Back
Singapore places significant emphasis on workforce outcomes. Returning after 25 years in UK higher education, I noticed the evaluation systems straight away: graduate employment surveys, wage tracking, skills utilisation data, programme outcome metrics. While the UK still argues about whether its graduates are work-ready, Singapore asks for evidence. It is impressive and built with care to make sure that public money spent on learning actually delivers.
Are we measuring the right things, though? This singular inquiry remains unresolved despite rigorous analysis.
Did people find work? Were they paid well? Did employers find them capable? For a long time, asking those questions told us most of what we needed to know. The system rested on a premise that was, for decades, perfectly reasonable: that the knowledge and skills required for good work can be named, taught, and then confirmed by what happens in employment. Industries had recognisable skill requirements. Careers had visible pathways. What was learned and what was needed sat close enough together that measuring one told us something real about the other.
Recently, that closeness has been unravelling. The half-life of useful knowledge is shortening across many fields, not just technology. What someone needs when starting a role often differs from what they will need five years later, and in some areas, even the definition of competence has broadened in ways it didn't a generation ago. AI is the most noticeable cause, but not the only one. The frameworks we rely on to track outcomes were designed for a stable landscape, where they work well. However, the ground is shifting, and it will not remain still.
The timing is hard to ignore. SkillsFuture Singapore and Workforce Singapore are about to become a single agency - Skills and Workforce Development Agency. When the bill went through Parliament in May 2026, members spent real time on exactly this question: how the new body ought to measure outcomes such as wages, job placement and employer behaviour. The evaluation system is currently being rebuilt, making this a good time to ask what it might accurately measure and what it might miss.
Different tools, one assumption
I have seen this before in a different form. UK higher education boasts one of the most advanced quality assurance (QA) systems anywhere, and even it is starting to show strain. When an institution aims to ensure that its graduates are prepared for the future, especially when neither the institution nor employers can clearly define what the future demands, the process of oversight continues, even as the very nature of what is being checked becomes harder to define.
Someone might reasonably argue that quality assurance and evaluation are distinct concepts. They are indeed different: quality assurance examines whether a provision meets an agreed standard, while evaluation assesses whether a programme or investment achieves its promised outcomes. One tests conformance, the other consequence. Different objects, different methods, different questions.
This difference explains why some have lost patience with quality assurance and have shifted to evaluation. QA is a slower process that confirms whether standards are met, but it doesn't reveal if anything truly valuable results. Evaluation appears more agile and asks the more honest question: whether something actually worked. This shift is understandable and often beneficial, but it offers less than it promises. Like QA, evaluation relies on the same hidden assumption: to judge an outcome, we must already agree on what the right outcome should be. When that expectation shifts, evaluation becomes less reliable for the same reasons as QA does, even though it serves a different purpose. The numbers keep appearing, but what they truly represent quietly diminishes.
The doubt is now emerging from the source itself. In late 2025, the OECD released its report on ‘Education for Human Flourishing’, which acknowledges that traditional schooling, focused on measurable cognitive outcomes, is increasingly misaligned with an evolving economy. There is now a growing emphasis on skills such as adaptive problem-solving, ethical judgment, and purposeful action in unfamiliar contexts. i.e. qualities that standardised assessments often fail to capture.
What the numbers miss
There is a second gap that Singapore is well placed to notice. A metric can confirm that a skill was learned. Whether that skill is ever used at work is a separate matter, and one the metric seldom touches. Decades of research on the transfer of training say as much: a great deal of what people learn in formal training never reaches the job, and whether it does has less to do with the training than with what the worker returns to, that is, the backing of a supervisor, the help of colleagues, the chance to actually practise (Hughes, Zajac, Woods and Salas, 2020; Blume, Ford, Baldwin and Huang, 2010).
In Singapore, these bite harder than they might elsewhere. Around 99% of enterprises here are SMEs (Singapore Department of Statistics, 2026), and many operate with lean teams and little spare capacity to release staff for training in the first place; a lack of manpower to cover duties is the most cited obstacle (Singapore Business Federation, 2023). A worker can hold a well-measured credential and still get little from it if the workplace cannot make use of the certification it provides. Skills that go unexercised at work carry a measurable pay penalty, and the latest OECD evidence describes a broken link between the skills people hold and those their jobs actually draw on (OECD, 2026). The outcome data registers none of that. It shows that training happened and that someone was employed, and yet says nothing about whether the skill ever took root.
An economy with no natural resources, built wholly on its people, cannot afford to measure the wrong things. Evaluation still earns its keep. It underwrites accountability, guides where money goes, and sustains public trust, and none of that falls away just because measurement has become harder. What it needs is some candour about its reach. The skills that are most valuable during times of uncertainty include the ability to learn progressively, to think interdisciplinary, and to make sound judgments in situations lacking precedent. These skills develop through education, culture, and the relationships between educators and learners, as well as through workplaces prepared to leverage them. They are tangible and significant. However, an employment survey lacks the means to accurately assess these skills.
A more candid discourse might begin by separating two things: what the metrics can tell us, and what lies outside their range. The cultures of good teaching, the communities of practice, and the workplace conditions that determine whether learning survives contact with the job all need their own attention. Singapore's own thinking about lifelong learning already leans this way, in its recognition that a worker at forty-five needs something different from what they needed at twenty-three, and that an evaluation logic built around the first job may fit the rest of a career poorly.
What has been established here offers a solid foundation, but it's important not to see it as the complete picture. A better starting point is to precisely define where the measures end and consider what lies beyond them.
Are we measuring the right things, though? This singular inquiry remains unresolved despite rigorous analysis.
Did people find work? Were they paid well? Did employers find them capable? For a long time, asking those questions told us most of what we needed to know. The system rested on a premise that was, for decades, perfectly reasonable: that the knowledge and skills required for good work can be named, taught, and then confirmed by what happens in employment. Industries had recognisable skill requirements. Careers had visible pathways. What was learned and what was needed sat close enough together that measuring one told us something real about the other.
Recently, that closeness has been unravelling. The half-life of useful knowledge is shortening across many fields, not just technology. What someone needs when starting a role often differs from what they will need five years later, and in some areas, even the definition of competence has broadened in ways it didn't a generation ago. AI is the most noticeable cause, but not the only one. The frameworks we rely on to track outcomes were designed for a stable landscape, where they work well. However, the ground is shifting, and it will not remain still.
The timing is hard to ignore. SkillsFuture Singapore and Workforce Singapore are about to become a single agency - Skills and Workforce Development Agency. When the bill went through Parliament in May 2026, members spent real time on exactly this question: how the new body ought to measure outcomes such as wages, job placement and employer behaviour. The evaluation system is currently being rebuilt, making this a good time to ask what it might accurately measure and what it might miss.
Different tools, one assumption
I have seen this before in a different form. UK higher education boasts one of the most advanced quality assurance (QA) systems anywhere, and even it is starting to show strain. When an institution aims to ensure that its graduates are prepared for the future, especially when neither the institution nor employers can clearly define what the future demands, the process of oversight continues, even as the very nature of what is being checked becomes harder to define.
Someone might reasonably argue that quality assurance and evaluation are distinct concepts. They are indeed different: quality assurance examines whether a provision meets an agreed standard, while evaluation assesses whether a programme or investment achieves its promised outcomes. One tests conformance, the other consequence. Different objects, different methods, different questions.
This difference explains why some have lost patience with quality assurance and have shifted to evaluation. QA is a slower process that confirms whether standards are met, but it doesn't reveal if anything truly valuable results. Evaluation appears more agile and asks the more honest question: whether something actually worked. This shift is understandable and often beneficial, but it offers less than it promises. Like QA, evaluation relies on the same hidden assumption: to judge an outcome, we must already agree on what the right outcome should be. When that expectation shifts, evaluation becomes less reliable for the same reasons as QA does, even though it serves a different purpose. The numbers keep appearing, but what they truly represent quietly diminishes.
The doubt is now emerging from the source itself. In late 2025, the OECD released its report on ‘Education for Human Flourishing’, which acknowledges that traditional schooling, focused on measurable cognitive outcomes, is increasingly misaligned with an evolving economy. There is now a growing emphasis on skills such as adaptive problem-solving, ethical judgment, and purposeful action in unfamiliar contexts. i.e. qualities that standardised assessments often fail to capture.
What the numbers miss
There is a second gap that Singapore is well placed to notice. A metric can confirm that a skill was learned. Whether that skill is ever used at work is a separate matter, and one the metric seldom touches. Decades of research on the transfer of training say as much: a great deal of what people learn in formal training never reaches the job, and whether it does has less to do with the training than with what the worker returns to, that is, the backing of a supervisor, the help of colleagues, the chance to actually practise (Hughes, Zajac, Woods and Salas, 2020; Blume, Ford, Baldwin and Huang, 2010).
In Singapore, these bite harder than they might elsewhere. Around 99% of enterprises here are SMEs (Singapore Department of Statistics, 2026), and many operate with lean teams and little spare capacity to release staff for training in the first place; a lack of manpower to cover duties is the most cited obstacle (Singapore Business Federation, 2023). A worker can hold a well-measured credential and still get little from it if the workplace cannot make use of the certification it provides. Skills that go unexercised at work carry a measurable pay penalty, and the latest OECD evidence describes a broken link between the skills people hold and those their jobs actually draw on (OECD, 2026). The outcome data registers none of that. It shows that training happened and that someone was employed, and yet says nothing about whether the skill ever took root.
An economy with no natural resources, built wholly on its people, cannot afford to measure the wrong things. Evaluation still earns its keep. It underwrites accountability, guides where money goes, and sustains public trust, and none of that falls away just because measurement has become harder. What it needs is some candour about its reach. The skills that are most valuable during times of uncertainty include the ability to learn progressively, to think interdisciplinary, and to make sound judgments in situations lacking precedent. These skills develop through education, culture, and the relationships between educators and learners, as well as through workplaces prepared to leverage them. They are tangible and significant. However, an employment survey lacks the means to accurately assess these skills.
A more candid discourse might begin by separating two things: what the metrics can tell us, and what lies outside their range. The cultures of good teaching, the communities of practice, and the workplace conditions that determine whether learning survives contact with the job all need their own attention. Singapore's own thinking about lifelong learning already leans this way, in its recognition that a worker at forty-five needs something different from what they needed at twenty-three, and that an evaluation logic built around the first job may fit the rest of a career poorly.
What has been established here offers a solid foundation, but it's important not to see it as the complete picture. A better starting point is to precisely define where the measures end and consider what lies beyond them.
Lim Wai Mun, principal research at IAL’s research division earlier published a version of her opinion in the Higher Education Policy Institute (HEPI) website on 7 June 2026. HEPI is the UK’s only independent think tank for higher education. The views expressed in the article are those of the author and do not necessarily reflect the views of IAL.