Plantation and Mill Economics

# Plantation and Mill Economics The numbers that shape what Rimau can sell at turn two of [[Rimau/The Flywheel and Rimau Sentient|the flywheel]]. ## The mill: OER is everything > [!important] One percentage point of OER > **Oil extraction rate** cuts input cost by roughly **RM200 per ton of oil** per point. OER ranges from ~19% (bad mill) to 21%+ (good mill) — and that spread is the difference between losing money and roughly **RM700 per ton of headroom**. **A mill can only lose oil, never make it.** The loss points: - Injury to fruit at the ramp - **Steriliser** — ~1%, and timing is experience-based fine-tuning by the mill manager - **Press fibre** — 2–3% Mills cost **RM50–70M** to build and are a volume game, targeting 80–100% capacity. ## Price forecasting as a live pain MPOB issues a **monthly average price** that mills can commit volumes to. Forecasting errors force spot sales — so forecasting accuracy is a named, quantified mill pain today. Local mills behave like a cartel on the posted price, then compete through side incentives. This is the clearest first AI use case after compliance: the data comes from the same customer, and the ROI is arithmetic. ## The grading loop Mills grade incoming fruit — unripe and rotten rejected — weigh trucks in and out, and pay only for accepted fruit. **Rejected fruit rides home at the plantation's transport cost.** Some estates send staff to the mill to learn its grading standard and align. That is the tightest optimisation loop between plantation and mill, and an open question the team wants quantified with real data. > [!warning] The invalidated assumption > **AI grading alone can double cost rather than cut it** if humans must still act on the output. Identification first, actuation later. Collection centres exist for haulage economies of scale (20-ton trailers vs 5-ton lorries) and usually have a mill behind them. ## The plantation lifecycle ```mermaid graph LR S[Seeds licensed DxP only] --> N[Nursery 1 year] N --> C[Land clearing excavators; burning is criminal] C --> T[Terracing and drains] T --> L[Lining ~135 palms/ha] L --> I[Immature 3 years] I --> P[Peak yield years 7-18] P --> R[Replant year 25] R --> S ``` - **Density** — ~135 palms/ha is the sweet spot. Up to 148 gives fast early yield but forces thinning after year ten. - **Seeds** — licensed producers only in Malaysia; DxP crosses of dura and pisifera. Generational claims of up to 24% oil and 27 t/ha are *lab numbers*; real fields never hit them. Seed **brand** matters less than assumed; licensed **generation** matters more. - **Nursery disease control is the hardest problem in the whole operation** — especially for third-generation clones bred for oil at the cost of resilience, and especially in rain. Pests: rats, birds, rhinoceros beetle (kills the shoot), wild boar, monkeys. - **Ganoderma** drives the requirement to remove roots at replanting. > [!failure] The market that looked real but isn't recurring > Drone and contour-mapping tools that auto-plan roads, terraces and lining exist and work — but a full replant happens **once every 20–25 years**. Per-estate demand is episodic, not SaaS. Mapping freelancers are increasingly available anyway. ## The staffing consequence of MSPO Estate structure moved from *manager + two field assistants* to an added **administration and sustainability assistant** with staff below — a whole function created by regulation, and Rimau's user. ## Related - [[Rimau MOC]] - [[Rimau/Palm Oil Value Chain Primer]] - [[Rimau/The Flywheel and Rimau Sentient]] - [[Rimau/Workforce and the Social Layer]] - [[Agritech]]