Singapore Aquaculture Intelligence Report

Research signal · Singapore · 27 Oct 2026

Tropical aquaculture R&D: what buyers should compare

The Singapore Food Agency (SFA) announced a new research allocation and an exploratory R&D memorandum of understanding (MOU) with Japfa Pte. Ltd. on tropical aquaculture. Neither creates a purchasing opportunity. Both are a reason to ask every offer for the same, comparable evidence.

Comparison fields to request from every offer

Use neutral offer references. Provider identities and evidence documents stay in the confidential, human-reviewed process.

Species and life stage

Exact species (and line or strain where relevant) and the life stage covered: broodstock, fry, fingerling, nursery or grow-out.

Evidence for one species or stage does not transfer to another.

Water conditions

Salinity, temperature range, dissolved oxygen, pH/alkalinity and water source in the trial and at your site, with tropical conditions stated explicitly.

Different water conditions make results non-comparable.

Trial design

Number of farms or units, replicates, comparison group, duration, who ran the trial and whether it was independently reviewed.

A single demonstration is not a controlled trial.

Disease baseline

Pathogens present or screened, diagnostic method and laboratory, and the site's prior disease history during the trial.

Without a baseline, a lower disease rate cannot be attributed to the product.

Sensor and detection accuracy

Stated accuracy, sensitivity and specificity against a reference method (laboratory test or manual count) under your water conditions, plus drift and fouling behaviour.

Accuracy measured in other water or species is not your accuracy.

False-alert handling

How false alarms and missed detections are measured and reported, who reviews an alert, and which qualified person decides the response.

An AI alert is a prompt for checking, not a diagnosis.

Offline operation

Which functions keep working without internet or cloud access: local alarms, data buffering and manual fallback during power or network loss.

Cloud-only functions stop when connectivity fails.

API and data ownership

Export formats, API access, who owns farm data and any models trained on it, data retention, and deletion when the contract ends.

Unclear data rights can lock a farm into one provider.

Cybersecurity

User access controls, update and patch policy, encryption, remote-access arrangements and incident notification.

Connected control systems add operational risk that needs an owner.

Local calibration and service

Who calibrates sensors and models locally and how often, response times, spare parts and operator training in your country.

Remote support alone may not keep sensors or models calibrated.

Feed-conversion assumptions

FCR basis (biological or economic), feed type, fish size range, temperature, period and how feed and biomass were measured.

FCRs on different bases cannot be compared directly.

FCR calculator

Mortality evidence

Survival or mortality records by stage, with counting method, period, cause attribution and a comparison group.

Extrapolated or uncounted survival claims are not evidence.

Survival and harvest guide

Traceability records

Lot or batch identifiers, origin and movement records, what is captured automatically versus manually, and whether records export for buyer or authority checks.

A traceability label does not prove record completeness.

Scale-up and commissioning plan

Steps from trial to farm scale, acceptance tests, commissioning responsibilities, staff training and what happens if performance is not reproduced.

Pilot results are not a commissioning guarantee.