Sector notes
Robotics ROI on the SME floor: where it's real, and where it's oversold
A cobot handles the four-thousandth repetition exactly like the first. That is the entire case for buying one — and it stops being the case the moment the task changes more often than the cycle repeats.
Executive summary
35%
median local-staff salary rise at Panasonic's Singapore operation after three years of cobotisation
50–70%
of qualifying capital costs the Enterprise Development Grant funds
Core conclusions
- The case is real where high repetition, ergonomically poor work, and stable product geometry hold together — not separately.
- Low-volume, high-mix work fails quietly: changeover and reprogramming time consumes the labour saving, and without a recorded baseline nobody checks.
- Panasonic and Multi-Wing CZ both automated the repetitive station first and moved people into supervision and judgement work — a reallocation, not a headcount cut.
Most robotics proposals on an SME factory floor get evaluated on the wrong question. The one that matters is not whether the arm can do the task demonstrated at the trade show. It is whether that task repeats often enough, in a form consistent enough, for the capital to pay back before the product line changes underneath it.
Get that wrong and the cell sits behind a reprogramming backlog nobody budgeted for. Get it right and it removes exactly the shift nobody wants to staff.
Where the case is real
Three conditions, and they have to hold together, not separately:
- High repetition: the same motion, hundreds or thousands of times a shift, not a batch that changes weekly.
- Ergonomically poor work: repetitive lifting or awkward reach that is already producing the injury claims and the absenteeism, not just the labour line item.
- Stable product geometry, so the end-of-arm tooling and the programme do not need reworking every time the SKU changes.
Where it gets oversold
Low-volume, high-mix work is the case that fails quietly rather than visibly. A cell built to demonstrate flexibility across a handful of SKUs a month rarely earns back its capital, because the changeover and reprogramming time consumes the labour saving it was supposed to deliver. The project is not cancelled — it is simply never quite as productive as the proposal said, and nobody goes back to check because there was no baseline recorded to check it against.
What this looks like when it works
Panasonic's Singapore operation is worth studying because it names the mechanism, not just the outcome. Over three years, supported by the Economic Development Board, the company moved from manual lines to what it calls "cobotisation": workers retrained to work alongside the machines rather than displaced by them. The median salary of its local staff rose 35 percent as the workforce moved into the supervision, programming and exception-handling roles the automation created.
Multi-Wing CZ's Singapore plant, building tailor-made ventilation systems, put cobots on the strenuous, repetitive stations on its line and moved people onto the work that needed judgement — a reallocation of labour, not a headcount cut. Both cases automated the ergonomically poor, high-repetition station first, and neither tried to automate the variable work.
We have delivered the same shape of project ourselves: the Enterprise Development Grant-funded robotics line at Raymang Eggs automated the repetitive, high-volume handling stations, which is precisely the profile the EDG's own criteria ask for — a defined business problem, not a general-purpose flexibility upgrade.
The test before committing capital
| Question | What it should show |
|---|---|
| Repetition | Cycles per shift, sustained over months, not a single busy week |
| Variability | How many SKU or format changes the line sees per quarter |
| Consequence today | Injury claims, absenteeism or turnover already attributable to this specific task |
| Baseline | A recorded OEE or labour-hours figure before the cell goes in — without it, improvement is an impression, not a measurement |
The grant changes the arithmetic, not the decision
Enterprise Singapore funds up to 50 percent of qualifying costs under the Enterprise Development Grant for a standard SME project. From April 2023 a sustainability-linked project could claim up to 70 percent, a window that closed at the end of March this year. EDG itself is now on its way out: Enterprise Singapore is folding it into a single unified grant, EDGE, alongside the Productivity Solutions Grant and Market Readiness Assistance, launching later this year.
None of that changes which tasks are worth automating. It changes how much of the capital risk the operator carries while finding out. A grant makes a marginal case fundable; it does not make a bad case good. The falsifiable version of the argument above is this: a robotics proposal justified primarily on flexibility across a variable product mix, with no baseline measurement taken before commissioning, is the one most likely to be running below its proposed utilisation within a year — and it is checkable, on any given floor, by asking whether that baseline was ever recorded.