No site wins on all four criteria. Rank them for your case before you visit anywhere, and rank raw material first.
RCN is 64–90% of your total cost depending on the nut price — a share set by the market, not by your efficiency — and access to it is the first of the documented causes of plant failure. The standing guidance is to locate near the RCN supply, in an industrial area, with labour proximity or staff transport.
The other three sort themselves by how measurable they are. Port distance is a freight and lead-time cost you can calculate. Power quality is a downtime risk you cannot. Labour is the constraint that arrives in year two.
Power
You need three-phase supply, and you need outage frequency and duration before you sign, not the tariff.
For orientation, a 10 TPD automatic plant runs on the order of 150 kVA at 380/415 V 50 Hz, with a 1,000 kg/hour steam boiler and around 50 HP of compressed air for the peeler.
Utilities have been recorded at only about US$20 per tonne of RCN — under 2% of total cost. So poor power does its damage through downtime and quality, not through the bill. An interrupted drying cycle or humidification stage costs kernels, not kilowatt-hours.
Size the generator for the process you cannot stop, not for the whole plant. That is a much smaller and much more affordable machine, and it protects the material already in process.
Water and effluent
You need process water for steaming and cleaning, and a lawful discharge route.
The problem effluent is waste water from the cooking stage: condensate mixed with leaked CNSL. Where shells are stockpiled on open ground, CNSL leaches into soil and reaches surface water — a finding from a study across 51 factories in eight countries. The documented remedy is decanting followed by phyto-purification, which proved effective in practice, plus a bunded shell store to stop run-off.
A genuine data gap: no BOD, COD, pH or TSS figure specific to cashew effluent exists in the published record, and this guide will not invent one. Characterise your own volume, then take your numeric limits from your own pollution-control consent before designing treatment.
The same applies to stack emissions. Shell-fired boilers produce black smoke because the CNSL does not burn completely; the documented fix is compact cyclones and gas scrubbers, locally manufacturable, removing up to 90% of particles.
Labour catchment
Count the working-age population inside a commutable radius, not the district total, and assume a workforce that is 80–90% female.
Then plan for the failure mode rather than the headline. A large Ghanaian processor with more than 2,000 employees traced a sustained productivity decline to lateness and absenteeism, and fixed it with satellite shelling stations, staff buses, a productivity bonus and a canteen.
The site-selection lesson is the one worth carrying: the factory goes where the power and the road are; the shelling goes where the people are. Those need not be the same place, and treating them as one site is what forces the compromise.
Zoning and environmental assessment
The environmental assessment comes before the building, and in most regimes it gates construction itself. It is also the only approval whose answer can be “not here”, which is why it belongs on the critical path from the day you have a site and a nameplate capacity.
Three items reliably draw an assessor’s attention: the shell-fired boiler, the cook-station condensate carrying CNSL, and the shell stockpile. Have an answer for each before you file.
Fee concessions and assessment thresholds change on political timetables, so confirm what is current with the authority rather than working from any published figure — including this one.
Land: buy, lease, industrial park or free zone
Buying gives you collateral, which matters more than it first appears. Where banks require land title as security, processors without it are pushed onto short facilities and end up running three to six months a year.
Leasing preserves cash.
An industrial park or reserved industrial land gives you power, water and road access without the connection fight — several producing countries reserve serviced land in cashew regions specifically.
Free-zone or EPZ status is the biggest fiscal prize and carries the sharpest condition. Packages commonly include a multi-year corporate tax exemption, VAT exemption on capital goods and withholding tax relief on foreign loan interest and dividends. But these regimes are export-conditioned by design: output is expected to be exported, and domestic sales are typically treated as imports and taxed accordingly.
That bites where you need it least. Year-one off-spec production and broken grades often sell best locally, and a zone condition can turn your most accessible market into your most expensive one. Get the current instrument from the zone authority before you sign.
Logistics
Container access and road quality decide your real port distance. A 20-foot container holds 18–20 t of RCN or 14–15 t of kernels, and one full container load is the practical minimum kernel shipment.
Freight is a live exposure rather than a fixed cost, and container index movements plus peak-season surcharges can swing a meaningful share of a net margin that typically runs 4–8%. Sell FOB where you can, or index the freight in the contract.
And check that your access road takes a loaded 40-foot trailer in the wet season — because your harvest window is often the rainy one.
Score the sites before you see them
Score each candidate 1–5 against weighted criteria, multiply, and total. Set the weights for a first-plant, single-origin, export-oriented project — raw material heaviest, then power reliability, then labour catchment, then port and road, then land tenure and utilities.
Change the weights if your case differs. But change them before you visit the sites, not after. A scoring matrix built once you have a favourite is a justification, not a decision.