When Can a Portfolio Be Trusted?

The question 0:01

If a student hands an employer a portfolio of real work instead of a list of grades, how much of it can the employer actually trust?

This is WAC Signal, from the World Assessment Council. Evidence and ideas for the changing world of learning.

Episode nine. When can a portfolio be trusted?

The signal 0:30

Grades are losing their power to sort. At Harvard College, A grades made up twenty-four percent of all grades in 2005, and sixty point two percent in 2025. In October 2025 the college's dean of undergraduate education, Amanda Claybaugh, concluded, quote. Our grading no longer performs its primary functions and is undermining our academic mission.

In England, the share of graduates awarded first-class degrees rose from fifteen point eight percent in 2010-11 to a peak of thirty-seven point seven percent in 2020-21. It has fallen for three years since, but the regulator's model still could not explain nearly forty percent of the firsts awarded in 2023-24.

Employers are already moving. In the Job Outlook surveys of the National Association of Colleges and Employers, the share of employers screening graduates on grade point average fell from close to three-quarters in 2019 to thirty-seven percent in 2023, before rising to forty-six percent in the 2025 survey.

So universities are building something to sit beside the transcript. In April 2026 Brandeis University announced a Second Transcript of verified competencies, issued by the registrar. In Australia, a coalition including Commonwealth Bank, KPMG and three universities began design work in July 2026 on a learner-owned Human Capability Record.

The question is whether a record of real work can carry the trust that grades are losing.

The evidence 2:25

Start with what employers say. In an October 2020 survey for the Association of American Colleges and Universities, forty-eight percent of executives and hiring managers rated electronic portfolios very useful in evaluating candidates, against thirty-eight percent for transcripts. In 2015 the gap had been thirty-six percent against nine.

But those are stated preferences, from senior respondents, in surveys run by an association that champions applied learning. What employers do is a different story. In a 2008 study, Ward and Moser surveyed five thousand three hundred and ten employers and got a thirteen percent response. Among employers that did use ePortfolios, only sixteen percent used them at initial screening, and fewer than one percent after the first interview.

Richer university records have hit the same wall. In a 2013 survey of members of the Association of Graduate Recruiters, more than a quarter had not heard of the UK's Higher Education Achievement Report, and of those who had, just under a fifth said they would use it in recruitment. And when the Aspen Institute interviewed twelve employers in 2024, none had fully implemented systems that recognize digital credentials at scale. They wanted competency data and some verification, but not too much information.

Reading time is the real constraint. A 2023 survey of UK admissions staff found that the average personal statement gets about two minutes, and thirty-nine percent get one minute or less. A portfolio that takes twenty minutes to judge will rarely get them.

Then the harder question. Can portfolios be scored consistently? The classic test is Vermont's statewide portfolio program in the early 1990s, evaluated by RAND. In 1992, agreement between raters on writing portfolios ranged from zero point two eight to zero point five seven across scoring criteria. Daniel Koretz, the study's author, put it bluntly, quote. If you're not rating reliably, you're not rating.

Medicine reaches the same conclusion with more precise tools. A UK study of pediatric trainees' portfolios found that reaching a reliability above zero point eight would have needed four independent raters. And a nationwide Taiwanese study published this month, covering forty-five thousand five hundred and twenty-six workplace observations of four hundred and sixty-six specialty trainees, found that a single observation could not support a summative judgment. About thirty observations, spread across six raters, six activities and four settings, reached a dependability of zero point eight.

The lesson is that a portfolio becomes trustworthy through sampling: many pieces, several trained raters, varied settings. That costs time. In Vermont's second year, most mathematics teachers disagreed that scoring portfolio work was not too time-consuming.

The tension 6:03

AI makes a finished artifact much weaker proof of authorship. In a 2025 study, medical educators tried to tell student-written reflections from AI-written ones. Their hit rate for the AI pieces ranged from thirty-six to sixty-four percent, and the authors concluded they could not reliably tell them apart.

Public work signals can be faked too. A study of GitHub found roughly six million suspected fake stars, with a surge in 2024, most of them promoting short-lived malware repositories. Once a signal matters, someone inflates it.

And access to portfolio-worthy work is unequal. The National Survey of College Internships, a pandemic-era pilot at seventeen self-selected institutions, found that fifteen point eight percent of first-generation students had taken an internship in the previous year, against twenty-three point four percent of their peers. Of the students who had not interned, sixty-seven point three percent had wanted to, and among those blocked, forty point one percent said they needed to work a paid job.

Finally, the evidence on impact is missing. Our white paper found no randomized or field experiment showing that an enriched learner record or a verified portfolio changes hiring or admission outcomes compared with a transcript alone.

The implication 7:43

A portfolio is not automatically better evidence than a transcript. A folder of polished work, uploaded by the student, with no record of the conditions, no rater agreement and no check on authorship, is a showcase. It can start a conversation. It should not be used for any screening decision.

But the alternative is not to give up on records of real work. Grades are compressing, and unsupervised written work is a weaker signal of who did the thinking. The question is what has to be true before a reader can rely on a portfolio.

The WAC view 8:26

In our white paper, Beyond the Transcript, we propose the TRACE Ladder. It names five seals that a record of demonstrated work has to earn.

Traceable. The institution, not the student, records where the work came from: the course, the task, the date, the conditions and the AI rules. Rated. It was judged against a public rubric by trained raters. Anchored. There is evidence that another assessor would agree, from double-scoring, reported agreement and enough pieces and raters. Checked. A live defense, supervised demonstration, process history or supervisor attestation confirms the person did it. And Exportable. It comes as a signed, standard record that a reader can verify quickly, with a short summary on top.

The seals stack into rungs. A showcase with no seals is fine for an interview conversation or self-presentation, and nothing more. A record that is traceable and rated can help screen candidates in, but not out. Only a record with all five seals is strong enough for admission credit, progression or credit transfer.

Three design rules follow. Sample, do not curate: dependable judgments need many observations, not a handful of best pieces. Verify authorship at a few points, not everywhere. And summarize for a two-minute reader, with verified claims first.

Elon University shows what traceable looks like in practice. It has collected co-curricular data since 1994, and in 2013 it gave that record a registrar-issued format beside the academic transcript. The institution enters the data, not the student, and the average student records eight experiences.

For universities, we recommend classifying every record they issue on the ladder, setting minimum evidence rules before any portfolio claim counts for credit or admission, and building portfolio-worthy work into required courses, so access does not depend on free time.

What to watch 10:47

One. Whether anyone runs the field experiment that is missing, testing whether verified records change who gets hired, especially for first-generation students.

Two. How employers and admissions offices actually use the new records from Brandeis, Australia's Human Capability Record and others. Watch for data on who opens them, not only surveys.

Three. Whether universities can agree what a verified competency means across institutions, so that proficient at one university means the same at another.

The sources for every figure in this episode are listed on the episode page, along with the full white paper, Beyond the Transcript, at worldassessmentcouncil.org.

WAC Signal is produced by the World Assessment Council, with AI-assisted research and AI voices, and every claim is checked against its source before publication.

This has been WAC Signal. Evidence and ideas for the changing world of learning.