Stage 4 of eight. This article is one step in the production line buyer’s roadmap — the full sequence from defining the product to the first year of operation.
Who this is for
You are choosing between two versions of the same line. The semi-automatic option is materially cheaper and needs more people. The fully automatic option costs more, needs fewer people, and the supplier's business case is built almost entirely on the wage bill you will avoid.
That business case was written for a different labour market. This article is about how the sum actually works in South Africa and the region, and how to decide which steps to automate rather than how much automation to buy.
The imported business case, and why it does not transfer
Most automation economics you will encounter assumes two conditions: labour is the dominant variable cost, and when something breaks a technician and a part are available within a day. Both hold in Germany. Neither holds here in the same way.
What is true locally instead:
- Direct labour is comparatively inexpensive, so the payback period calculated on headcount is longer than the brochure's.
- Skilled labour is not. Millwrights, instrument technicians and controls people are scarce and command a premium — and industry surveys of South African manufacturers consistently put skills availability among the top constraints, with production-management and engineering skills specifically named. Automation does not remove labour dependence; it converts a dependence on many semi-skilled people into a dependence on a few scarce ones.
- Downtime is expensive and slow to resolve. A failed servo drive is not a day's problem if the replacement ships from Asia. The more automated the line, the more single points of failure sit between you and output.
- The grid interrupts. Every automated restart sequence is a sequence that has to survive a stop it did not plan for.
- Volumes are frequently lower and more variable than the equipment was designed around, and automation is least forgiving at low utilisation — the capital is committed whether the line runs or not.
None of that is an argument for a manual factory. It is an argument for spending the automation budget somewhere other than where the brochure points it.
The three reasons to automate that survive the translation
Strip out headcount and three reasons remain, and all three get stronger rather than weaker in this market.
1. The step a person cannot hold
Fill weight to a gram, a seal at a temperature and dwell, a dose, a torque, a coating thickness. Where the specification is tighter than human consistency, automation is not a labour decision at all — it is the only way to make the product. Giveaway on a filling line is the clearest version of this: over-filling every container by a percent to stay legal is a raw-material cost that runs every hour the line runs, and it usually dwarfs the wage difference being debated.
2. The step that is unsafe
Guarding costs less than an incident and far less than a prosecution, and your obligations under the Occupational Health and Safety Act attach to the line as installed on your floor. Where a step involves a nip point, a blade, a press, a hot surface or a confined reach, automating the handling is usually cheaper than engineering a person safely into it.
3. The step that decides quality — and records it
Inspection, coding, traceability, batch records. A retailer audit, an export customer or a recall investigation asks for evidence, and evidence produced by hand is evidence produced under pressure at the end of a shift. This is also the cheapest automation on the list to add later, which matters for sequencing.
What to keep manual, deliberately
The corollary is worth stating explicitly, because buyers rarely hear it from a supplier:
- Steps with high product variety. A person changes over between three pack sizes in minutes. An automated station needs change parts, setting and validation for each one, and if your SKU mix is wide and your volumes modest the automation spends its life being reconfigured.
- End-of-line palletising at modest volume. Frequently the last thing worth automating and the first thing quoted.
- Judgement steps. Grading, sorting by appearance, dealing with the non-standard. Vision systems do this well at volume and expensively at low volume — the honest comparison is in our machine vision inspection cost guide.
- Anything where the automated version has no local support. A robot nobody within 500 km can program is a manual station with an expensive statue next to it.
The comparison, laid out
| Dimension | Semi-automatic | Fully automatic |
|---|---|---|
| Capital | Lower — often 40–60% of the automatic option for the same nominal rate | Higher, and the gap widens with the number of automated transfers |
| People | More operators, mostly semi-skilled and trainable in days | Fewer operators, but a real dependence on a millwright and a controls technician |
| Consistency | Varies by shift, operator and fatigue | Consistent while it runs — which is the main reason it wins on quality-critical steps |
| Changeover | Fast and flexible; a person adapts | Needs change parts and setting; the wider your SKU mix, the worse this gets |
| Failure behaviour | Degrades — one station slows, the line continues | Stops — more single points of failure, each one a full line stop |
| Load shedding | Restarts easily; people resume | Needs a designed restart sequence, and product in process may be lost |
| Spares exposure | Mechanical, mostly locally available | Drives, servos, sensors, controllers — lead times measured in weeks |
| Best at | Modest volume, wide SKU mix, uncertain demand | Steady volume, narrow SKU mix, tight tolerance, safety-critical handling |
The row that most first-time buyers under-weight is failure behaviour. A semi-automatic line has a graceful failure mode: a station slows, people work around it, some output continues. A fully automatic line is binary. That difference does not appear in a payback calculation and it dominates the first year of operation.
A sequence that beats a decision
The best answer is usually not a point on the scale but an order of operations. Buy the line at the automation level your current volume and SKU mix justify, and buy it prepared for the next level:
- Specify the space. Leave floor length in the layout where an automated station will later go — a case packer, a palletiser, an inspection station. Space is free on a drawing and unbuyable afterwards.
- Specify the interfaces. Conveyor heights, discharge positions and control signals that a future station can be dropped into. This is the item most often lost.
- Specify the controls headroom. A PLC with spare I/O and a network the next station can join, rather than a controller sized exactly to today.
- Automate the quality and record-keeping steps first. They are the cheapest to add, the most defensible to an auditor, and they do not depend on volume growth to pay back.
Done this way, the upgrade at year three is a purchase order rather than a project. Done the other way, it is a line rebuild. The same logic applied to controls specifically — what makes an imported line upgradeable at all — is in maintainable controls on an imported line, and the decision about when to replace rather than extend is in when to upgrade your PLC.
Four questions that settle it
- How many SKUs, and how often do you change between them? More than a handful of changeovers a week pushes hard toward semi-automatic or toward automation with proper quick-change tooling — which is a specification, not an assumption.
- Which steps have a tolerance a person cannot hold? Automate those regardless of the wage arithmetic. Include giveaway in the calculation.
- What breaks, and where does the part come from? Price the lead time of the three most likely automated failures before you buy them. Our spare-parts strategy covers how to build that list.
- Who maintains it on a Saturday? If the answer is one person, the automation has created a single point of failure that is not on the equipment list.
What this does not change
Two things hold at every automation level, and they are worth separating from this decision so they do not get traded away in it. Operators are part of the machine — a semi-automatic line run by untrained people produces worse output than an automatic one, and an automatic line with no trained fitter stops. And measurement is independent of automation level: you can capture real downtime and output on a semi-automatic line without a new controller, which is covered in measuring OEE on an old line without a new PLC. Choosing less automation is not a reason to choose less visibility.
What CISH does in this part of the process
We set the automation level per step rather than per line. That means going through the process route with you, identifying which steps carry a tolerance, a safety exposure or a record-keeping requirement, and automating those — then specifying the space, interfaces and controls headroom so the remaining steps can be automated later without rebuilding. Where a supplier's quotation is built on a headcount saving, we re-run the case on your actual wage structure, your SKU mix and your parts lead times, and tell you what it looks like then.
Frequently asked questions
Is a fully automatic line worth it in South Africa?
It depends far less on wages than the usual business case suggests. Direct labour here is comparatively inexpensive, so payback calculated on headcount is slower than the brochure assumes — while skilled maintenance labour is scarce and expensive, spare parts for drives and servos arrive on shipping lead times, and grid interruptions force restarts. The reasons that do survive are tolerance, safety and quality records. If a step needs a consistency a person cannot hold, automate it; if the only argument is fewer operators, test that arithmetic on your own numbers first.
What is the difference between a semi-automatic and a fully automatic line?
In a semi-automatic line people perform some transfers, loading or changeover steps between powered machines; in a fully automatic line those transfers are mechanised and the line runs as one sequence. The practical difference is failure behaviour: a semi-automatic line degrades when a station has a problem and some output continues, while a fully automatic line stops. That difference rarely appears in a payback calculation and tends to dominate the first year of operation.
Which steps should I automate first?
Three categories, in this order: steps where the specification is tighter than human consistency (fill weight, seal temperature and dwell, dose, torque — include over-fill giveaway in the sum, because it runs every hour the line runs); steps that are unsafe for a person to be near; and steps that decide or record quality, such as inspection, coding and traceability. The last group is usually the cheapest to add later and the most useful in an audit.
How do I buy a line now but automate more later?
Specify four things at purchase: floor space in the layout where the future station will sit; the mechanical interfaces it will connect to, meaning conveyor heights, discharge positions and transfer geometry; the control signals and spare PLC I/O it will need; and a network the new station can join. All four are close to free at design stage and effectively unbuyable afterwards. With them the upgrade is a purchase order; without them it is a line rebuild.
Does more automation mean fewer skilled staff?
It means fewer people and more skill per person. An automated line converts a dependence on several semi-skilled operators into a dependence on a millwright and a controls technician — exactly the roles South African manufacturers report the most difficulty filling. Before choosing a higher automation level, answer a practical question: who maintains this line on a Saturday, and what happens when that person resigns?