Sources & calculations · Yaschin Mohabir
The evidence behind
the report.
The records were collected in person by Yaschin Mohabir in September 2026. The observation cutoff is 12 September 2026. They sit in one workbook, SA_Artisan_Hiring_200_Responses_v1.xlsx, with one hiring journey per row. The report publishes aggregates only. This is an internal analysis and has not been independently audited.
1. The study group
200 hiring journeys: 100 plumbing and 100 welding. Each trade has 35 direct-employment, 30 agency-placement and 35 commissioned-contractor journeys. A journey follows one role or work requirement and can contain several rounds and starts. The journeys contain 560 rounds. Figures describe these records. They are not national estimates, and route comparisons do not isolate the effect of the hiring route.
Source: Journeys; Responses (one journey per row).
2. Selection effort and time
Round means use round records, and journey means include every round in the journey. Counts can include repeated contact with one person. Calendar time is kept separate from active hours. Missing values are excluded, and each figure shows its valid count.
Source: Rounds; Journeys.
3. Agreements, starts and outcomes
602 agreed engagements: 400 actual starts, 191 no-shows and 11 not yet assessed. No-show rates use the 591 assessed agreements. The 400 starts split into 275 failed, 35 continuing with the employer dissatisfied and 90 completed or continuing satisfactorily. Dissatisfaction among continuing engagements uses 70 records on each side. Dissatisfaction after replacement uses a separate base of 93 and isn't added to these branches. 151 journeys required a replacement, and there were 217 replacement starts.
Source: Engagements; Rate summary; Journeys.
4. The employer cost waterfall
The waterfall uses the 189 journeys with complete employer cost records, the same set for every component. Mean full cost is R96,615.84: R42,545.26 ordinary plus R54,070.57 additional failure/restart spending. By basis, that is R41,158.86 net recorded cash, R23,422.96 valued internal time and R32,034.01 accrued labour. It isn't all cash paid, loss or saving. The separate additional-cost metric has 193 complete records (mean R53,705.57) and isn't added to the waterfall. Each ledger line counts once, recoveries are subtracted once, and the workbook's overlapping repeat-recruitment subtotal isn't added again.
Source: Cost ledger; Employer costs.
5. The artisan's costs and payment
Net cash spending (travel, data and fees, tools and test materials, and worker-paid rework, less reimbursements) has 200 complete journey totals. Unpaid time has 188 and has no cash value attached. 333 of 728 assessed due payment obligations were paid late or are overdue.
Source: Artisan costs; Payments.
6. The issue tree
The fourteen branches, seven employer and seven artisan, come from the categories in SA_TRADES_RESEARCH_CORPUS.xlsx. They are categories, not a ranking of how often each occurs. Quoted response excerpts each come from one recorded journey.
Source: SA_TRADES_RESEARCH_CORPUS.xlsx, Taxonomy.
7. The four rework cases
Case JOB110/J110, stainless assemblies and pipe fittings: original R27,788.48 + removal R8,532 + repair R19,081.78 + replacement R22,939.77 + retesting R1,947 + disruption R11,368 − recovery R1,419 = R90,238.03. It took 88 extra labour hours and 20 days of delay, and was accepted after remedial work. The other cases are JOB101, JOB102 and JOB002, and all four reconcile. Of the 183 assessed jobs, 111 needed rework, and 4 of those 111 kept their original worker.
Source: Case examples; Jobs; Rework.
8. Calculations and checks
P10 and P90 use inclusive linear interpolation on sorted valid values at position (n−1) × p. They show spread, not confidence intervals. Every filter is recalculated from the record values in Analysis values, never by averaging group medians.
An internal arithmetic re-check rebuilt the ledger amounts, cost totals, rework components, rate numerators and denominators, and case totals, and they reconcile. All 588 metric groups in Metric summary (rows 6–593) match on count, mean and median, and P10/P90 were recalculated. The original workbook is unchanged.
9. The scope of a named skill check
An UpSkillMe task record names the task, conditions, date and responsible assessor. It is not a Red Seal, NQF award, plumbing CoC, coded-welder qualification or SAIW certification. It does not replace the employer’s formal requirements, induction, safety checks or quality assurance.