Employer-side screening uses an assessment to sharpen one firm's hiring decision. Candidate-side certification gives the same result to the candidate to carry between employers. In a South African field experiment, a shareable certificate raised employment by 17% and weekly earnings by 34%; the identical scores, issued in a form candidates could not credibly share, moved employment not at all.
That pair of results comes from one of the few studies built to separate the two. Carranza, Garlick, Orkin and Rankin (2022) assessed 6,891 unemployed and underemployed young workseekers in urban South Africa on non-specialist skills — communication, numeracy, grit — then randomised who got to know the results. The finding is not that assessment works. It is that where the result sits decides most of the return.
Why this question is live in offshore services markets
South Africa's global business services sector has become the clearest test case. It has created roughly 200,000 cumulative jobs since 2006, and in October 2026 industry body BPESA revised its own outlook, telling ITWeb that about 150,000 additional jobs by 2030 is realistic against the 2021 master plan's 500,000 — partly because the growth is moving toward complex roles that "are not going to scale."
The hiring pool those roles draw on carries very little documentary signal:
- Labour market, Statistics South Africa — the Q2 2026 Quarterly Labour Force Survey put official unemployment at 33.6% and youth (15–34) unemployment at 47.4%.
- Disconnection from work and study — in Q1 2026, 45.6% of 15–34 year olds were not in employment, education or training, and the 15–24 unemployment rate was 60.9%.
- Who the sector actually hires — BPESA's job creation report for April–June 2025 recorded 8,180 net new international jobs, of which 90% went to workers aged 18–34, and 71% were frontline voice roles.
A first-time workseeker in that pool has no employer reference, no performance record and often no post-secondary qualification. Everything an employer could screen on has to be generated, not retrieved.
What the evidence says for employer-side screening
Putting the assessment inside the employer's funnel has a long, solid record on the outcome employers care about — retention and performance of the people they hire.
- Autor and Scarborough studied a national retailer whose 1,363 stores switched from informal to test-based screening over a year. Median tenure rose by more than 10%, and the switch produced no measurable reduction in minority hiring rates, with productivity gains large for minority and non-minority hires alike.
- Hoffman, Kahn and Li examined 15 firms hiring into low-skill service roles where median tenure was 99 days. Greater reliance on the job test raised average tenure of hires by about 15%, and a one-standard-deviation increase in managers' rate of overriding the test was associated with roughly 5% shorter tenure.
- Carranza and colleagues' third experiment points the same way from the other direction: giving assessment results only to firms shifted their interview decisions.
Employer-side screening is not the weak option. It is the right tool when the question is which of these applicants to hire into this role.
What the evidence says for candidate-side certification
Certification answers a different question: how does a person with no record get into the right funnel at all?
- In the South African experiment, workseekers randomised into public certification received electronic and physical certificates showing their assessment results, their names and national identity numbers, and the branding of the agency that ran the assessments. Over the following three to four months their employment rate rose 17% — five percentage points above the control group — weekly earnings rose 34% and hourly wages 20%. The authors note the employment effect was almost three times the mean effect of the active labour market programmes reviewed by Card, Kluve and Weber, and that the first three months' earnings gain was 5.6 times the variable cost of bolting certification onto an assessment programme that already existed.
- Bassi and Nansamba (2022) ran a complementary experiment with small firms in Uganda, disclosing certificates of young workers' non-cognitive skills to both sides before interviews. Workers raised their expectations, higher-ability managers revised their assessments of worker skill upward, positive assortative matching increased and earnings rose for those employed.
- Abel, Burger and Piraino (2020) tested the nearest available alternative — an employer reference letter. Attaching one raised employer response rates from about 4.2% to 6.7% and interview requests from 2.4% to 3.9%, with the effect concentrated among women, whose response rates roughly doubled. Encouraging workseekers to obtain a letter raised women's employment by 5.9 percentage points on an intent-to-treat basis. The catch is structural: a reference letter requires a previous employer, so it is unavailable to exactly the first-time workseekers who need a signal most.
- Pallais (2014) explains why the market under-supplies this on its own. She hired 952 inexperienced workers on an online labour platform and gave them detailed or coarse public evaluations; public information about an unproven worker improved their later outcomes, but the hiring firm captures none of that value, so firms rationally hire too few inexperienced workers in the first place.
The distinction that carries the whole result
The sharpest finding in the South African study is the comparison between its two certificate arms. The private certificate showed the same scores but deliberately omitted the features that made the result credible to an employer — the issuing agency's branding and the candidate's identifying information. Private and public certification shifted candidates' beliefs about their own skills and their search targeting almost identically. Only public certification moved employment.
The implication is uncomfortable for a lot of assessment practice. A score that tells a candidate something true about themselves but cannot be verified by a third party produces insight without employment. What converted insight into a job was a verifiable issuer, an identity binding and a legible description of what was measured — not better psychometrics. That is a more practical requirement than cryptographic tamper-proofing, and a more demanding one than most self-service score reports meet. A portable record such as an AssessAll Skill Passport is useful only to the extent it carries those three things.
Three limits worth stating plainly
Signal congestion is real. The same study ran a correspondence experiment, submitting real résumés to real vacancies. When one application at a vacancy carried a certificate, it was 11% more likely to get an interview — but that advantage vanished as more certified applications arrived. Certification is partly positional. A scheme that certifies everyone in a labour market should expect its per-candidate interview advantage to shrink, even if match quality improves.
Certification sorts horizontally, not vertically. The authors found the effects more consistent with helping firms work out which candidates suit which jobs than with identifying a subset who suit everything. A certificate is a matching instrument, not a ranking.
The effect is largest where other signals are absent. Gains concentrated among workseekers lacking work experience or university education. In a senior or credential-rich pool, a certificate is adding information to a file that already has some.
When each approach is right
- Use employer-side screening when you control the funnel, the role is defined, volume is high enough to justify calibration, and you need defensible within-role comparisons. Ownership of the item bank and the scoring rules matters here, and the cost model should scale with candidates rather than licensed seats — pay-as-you-go credits of the kind AssessAll uses, at ₹30 or US$0.50 each, change what you can afford to measure at the top of a funnel.
- Use candidate-side certification when the pool is dominated by first-time workseekers, when several employers draw on the same pool, when you are recruiting across a region rather than a city, or when a public programme or sector body — not a single firm — is paying. The returns show up as market-wide match quality, which is precisely why no individual firm funds it.
- Use both when it is the same assessment. Nothing in the evidence requires choosing: the South African certificates came out of an assessment programme that already existed, and adding the candidate-facing artefact cost a fraction of what it returned.
The practical takeaway is that an assessment result has an owner, and the owner determines the return. Decide deliberately whether yours is built to improve one hiring decision or to travel with a person who has nothing else to show.