By Thuto Motsie CEO of Thamani (CA)
It may seem contradictory, but the cheapest hire you ever made is also the most expensive person on your payroll.
In 2019 our business reached an inflection point that required us to hire an additional manager. We needed an individual who was not only technically strong but had the ability to operate independently and become the go-to for our clients.
We ended up going with the cost-conscious option and hired an individual who was technically sound but lacked the client and relationship management skills. Inevitably, this resulted in me and the other partners having to supplement the gaps in client management. The nature of our business is such that if every staff member is not operating at their optimal level, in terms of their role requirements, the role level above needs to step in.
Yes, in monetary terms, the business “saved” 20% on annual compensation. However, as a director I lost valuable time getting involved in execution matters rather than spending time on strategic objectives like growing the business. It took us 12 months to realise this was not the best hire for the business and it was difficult to quantify the time and opportunity cost of not focusing on the strategic and growth objectives of the business.
When you hire below the market, the savings appear immediately, on a line you can admire every month.
But the cost appears somewhere else: in deals that close more slowly, in work that comes back, in the disproportionate share of your own week that a weak appointment quietly consumes, and eventually in the good people who leave because they got tired of carrying someone.
What the numbers say
There is research that puts a figure on this, and it is older and considerably less fashionable than what usually gets quoted about talent. In 1990, Hunter, Schmidt and Judiesch measured how much individual output varies between people doing the same job. In low-complexity roles, the standard deviation of output was about 19% of the average. In medium-complexity roles, 32%. In high-complexity roles — the judgment-heavy work most of us are actually hiring for — 48%.
Sit with that last number. In a complex role, a strong performer produces roughly twice what a weak one does. The salary gap between them is rarely more than 30% or 40%. Cost scales in a straight line. Output does not.
The uncomfortable part is what the same field says about how we choose people.
For 25 years, hiring practice rested on a 1998 meta-analysis by Schmidt and Hunter that ranked cognitive ability tests and work samples as the strongest predictors of job performance. In 2022, Sackett and colleagues showed those estimates had been systematically inflated by a statistical over-correction, and the ranking changed. Cognitive ability testing has since been revised down twice, and now sits around 12th.
Two numbers from that revision should concern any executive involved in hiring. Years of experience predicts job performance at 0.07 — close to nothing. Unstructured interviews, the conversation most of us trust and nearly all of us conduct, come in at 0.19. Structured interviews — the same conversation, with the same questions put to every candidate and scored against a defined scale — come in at 0.42, the strongest of any method tested.
The distance between 0.19 and 0.42 costs nothing. It is not a system, a budget line or a consultant. It is the decision to write the questions down beforehand and ask everyone the same ones.
The part most of us get wrong
Which brings me to remuneration.
Paying people more does not make them work harder. That has been tested in the field. Gneezy and List found an unexpected pay increase lifted effort for the first few hours, but then the effect vanished.
If your case for paying properly is that it buys discretionary effort, the evidence does not support you.
The finding that does hold is asymmetric, and it is far more useful. Kube, Maréchal and Puppe ran a field experiment on wage cuts. Output fell by more than 20%, and it stayed down. The same study found that equivalent wage increases produced no gain whatsoever. Underpayment destroys output durably. Overpayment does not buy anything extra.
You are not buying effort when you pay properly. You are avoiding a large, persistent and almost entirely invisible penalty — and you are buying two things the market does price honestly: who applies to you and who stays.
George Akerlof described the mechanism more than 40 years ago. Employment, he argued, is not a clean market trade of effort for money. It is partly a gift exchange: the firm gives more than it strictly must, and people reciprocate with more than they strictly must. Cutting to the cheapest wage the market will bear is therefore not costless. It buys you the cheapest labour and the least effort, and it can raise your unit costs while appearing to lower them.
The objection I have to concede
Here is where the argument has to give something up, because the version that concedes nothing deserves to be ignored.
Most of us cannot outbid the market. A business scaling in a capital-constrained economy has a wage bill that is real, monthly and unforgiving, and “pay everyone well” is easy advice to give from a position that is not funding it.
So the answer is not uniform generosity, but selective generosity — and the variance research tells you where to point it. Pay at the top of the band for the roles where output variance is highest: the complex, judgment-heavy positions where the gap between excellent and adequate is a factor of two. Pay at market for the rest, and run those roles well.
What we misread about the South African market
Those of us running businesses here look at an official unemployment rate of 33.6%–43.8% on Statistics South Africa’s broader measure, in the figures released in August, and conclude that we are operating in a buyer’s market for labour.
But one row down from the same survey, unemployment among graduates is 12.4%. Among people without matric, it is 39.2%. Same country, same quarter, a 27-point gap.
That is not a labour surplus. It is a skills mismatch, and the two demand opposite responses.
In a surplus, you can bid down. In a mismatch, you are competing for a thin band of people, while many of your fellow citizens cannot get through the door at all. This is why the state maintains a critical skills list.
There is a second thing we misread, and it is a bigger opportunity. South African companies already spend heavily on developing people, by law. The Skills Development Levy takes 1% of payroll. The B-BBEE scorecard sets a skills development spend target of 6% of the levyable amount. Most of us treat that as a compliance cost: administered by HR, reported annually, optimised for points.
That is a strategy budget being run as a tax return. If capability is scarce, and you are obliged to spend on building it, why is your executive team not deciding where it goes?
The question everyone is asking
I do not believe artificial intelligence reduces the premium on good people. I think it raises it. If the output gap between strong and weak performers widens with the complexity of the work, then a technology that strips out the simple work and leaves the judgment makes talent matter more. The leverage lands with the people who can direct it well, and the distance between them and everyone else widens. If that is right, then businesses that respond to AI by hiring more cheaply are about to make a larger version of the mistake they were already making.
Every business reaches a point where growth depends on who is carrying the product. This is uncomfortable because the things that got you there are not the things the next stage asks of you.
At that point you stop building the thing and start building the people who build the thing. Most of us make that shift later than we should, and most of us make it only after a hire like the one I opened with.
The cost saving is not a saving. It is a transfer — from a line you watch to a dozen you don’t.