The Rand You Can't See: Cost Every Repeat Test Before You Ask a Funder for Money

The Rand You Can’t See: Cost Every Repeat Test Before You Ask a Funder for Money

It’s 4pm on a Thursday and your best technician is re-running a batch that failed QC this morning. The reagents went in twice. Her afternoon is gone. The samples behind hers are now late, and tomorrow starts with a backlog instead of a clean bench. Down the corridor, the analyser everyone complains about needs three attempts to pass calibration before it’ll accept a single sample.

None of that shows up as a line on your income statement.

You see the reagent invoices. You see the salaries. What you don’t see is the slice of both that gets burned doing the same work twice — the repeat test, the re-run batch, the supervisor hours spent fixing yesterday’s error. Labs that bring testing in-house to save money routinely miss this, because the full cost picture only shows up after the commitment is made.

Here’s the thing most lab owners get wrong when they finally decide to ask a bank or a development funder for money: they walk in with a wish list. A new analyser. More staff. Better training. What a funder actually lends against is a number — a quantified loss you can prove, and a believable date by which the investment pays for itself. This is how you build that number from your own books.

The Opportunity

The win here isn’t “better quality” in the abstract. It’s recovering margin you’re already losing, and converting it into an investment case a funder will approve.

Cost-per-test isn’t a vanity metric. It’s an established, optimisable operating KPI — the one lab managers track alongside turnaround time precisely because you can move it without compromising diagnostic quality. The moment you can show your cost-per-test falling, you’re speaking the language a lender understands.

And capability scales with cost. A broader test menu, faster turnaround, tighter QC — each costs more to run, which means every investment decision is really a capability-versus-cost trade-off. That’s good news. A trade-off can be modelled. A “feeling that the old kit is hurting us” cannot.

The payoff is a single, honest sentence you can say across a desk at the IDC, SEFA, the NEF, or your bank: “This investment removes R X of cost a year, it costs R Y, so it pays back in Z months — and here’s my cost-per-test before and after.” That sentence is worth more than any brochure about the equipment you want to buy.

One caution before we build it. Every rand input below has to come from your own records or from a verified source — SARS, SANAS, your equipment supplier. The published research on lab costs is either qualitative or in US dollars. So this guide hands you the formulas and tells you to fill in your own figures. Anyone who quotes you a tidy local benchmark for “the cost of a repeat test in South Africa” is guessing. Don’t build a funding application on a guess.

How It Works

Six steps. Work through them in order.

Step one — build the denominator. Count validated tests produced in a period, not tests started. A test you ran twice is one validated result and two lots of cost. Counting starts, not finishes, is exactly how the hidden loss stays hidden.

Step two — layer the three cost tiers onto each test. Direct costs are the obvious ones: reagents, consumables, technician hours. Indirect costs are the ones you pay whether or not a sample moves: instrument depreciation, maintenance contracts, calibration, accreditation, utilities. Hidden costs are the ones nobody invoices you for: repeat tests from error, downtime, rework, failed QC, lost throughput. Add the three tiers, divide by validated tests, and you have a true cost-per-test — usually higher than the one in your head.

Step three — put a rand figure on the untrained technician. This is where a feeling becomes a number. The cost works out as: (repeat-test rate × cost-per-test × annual volume) + (QC-failure rate × cost to re-run a batch × number of batches) + (supervisor rework hours × that supervisor’s loaded hourly rate) + (tests lost to slower throughput × the margin on a test slot). You measure each rate yourself over a quarter. I’m not handing you a “10% repeat rate” industry fact, because no credible South African one exists — the multiplier has to be your own, measured on your own bench.

Step four — put a rand figure on the ageing instrument the same way. Add rising maintenance and repair spend, plus unplanned downtime hours multiplied by your revenue-per-hour, plus the extra reagent and calibration consumption a tired machine eats, plus the repeat tests it causes. Then compare that annual figure against the amortised cost of a replacement. All lab equipment eventually becomes outdated and needs reinvestment — so an old analyser isn’t a one-off purchase you’ve avoided, it’s a recurring cost you’re paying in instalments you never see.

Step five — build the one-page funder dashboard. Six things, no more: cost-per-test trend over the last few months, your repeat/rework rate, instrument downtime hours, QC-failure rate, technician competency and training status, and your projected cost-per-test after the investment. A lender doesn’t want your whole LIMS export. They want the trend and the projection on one page.

Step six — convert it to the ask. Annual hidden cost eliminated, divided by the investment, gives you payback in months. Show it beside the before-and-after cost-per-test. That’s the whole application in two figures.

On the South African specifics worth chasing down: technician training is usually the lowest-capital lever, and it may attract real support. SARS’s Section 12H learnership allowance and SETA grants — for labs, typically the HWSETA — exist to subsidise structured training, and there are wear-and-tear allowances (look at Sections 11(e) and 12C) that affect the after-tax cost of equipment. I’m not quoting you the amounts, because rates and eligibility change and the brief I’m working from flags every one of them as unverified. Confirm the current figures with a registered tax practitioner or directly with SARS before you put a single rand of allowance into your model. Same goes for SANAS accreditation costs under ISO/IEC 17025 and HPCSA registration rules for your technicians — check, don’t assume.

Case Study: Costing the Repeat Rate at a Gauteng Water Lab

Everything below is an illustrative worked example. The lab is composite, and every rand figure is a placeholder to show you the mechanics — not a real client and not a real result. Use the structure; supply your own numbers.

Who: an independent water-testing lab in Gauteng, twelve staff, profitable but stretched, doing municipal and industrial compliance work.

The problem: the owner suspected two things were quietly eating margin — a repeat rate that climbed whenever a junior ran a batch unsupervised, and a six-year-old analyser that kept dropping out of calibration. He had no number for either.

What changed: he ran the model for one quarter. Assume a true cost-per-test of R220 once all three tiers were added, and 18,000 validated tests a year. He measured an 8% repeat rate, with most repeats traced to two undertrained juniors. Plug those assumed figures in: 8% × R220 × 18,000 comes to roughly R317,000 a year in repeat-test cost alone, before counting supervisor rework hours and the throughput he lost while the senior tech cleaned up. The ageing analyser added its own assumed R140,000 a year in downtime and extra calibration.

Why training came first: a structured competency programme for the two juniors was modelled at an assumed R90,000 — far below the cost of replacing the analyser. It attacked the repeat and QC-failure cost directly. On these placeholder numbers, eliminating even two-thirds of the repeat cost pays the training back in roughly five months, and his projected cost-per-test drops below R205. The equipment replacement became the second, larger case the same dashboard supported — a stronger application once the training case had already proven the method.

The result: on these illustrative figures, a funder sees a five-month payback on a small ask, backed by a measured before-and-after. That’s an easy yes compared with “I’d like R600,000 for a new machine, trust me.”

The friction: the first quarter of data was messy. His technicians, sensing the repeat rate was really a competency score, started quietly re-logging repeats as new tests — which corrupted the denominator. He almost abandoned the exercise. What saved it was being direct with the team: the number measures the lab’s training gap, not their job security. Once that landed, the data held. If you skip that conversation, you’ll measure your team’s fear instead of your repeat rate.

Frequently Asked Questions

I already know my lab is expensive to run. Why model all this?

Because “expensive” isn’t fundable. The model splits unavoidable cost from recoverable hidden cost, and the recoverable part is the only thing a lender lends against. Knowing your lab is expensive changes nothing. Knowing that R317,000 of that expense is repeat work you can fix for R90,000 changes the conversation entirely.

Won’t a lender just want collateral, not a dashboard?

Asset finance still needs a repayment story. Collateral protects the lender if you fail; the dashboard tells them why you won’t. Before-and-after cost-per-test and a payback in months is that story, and it strengthens any application — to a bank, the IDC, SEFA or the NEF. I won’t quote you their exact approval criteria, because those vary and I haven’t verified them; ask each funder directly what metrics they want. The dashboard makes you the applicant who already has answers.

Training is a soft cost. Isn’t a machine easier to justify than sending staff on a course?

It feels that way, which is exactly why most owners over-invest in equipment. Training is usually the lowest-capital lever and it hits repeat and QC-failure cost head-on. It may also attract SARS Section 12H learnership support and a SETA grant — but verify those amounts with a tax practitioner before banking on them. The machine is real and tangible. So is a repeat rate you’ve written down for three months running.