Debottlenecking begins with measurement, not with a machinery quotation. Measure actual throughput at every stage across a full week — in tonnes per hour of running time and in tonnes per calendar day — and compare both against your line-balance design.
The constraint is the stage with product queued in front of it and idle capacity behind it. That is the only definition that matters, and it is frequently not where the operations manager thinks it is.
Expect two answers that are not machines
Labour availability. Absenteeism is the largest documented operational challenge in this industry. A plant that could run with 60% of its current staff if absenteeism were eradicated does not have an equipment problem, and buying a faster peeler will not fix it.
Raw material. A plant that stops in month nine is constrained by its buying plan and its working capital. No amount of debottlenecking inside the fence changes that, and the money spent trying would have been better placed in the buying facility.
Where the constraint genuinely is a machine, it is usually peeling or grading, because manual finishing never fully goes away. Peeling machines are reported by factories at rates as low as 70% against manufacturer claims of 80–95%, and mechanised grading gives five to eight grades against the twenty-four or more the trade recognises.
Yield projects, ranked by return
Rank with the arithmetic rather than by enthusiasm. One production KOR point is worth about US$68 per tonne of RCN; one whole-kernel point about US$2.53.
| Project | Mechanism | Cost | Value per 1,000 t RCN |
|---|---|---|---|
| Calibration discipline and blade adjustment by size class | Fewer uncut, less breakage | Near zero — supervision only | 2 whole-kernel points ≈ US$5,060 |
| Blade sharpening interval set from data | Same, sustained | Consumables and labour | 1–2 points ≈ US$2,530–5,060 |
| Steam parameter tuning on your own nuts | Cut-open rate up, breakage down | One month of controlled experiment | Varies — measure it |
| Dryer tray rotation and air circulation | Less scorching, better colour class | Retrofit fans on older dryers | Grade mix, not yield; often larger than either |
| Humidification control | Better testa release, less hand finishing | Instrumentation | 1–2 peeling breakage points |
| Reducing rework | Less double handling, less breakage | Process discipline | Labour and breakage together |
Do the free ones first, and measure each in isolation. A plant that runs six improvements at once learns nothing from any of them, and will repeat the expensive one next season because it got the credit.
Cost reduction that does not cost quality
Shell is the reliable one. Factories consume only 5–25% of the shell they produce for their own thermal needs — a surplus of four to twenty times over. Burn what you need and sell the rest.
Beyond that: measure energy per tonne so boiler efficiency becomes visible; rationalise packaging specifications against what buyers actually require rather than what the first buyer once asked for; and review the standing overtime that accumulated around a constraint you have since removed.
What not to cut: QC staffing, blade replacement, moisture metering, retention samples, and security. Each is cheap relative to the loss it prevents, and each is the first thing a cost programme reaches for.
Automation retrofits, in order
Sorting first. Colour is the first axis of grading and it drives price directly. Optical sorting is the one mechanised grading function that consistently outperforms hand sorting on consistency. Hand sorting survives regardless.
Steam second. Consistency at the front of the line propagates through every stage after it, and a steam vessel with reliable pressure and time control removes a variable that otherwise contaminates every experiment you run downstream.
Then peeling, then shelling — and only against measured constraint data.
Size the ambition against the thresholds
Published viability thresholds are blunt but useful: standalone machines from about 500 t/year, semi-mechanised viable from about 1,200 t/year and described even then as “not ideal”, full mechanisation needing 10,000 t/year or more.
Note what that does to nameplate arithmetic. A 30 t/day plant on a 120-day season is a 3,600 t/year plant — well below the full-mechanisation threshold whatever the plate says.
And keep the economics honest. Mechanisation saves roughly US$30 per tonne against a competitiveness gap with Asia of US$150–350 per tonne. It is a throughput and consistency decision, not a cost-parity fix.