The worker says, “I can do it.” The site asks, “How do I know?” Too often, the answer comes after the shift has started.
The same problem shows up on construction sites, in factories and workshops, inside facilities teams and across contractor networks. UpSkillMe starts with Johannesburg plumbers and welders. The problem is national.
Sites and employersAgenciesPlumbersWeldersEvidence cut: 19 Aug 2026
MarketSouth AfricaNational problem; regional labour markets and work settings differ.
First launch areaJohannesburgPlumbers and welders on participating sites.
Proof objectPerson + task + dateA named check, not a broad occupational badge.
UpSkillMe starts with plumbers and welders in Johannesburg because the first rollout needs enough sites and artisans close together. The trust gap does not stop at the city boundary.
Across South Africa, employers hire through referrals, WhatsApp groups, CVs, job boards, labour intermediaries, contractor panels and walk-ins. Artisans carry qualifications, references, photos and years of experience. Yet the final question is still answered late: can this person do this exact job, to this site’s standard, now?
1,167source-and-issue eventsEvidence records, not people surveyed.
427unique public sourcesOfficial, standards, job, procurement, public voice and other classes.
158job and employer-demand observationsDated demand signals, not a vacancy census.
432de-identified public-voice eventsExploratory evidence, not population prevalence.
Figure 01 · research coverage
What the evidence base actually covers
Counts show where the corpus has evidence. They do not show how common a problem is in the labour market.
Plumbing
559
Welding / fabrication
326
Cross-trade
273
Market-wide
9
Evidence strength
Half of retained events are high-confidence
High585
Medium473
Exploratory109
The corpus unit is a source-and-issue event. A single source can support more than one coded issue, but only once per issue. Job listings are demand signals. Public comments and reviews are used to find hypotheses, not to estimate incidence.1
01
diagnosis
Two sides of the same failure
The national picture is not simply “too few artisans”. Demand and registered work seekers can coexist while a site still cannot match the right person to the right task with proof it can use.
The national paradox
Demand + peoplecan still produce an empty shift.
What the official list says
Plumber and welder both appear on DHET’s 2024 high-demand list.
Both also carry an asterisk for comparatively high numbers of registered unemployed work seekers. The gazette points to geography, training fit, database quality and unverified qualification claims as possible reasons. It also says the list is not a shortage list.2
The practical conclusion is plain: “people available” and “work available” do not guarantee a match. Location, task fit and proof still have to line up.
01right place
02right task
03credible proof
Site employer
“I have applicants. I still do not have an answer.”
A CV can show where someone worked. A certificate can show formal achievement or status. A reference can tell a site who is willing to vouch for the person. None of these, by itself, shows the exact task, material, process, finish or test the employer needs today.
So the site builds confidence again: another phone call, another document check, another practical test, another supervisor watching the first shift. The answer arrives after the programme has started carrying the risk.
More applicants do not solve it. The site needs proof it can use before access, induction and productive work begin.
Artisan
“I have done the work. The proof never follows me.”
Practical experience is often reduced to a job title, a list of employers and whatever paperwork can still be found. Photos sit in old WhatsApp threads. Job cards remain with the contractor. A supervisor’s knowledge leaves with the supervisor.
The artisan pays through repeat applications, transport to tests, unpaid proof days, delayed starts and fewer employers willing to take the first risk. A capable worker can lose to a better-connected or better-documented worker.
Experience loses value when the proof cannot travel. The next employer still has to take the first risk.
Figure 02 · directional issue signal
Where the desk evidence puts the pressure
A research-priority index out of 100, not the percentage of sites or workers affected.
Employer / buyer view
Quality, rework and safety risk96
Scope and quote ambiguity95
Reliability and status visibility91
Payment-chain friction87
Distinguish skill before the job81
Worker / artisan view
Late or unpaid additional work88
Unstable subcontract or project work85
Irregular work and opaque discovery83
Training and market alignment83
Portable proof of competence75
Employer signal shown in orangeWorker signal shown in lime
Scores come from the South Africa trades labour-market workbook. Formula: frequency 30%, severity 25%, economic consequence 20%, evidence breadth 15% and unmet need 10%, scaled to 100. Frequency is normalised within each trade-side ranking. This ranks the desk-research signal; it does not estimate national incidence.1
02
evidence
What each side runs into when hiring or proving skill
There is no single broken hiring channel. Each route carries part of the answer. None consistently carries enough proof to settle the decision.
01 / ROUTES
Different hiring routes. Same question at the end.
The research shows direct referrals, community and WhatsApp groups, public job boards, recruiters and labour brokers, walk-ins, contractor panels and formal procurement. These routes operate in different parts of the market. We do not have a nationally representative split showing which route produces what share of plumber or welder hires.1
01Referral or former supervisor
02WhatsApp or community group
03Walk-in or casual hiring point
04Job board or direct advert
05Recruiter or labour intermediary
06Panel, contractor or procurement route
Channel map, not market share. Every route still needs to answer task fit, recency, site readiness and proof.
02 / PROOF
Each proof type answers a different question
The common mistake is to ask one piece of evidence to do every job. A qualification should not be dismissed because it cannot prove today’s task. A practical test should not be promoted as a qualification because it showed one bounded skill.
Proof used
What it can tell the site
What remains open
Verbal claim
How the person describes their own experience.
Independent observation, scope, recency and comparability.
CV
Employers, roles, tenure and stated responsibilities.
The exact work done, quality reached and who observed it.
Qualification or registration
Formal achievement or status within the issuing body’s rules.
Current fit for this material, process, site and task.
Reference
A past employer or supervisor’s view and context.
Consistency, detail, recency and portability.
Site test or work sample
Direct observation of a specified task.
It is usually repeated at the next employer and often stays in the site file.
Named, dated skill check
Who did which task, when, under what stated scope, with what recorded outcome.
Everything outside that scope. It is not a qualification or blanket trade claim.
03 / TASK
“Plumber” and “welder” are starting points, not complete work briefs
A plumbing decision changes with the system, pipe and fitting type, installation or repair context, access conditions, testing requirement and compliance boundary. A welding decision changes with process, material, joint, position, thickness or range, drawing or WPS/work instruction, equipment and acceptance route.
Public South African vacancy and procurement records reflect this specificity. Employers ask for internal welding tests, technical drawings, job cards, named personnel, materials records, arrival and departure times, pressure or leak testing and documented close-out.45
Figure 03 · practical proof anatomy
What “show me the pipe / show me the weld” means in practice
This is the anatomy of a useful site check, not an official South African rubric. The employer sets the task, safety controls, equipment, supervision, standard and acceptance criteria.
Plumbing
Name the task before judging the person
1
Define. State the system, work type, pipe or fitting, site constraint and required test.
2
Observe. Watch preparation, method, tool use, sequencing and safe handling within the site’s process.
3
Check. Inspect the completed work and run the required functional, pressure or leak test.
4
Record. Attach the result to the named person, date, task scope and observer or approved assessment process.
Welding
Match the person to the weld, not the title
1
Define. State process, material, joint, position, drawing or WPS/work instruction, and acceptance route.
2
Observe. Watch setup, consumables, parameters, preparation, technique and site safety controls.
3
Check. Apply the employer’s visual, dimensional or inspection requirement, including NDT where the work calls for it.
4
Record. Attach the bounded result to the named person and date. Do not turn it into coded-welder or blanket trade status.
04 / TRAVEL
The strongest evidence is often created after the match, then trapped inside the employer
Sites already produce useful records: practical-test outcomes, signed job cards, work descriptions, materials used, arrival and departure times, supervisor sign-off, inspection results and rework history. The information solves one employer’s immediate problem. It rarely moves with the artisan in a consistent, consented form.
Today
Site-owned record
Test result, job card, photo or supervisor knowledge remains in one organisation.
→
UpSkillMe
Worker-consented proof
A named, dated skill check is exposed with its scope and source. Nothing broader is implied.
→
Next decision
Faster screening
The next site sees a bounded proof object before deciding what else it must verify.
03
money
Where the money goes
The wage is the easiest line to see. The expensive lines appear when the wrong match reaches the workface, blocks other people, consumes material or has to be undone.
Legal floor, not artisan benchmark
R30.23
per ordinary hour from 1 March 2026. The national minimum wage excludes allowances such as transport, tools, food and accommodation. It does not represent plumber or welder market pay, agency cost or loaded site cost.6
No made-up national number
We did not find a defensible national figure for the cost of a bad plumber or welder hire.
So this edition shows the cost lines to count and the method for building a real benchmark. A 2012 Cape Peninsula rework study shows why indirect costs are easy to miss, but it is not a 2026 national artisan-hire benchmark.7
Figure 04 · cost cascade
The bill grows as the mismatch moves downstream
The block heights show sequence and compounding, not a researched rand value.
01Sourcing and agency effort
02CV, document and reference checks
03Practical test, induction and access
04Paid time before the mismatch is seen
05Supervisor, helper, plant and crew idle time
06Material loss, rework, retest or callback
07Replacement, resequencing and delay
Cost of a wrong match = sourcing + screening + assessment + access + paid time before detection + supervision + idle dependencies + rework/retest + replacement + delay.Count each component once. Separate skill mismatch from poor scope, missing inputs, unsafe conditions and weak site management.
Caught before work
Screening loss
Recruitment time, document checks, reference calls, practical test, transport or test-day cost, and supervisor time.
Caught during work
Production loss
Everything before, plus paid unproductive time, blocked dependencies, equipment or access time, and replacement mobilisation.
Caught after completion
Rework and liability
Everything before, plus strip-out, material, redo labour, inspection, NDT or pressure testing, callback and programme slippage.
Figure 05 · selected public adverts
Two job titles. Very different work and advertised pay.
Listing-specific examples captured in August 2026. These are not market wage bands, accepted pay or all-in employer cost.10
R0kR10kR20kR30kR40k
Temporary plumbing apprenticeR9k–R10k
General plumbing roleR15k–R19k
Registered, customer-facing plumberR20k–R36k
Stainless TIG / finishing welderabout R8k
Laser welding roleR12k–R15k
Robotic welder programmerup to R40k
Plumbing exampleWelding example
The useful signal is not that one number is “right”. It is that broad job titles conceal different processes, formal requirements, independence, customer exposure and responsibility. A national wage-gap claim would require a matched sample by task, region, work form and worker profile.
How to produce the first credible South African number
Track every participating hire from requisition to day 30
Use Johannesburg for the first tracked rollout, then repeat the method in other provinces and work settings before publishing a national range. Keep plumbers and welders separate. Keep simple, specialist and safety-critical tasks separate.
01RequisitionTrade, named task, location, urgency and formal requirements.
02ChannelReferral, WhatsApp, job board, agency, walk-in, panel or direct hire.
03Proof pathCV, document, reference, site test, named skill check and supervisor hours.
04StartAccess date, first shift and first productive shift.
05OutcomeDay-30 status, reassignment, exit, callback, rework or replacement.
06MoneyPay, invoice, fee, idle hours, material loss, retest and delay.
Still needs primary data
national channel sharesactual agency economicssite-level idle and rework costproof-to-hire conversionmatched wage outcomes
04
voices
What the trust gap sounds like
These are composite, illustrative lines. They are not verbatim interviews and they are not presented as a survey result. They condense recurring mechanisms from public voice, job adverts, procurement records and the research corpus.
“The CV says welder. Fine. Which process, which material, which position, and when did someone last see it?”
“I pay an agency, then I still run my own test. I am paying twice for confidence.”
“A wrong first shift does not cost one person’s day. It pulls in the supervisor, the helper and the schedule.”
“A reference tells me the person worked there. It does not tell me what they actually did.”
“I have done the work for years. The photos are on an old phone and the job cards stay with the contractor.”
“A test day is not free for me. It is transport, food and a day without pay.”
Workers
Put a real skill on record
Start through the South Africa worker line. A named skill check records one bounded piece of practical proof. It does not award Red Seal, NQF, CoC, coded-welder or “certified artisan” status.
See how UpSkillMe shows named, dated skill checks before the decision while the employer keeps control of formal requirements, induction, safety and quality assurance.
This first public edition is built from governed desk research. The evidence is strong enough to describe recurring frictions and state what UpSkillMe will test. It is not strong enough to estimate national prevalence, a national bad-hire cost, a representative hiring-channel split or causal wage effects.
UpSkillMe South Africa Plumbing & Welding Research Corpus, build date 19 August 2026. The corpus contains 1,167 source-and-issue events from 427 unique public sources, including 158 job/employer-demand observations and 432 de-identified public-voice events. Its unit is an evidence event, not a respondent. Public voice is exploratory; job observations are dated demand signals.
Department of Higher Education and Training.2024 National List of Occupations in High Demand. The gazette includes Plumber and Welder, explains the asterisk used for occupations with comparatively high unemployed registrations in ESSA, and discusses geography, training fit, database quality and qualification validation. It expressly states that the list is not a shortage list.
Department of Higher Education and Training / National Artisan Development Support Centre.Artisan Recognition of Prior Learning. Used to show that formal recognition can combine a portfolio of evidence, supporting documents, technical evaluation and practical assessment.
Public vacancy observations retained in the UpSkillMe corpus. Examples include South African welding listings requiring process-specific experience, drawings or symbols and an internal practical test. Dynamic adverts are used as requirement signals, not as a census or proof of worker competence.
Independent Communications Authority of South Africa.Request for Quotation: Plumbing Services. Used as a public buyer-workflow example covering qualifications, references, named personnel, signed job cards, work and material records, times, close-out and testing. It is not treated as a national market estimate.
South African Department of Employment and Labour.2026 National Minimum Wage announcement, 3 February 2026. The R30.23 ordinary-hour floor applies from 1 March 2026 within the department’s stated rules and exclusions. It is not an artisan wage benchmark.
Simpeh, Ndihokubwayo and Love.Evaluating the direct and indirect costs of rework in construction, 2012. The Cape Peninsula study is used only to show that indirect rework costs can be under-recorded. Its percentages are not applied as a current national plumber or welder estimate.
Plumbing Industry Registration Board.Requirements to become a Licensed Plumber. Used to maintain the boundary between platform evidence and licensed-plumber status, registration and CoC-related functions.
Southern African Institute of Welding.Certification. Used to maintain the boundary between an UpSkillMe skill check and formal welding certification, company quality certification, inspection and NDT roles.
Public advertised-pay observations retained in the UpSkillMe labour-market workbook. Selected August 2026 examples are shown only to demonstrate role dispersion. They are listing-specific, can expire and do not establish accepted pay, national medians, total remuneration or all-in employer cost.