evaluation into innovation design and governance
Build evaluation plans into projects earlier. Government Office should strongly encourage evaluation planning and coordination in the kick-off of an innovation project. Evaluation is most effective when it is designed at the outset of an innovation. Early planning allows teams to articulate their intended outcomes, identify appropriate comparators or counterfactuals, collect baseline data, develop narratives explaining the evaluation approach, and ensure that relevant data can be collected (e.g. baseline data, participant information, outcome data, etc.). When evaluation is introduced late, it is significantly harder to conduct an impact evaluation. Oftentimes, baseline data no longer exists, there is no way to construct a comparator group, and evaluation timeframes are markedly shorter. The sooner evaluation is built into an innovation, the more latitude the researchers will have to design it, and the more rigorous its method can be. Integrating evaluation plans into innovation plans does not require fully specifying methods at the start, particularly in adaptive or exploratory initiatives. Early in the innovation lifecycle, evaluators may articulate evaluative questions, identify decision points where evidence will be required, and outline how the evaluation approach will evolve as the innovation matures. This staged approach is well‑suited to innovation contexts characterised by uncertainty and iteration. As the adage goes: malum est consilium, quod mutari non potest - it is a bad plan that cannot be changed.
Invest in the enabling infrastructure for evaluation
Build the data infrastructure for evaluation. Robust impact evaluation depends on the availability of timely, high‑quality data. Many public administrations collect extensive administrative data, but it is often not structured or accessible in ways that support evaluation of innovative interventions. Government Office, with the Ministry of Education and Research, should help build stronger data infrastructure by supporting standardised data collection, reporting requirements and comparisons across initiatives. This should be balanced with a “safe‑to‑fail” approach, recognising that not all innovations will succeed and that learning from failure is a legitimate outcome. Overly punitive reporting regimes may discourage honest evaluation.
Harness evidence-informed systems to drive public sector innovation
Evolve ministerial science advisors into the public sector innovation ecosystem and cross-ministerial agile policymaking teams. Currently, the Science Adviser Network and most innovation cells operate in parallel rather than in tandem. To foster systemic public innovation, science advice must be injected directly into the co-creation and prototyping stages of new public services. Redefining the science advisor's job profile to include innovation competencies ensures that scientific evidence explicitly informs experimental policy designs (such as behavioural nudges or data-driven pilots). The Government Office Innovation Team, alongside the Ministry of Education and Research, as well as the Estonian Research Council, should formalise a workflow where science advisors are automatically onboarded into sectoral innovation sprints and long-term innovation projects. Science advisors could act as the principal conduit to academia, translating scientific publications into the experimental design phase, while helping the innovation teams establish rigorous, evidence-informed key performance indicators (KPIs) for testing.
Establish a standardised cross-ministerial framework and funding pool for ex-post impact evaluation of innovative public services. Estonia’s current impact assessment focus is almost exclusively ex-ante. Without structured ex-post evaluations, there is no institutional mechanism to capture the learnings of innovative state interventions, resulting in a culture that is risk-averse or prone to repeating invisible policy failures. Led by the Ministry of Justice (building on their current legislative quality oversight) and the Ministry of Finance, the government should introduce an agile ex-post evaluation methodology specifically tailored for pilot projects and innovative public service rollouts. A dedicated financial envelope within the strategic cross-ministerial research funding framework should be ring-fenced to allow line ministries to commission independent academic evaluations 12 to 24 months post-implementation.
Embed portfolio-level value-for-money evaluation practices across relevant programmes
Prioritise incremental improvements to innovation evaluation practices in the short term. Government Office, and government ministries writ-large, can strengthen the quality and usability of evaluation data without requiring legislative change or significant additional resources. These include tagging projects by maturity stage at the point of award and at each reporting period, enabling stage-aware assessment across initiatives; introducing a standard post-mortem template for terminated projects to convert closures into structured learning; developing a shared indicator panel linking programme participation (e.g. Sprints, Nudgeathon) to existing ministry-level indicators such as those in the Tõetamm dashboard; and ensuring full cost transparency for major programmes, including disclosing programme-level costs (e.g., Accelerate Estonia), as a prerequisite for assessing value for money.
Build the infrastructure required for portfolio-level assessment over the medium term. Moving from project-level evaluation to portfolio-level understanding requires strengthening the underlying data and system architecture. Government Office could establish baseline measurements at the point of project initiation to enable before-and-after comparison; develop registers for currently unobserved areas of government innovation spending, including below-threshold procurement; and create a common data structure across programmes with consistent cost and outcome fields. These steps enable aggregation and comparability across initiatives, allowing Estonia to move from limited project-level evidence to a coherent portfolio-level view of performance.
Progress toward system-level evaluation capability over time. Once portfolio-level foundations are in place, Government Office could assess innovation investment at the system level. This would require developing approaches to address key methodological challenges, including attributing shared outcomes across programmes, estimating the value of non-market public outcomes, and correcting for survivorship and selection effects in innovation portfolios. Embedding portfolio- and system-level evaluation into decision-making processes could enable the Government Office to assess the extent to which overall government innovation effort delivers public value relative to its cost. Further methodological guidance, including on addressing these challenges, will be developed through ongoing OECD work on the value-for-money of government innovation investments.
References
[5] Accelerate Estonia (2025), Accelerate Estonia cost-benefit analysis (summary), Unpublished summary presentation; analysis with University of Tartu (RAKE).
[35] Arundel, A. and D. Huber (2013), “From too little to too much innovation? Issues in measuring innovation in the public sector”, Structural Change and Economic Dynamics, Vol. 27, pp. 146-159, https://doi.org/10.1016/j.strueco.2013.06.009.
[1] Australian Centre for Evaluation (2023), Commonwealth Evaluation Toolkit, https://evaluation.treasury.gov.au/toolkit/commonwealth-evaluation-toolkit (accessed on 9 July 2026).
[22] Avnimelech, G. and M. Teubal (2006), “Creating venture capital industries that co-evolve with high tech”, Research Policy, Vol. 35/10, pp. 1477-1498, https://doi.org/10.1016/j.respol.2006.09.017.
[41] Berger, M., C. Criscuolo and A. Dechezleprêtre (2025), “Understanding Government Venture Capital: A primer and a taxonomy”, OECD Science, Technology and Industry Working Papers, No. 2025/06, OECD Publishing, Paris, https://doi.org/10.1787/2157a5b3-en.
[33] Bryson, J., B. Crosby and L. Bloomberg (2014), “Public Value Governance: Moving Beyond Traditional Public Administration and the New Public Management”, Public Administration Review, Vol. 74/4, pp. 445-456, https://doi.org/10.1111/puar.12238.
[39] Cochrane, J. (2005), “The risk and return of venture capital”, Journal of Financial Economics, Vol. 75/1, pp. 3-52, https://doi.org/10.1016/j.jfineco.2004.03.006.
[34] De Vries, H., V. Bekkers and L. Tummers (2015), “Innovation in the Public Sector: A Systematic Review and Future Research Agenda”, Public Administration, Vol. 94/1, pp. 146-166, https://doi.org/10.1111/padm.12209.
[19] European Commission (2021), Better Regulation Toolbox, Publications Office of the European Union.
[36] Gompers, P. and J. Lerner (2001), “The Venture Capital Revolution”, Journal of Economic Perspectives, Vol. 15/2, pp. 145-168, https://doi.org/10.1257/jep.15.2.145.
[9] Government Office of Estonia (2025), Innovation Development Projects participation records: Sprint projects, cycles 1-8 (2020-2025) [Unpublished dataset].
[10] Government Office of Estonia (n.d.), Experimentation practicum / design sprint (Katsetamise praktikum), https://riigikantselei.ee/katsedisainisprint (accessed on 29 April 2026).
[7] Government Office of Estonia (2021-2025), Final presentations of projects financed by the Innovation Fund, https://riigikantselei.ee/valitsuse-too-planeerimine-ja-korraldamine/avaliku-sektori-innovatsioon/innovatsiooniprojektid/innovatsioonifondist-rahastatud-projektid (accessed on 29 April 2026).
[6] Government Office of Estonia (n.d.), Innovation projects (Innovatsiooniprojektid), https://riigikantselei.ee/valitsuse-too-planeerimine-ja-korraldamine/avaliku-sektori-innovatsioon/innovatsiooniprojektid (accessed on 29 April 2026).
[11] Government Office of Estonia (n.d.), Nudgeathon (Muksaton), https://riigikantselei.ee/muksaton-0 (accessed on 29 April 2026).
[16] HM Treasury (2022), The Green Book: Central Government guidance on appraisal and evaluation, https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1063330/Green_Book_2022.pdf (accessed on 3 June 2026).
[15] HM Treasury (2020), Magenta Book: Central government guidance on evaluation, https://www.gov.uk/government/publications/the-magenta-book.
[23] Howell, S. (2017), “Financing Innovation: Evidence from R&D Grants”, American Economic Review, Vol. 107/4, pp. 1136-1164, https://doi.org/10.1257/aer.20150808.
[30] Independent Commission for Aid Impact (2018), DFID’s approach to value for money in programme and portfolio management, https://icai.independent.gov.uk/review/value-for-money/ (accessed on 3 June 2026).
[32] Jørgensen, T. and B. Bozeman (2007), “Public Values”, Administration & Society, Vol. 39/3, pp. 354-381, https://doi.org/10.1177/0095399707300703.
[38] Kaplan, S. and A. Schoar (2005), “Private Equity Performance: Returns, Persistence, and Capital Flows”, The Journal of Finance, Vol. 60/4, pp. 1791-1823, https://doi.org/10.1111/j.1540-6261.2005.00780.x.
[2] Kaur, M. et al. (2022), “Innovative capacity of governments: A systemic framework”, OECD Working Papers on Public Governance 51, https://doi.org/10.1787/52389006-en.
[13] KOCK, E. (ed.) (2025), Building capacity for evidence-informed policymaking in governance and public administration in a post-pandemic Europe., Publications Office of the European Union, https://doi.org/10.2760/4815761.
[42] Lerner, J. (2009), Boulevard of Broken Dreams, Princeton University Press, https://doi.org/10.1515/9781400831630.
[43] Mayne, J. (2012), “Contribution analysis: Coming of age?”, Evaluation, Vol. 18/3, pp. 270-280, https://doi.org/10.1177/1356389012451663.
[40] Mazzucato, M. and J. Ryan-Collins (2022), “Putting value creation back into “public value”: from market-fixing to market-shaping”, Journal of Economic Policy Reform, Vol. 25/4, pp. 345-360, https://doi.org/10.1080/17487870.2022.2053537.
[8] Ministry of Finance of Estonia (n.d.), Public Procurement Register (Riigihangete register), https://riigihanked.riik.ee (accessed on 29 April 2026).
[31] Moore, M. (1995), Creating public value: Strategic management in government, Harvard University Press.
[24] National Academies of Sciences, Engineering & Medicine (2020), Review of the Small Business Innovation Research and Small Business Technology Transfer Programs at the Department of Energy, The National Academies Press, https://doi.org/10.17226/25674.
[17] National Audit Office (n.d.), Assessing value for money. Successful commissioning toolkit, https://www.nao.org.uk/successful-commissioning/general-principles/value-for-money/assessing-value-for-money/ (accessed on 3 June 2026).
[29] OECD (2026), “Digital Government Index and Open, Useful and Re-usable Data Index: 2025 Results and Key Findings”, OECD Working Papers on Public Governance, No. 90, OECD Publishing, Paris, https://doi.org/10.1787/6347ec74-en.
[26] OECD (2025), OECD Regulatory Policy Outlook 2025, OECD Publishing, Paris, https://doi.org/10.1787/56b60e39-en.
[28] OECD (2022), OECD Guidelines for Citizen Participation Processes, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/f765caf6-en.
[14] OECD (2022), Recommendation of the Council on Public Policy Evaluation, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0478.
[20] OECD (2022), Tackling Policy Challenges Through Public Sector Innovation: A Strategic Portfolio Approach, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/052b06b7-en.
[25] OECD (2020), Improving Governance with Policy Evaluation: Lessons From Country Experiences, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/89b1577d-en.
[27] OECD (2017), Recommendation of the Council on Open Government, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0438.
[21] OECD (Forthcoming), Strengthening evaluation for public sector innovation, OECD Publishing.
[4] Riigikontroll (2024), Viis aastat tegevuspõhist riigieelarvet: Reformi tulemused ja vajalikud edasised otsused [Five years of activity-based budgeting: Reform results and necessary further decisions] (Annual report to the Riigikogu), National Audit Office of Estonia, https://www.riigikontroll.ee/auditiaruanded/riigikontrolli-aastaaruanne-viis-aastat-tegevuspohist-riigieelarvet-reformi.
[12] Susetyo, A. et al. (2026), Measuring systemic and long-term outcomes of investments in public sector innovation.
[37] Trigeorgis, L. (1996), Real options: Managerial flexibility and strategy in resource allocation, MIT Press.
[18] U.S. Government Accountability Office (2012), Designing Evaluations: 2012 Revision (GAO-12-208G).
[3] Varazzani, C. et al. (2023), “Seven routes to experimentation in policymaking: A guide to applied behavioural science methods”, OECD Working Papers on Public Governance, No. 64, OECD Publishing, Paris, https://doi.org/10.1787/918b6a04-en.
Annex 6.A. Value for money of innovation investment in government
Copy link to Annex 6.A. Value for money of innovation investment in governmentValue for money (VfM) assesses whether the impact of public spending justifies its costs. VfM applies at three points in the policy cycle: ex ante, weighing proposed investments against alternatives; monitoring, tracking delivery during spending; and ex post, evaluating effects once results are observed. This analysis makes costs and results comparable, directing scarce public funds toward uses that return the most public benefit.
The United Kingdom's audit and appraisal tradition anchors VfM in three criteria: economy, efficiency and effectiveness. The National Audit Office (n.d.[17]) applies them ex-post in audit, while HM Treasury’s Green Book applies them ex-ante in appraisal (HM Treasury, 2022[16]). VfM practice increasingly adds a fourth criterion: equity. The Green Book requires appraisers to weigh how costs and benefits are distributed across income groups and places, a requirement strengthened in its 2020 review (HM Treasury, 2022[16]). The former Department for International Development formalised equity to judge programmes on whether they reach those most in need, rather than on cost per head alone (Independent Commission for Aid Impact, 2018[30]).
Public administration scholarship holds that this audit lens is too narrow for what governments produce. Moore (1995[31]) reframed the public manager’s task as creating value for citizens, a standard distinct from private profit and irreducible to countable outputs. Later work catalogued plural public values, from efficiency to transparency, equity and democratic accountability (Jørgensen and Bozeman, 2007[32]). This research positioned public value governance as a paradigm placing democratic values alongside efficiency at the centre of public action (Bryson, Crosby and Bloomberg, 2014[33]). A VfM assessment incorporating public value must therefore address long-term, distributional and institutional outcomes beyond cost alone.
Public sector innovation breaks the assumptions of conventional VfM frameworks, which expect countable outputs, predictable outcomes and short causal chains. This misalignment creates an evaluation gap that project-level methods cannot close. Public sector innovation is heterogeneous in type and purpose. It is shaped by organisational and system-level conditions rather than a single production function (De Vries, Bekkers and Tummers, 2015[34]). Measuring it is difficult because innovation resists output indicators built for routine administration; poorly chosen measures can mislead managers (Arundel and Huber, 2013[35]).
Consequently, evaluation remains rare. OECD work finds that project-level public sector innovation evaluation is uncommon even in administrations with strong evaluation traditions, with few resources supporting VfM evaluation specifically (OECD, Forthcoming[21]). Standard project-level evaluation assesses each initiative against its own objectives. This approach cannot close the gap alone: it penalises early-stage projects for outcomes that cannot yet exist and cannot aggregate discrete initiatives into a statement about whether a portfolio returned value.
Innovation investment behaves like a portfolio of staged, uncertain bets. The venture capital literature, adapted to the public sector, supplies principles and tools to refine these assessments. Returns are highly skewed: a few initiatives generate most value, many produce modest gains and some fail by design. Value accrues over years through diffusion, adoption and capability building.
The venture capital literature supplies both a logic and correction tools: staged financing, where each round buys an option to continue, expand or abandon (Gompers and Lerner, 2001[36]); real-options valuation of that flexibility (Trigeorgis, 1996[37]); benchmarking fund performance against a counterfactual (Kaplan and Schoar, 2005[38]); and explicit correction for selection bias (Cochrane, 2005[39]). These tools must be transferred with care, as public value is not a financial return and public innovations lack an exit market.
Their public sector adaptation runs through mission-oriented innovation policy, which casts the state as an active shaper of markets rather than a corrector of discrete market failures (Mazzucato and Ryan-Collins, 2022[40]). It also links to government venture capital literature, which finds that public equity and grant programmes can crowd in private investment when well designed (Berger, Criscuolo and Dechezleprêtre, 2025[41]; Lerner, 2009[42]). Causal evidence reinforces this: staged public R&D grants raise recipient firms’ subsequent private investment and innovation output (Howell, 2017[23]).
Annex Box 6.A.1. Why portfolio-level assessment is the immediate priority
Copy link to Annex Box 6.A.1. Why portfolio-level assessment is the immediate prioritySystem-level value for money aggregates across every innovation programme a government runs to evaluate the entire system of investment. This is the long-term goal. Portfolio-level assessment evaluates the projects within a single programme together. This is the immediate priority, and (OECD, 2022[20]) has set out how to operate a coordinated portfolio approach in practice. Key reasons include:
The portfolio level is the most tractable place to start: a programme usually has one owner, one budget line and a comparable set of projects. This allows governments to organise the assessment while strengthening project data.
Portfolio assessment produces decisions that matter immediately: funding and prioritisation choices occur programme by programme.
It builds the foundation for system-level assessment: a government that cannot evaluate one portfolio credibly cannot evaluate the broader system.
Source: (OECD, 2022[20])
A VfM assessment of innovation investment must operate at the portfolio level and ultimately the system level. This shift demands that governments meet three conditions spanning governance, data and methods. These three levels nest: project-level evaluation judges a single initiative against its own objectives; portfolio-level assessment judges projects within one programme together; and system-level assessment judges every programme as one effort.
Where attribution to a single programme is infeasible, contribution analysis offers an alternative by assembling a credible causal account and testing it against rival explanations (Mayne, 2012[43]). The OECD argues for managing public sector innovation as a coordinated portfolio, ensuring risk, learning and synergies are handled across a set of initiatives rather than project by project (OECD, 2022[20]). A government can achieve this only through three conditions:
Governance arrangements that embed evaluation into routine funding and prioritisation decisions.
Data infrastructure that records costs, outputs, and outcomes consistently across programmes.
Evaluation practices capable of assessing staged, uncertain and long-horizon returns.