Four engagements over one journey — a warehouse, self-service BI, an AI assistant, and finance reconciliation — turned FishTalk, Innova, Navision and a fjord full of sensors into biomass forecasts that run from egg to harvest, and production decisions made weeks earlier.
Arnarlax is Iceland's largest salmon farming company, founded in 2009 in Bíldudalur, a village on the Arnarfjörður fjord in the Westfjords. The operation is fully vertically integrated — hatcheries, sea farms, harvesting and sales under one roof — which means the data is too: biology, feed, logistics and finance all describing the same fish at different stages.
Until this engagement, those descriptions lived in separate systems that didn't talk to each other. The partnership grew into four projects forming a single journey: centralise the data, open it to the departments, layer reporting and an AI assistant on top, and finally reconcile the finance numbers across every system that produces them.
1 — The data warehouse (6 months, 3 FTEs) centralised FishTalk, Innova, Navision, Maritech, custom APIs and IoT sensor data into a single platform: Azure Data Factory for ingestion, Azure SQL for the warehouse, Blob storage for staging, Azure IoT Hub for the sensors in the pens — with automated data-quality checks, because sensor data from the North Atlantic does not arrive clean. Estimated ROI: 150% in year one, payback in under a year, with efficiency gains showing up in feed cost, mortality rate and analytics-team capacity.
2 — Self-service BI (4 months, 3 FTEs) unified KPI definitions across departments and put Azure Analysis Services models on top of the warehouse — a one-stop data shop where departments answer their own questions. Report generation time and the data team's request backlog both dropped by 40%. 3 — Enhanced reporting and an AI assistant (12 months, 2 FTEs) built out Power BI on the KPIs that run a fish farm — ADG, FCR, feed consumption, dissolved oxygen, pH, temperature, labor utilization, equipment downtime, mortality — automated the regulatory and financial reporting, and added a data assistant on Azure Bot Service and Cognitive Services. Reporting time fell by 50%.
4 — Finance reconciliation (6 months, 2 FTEs) closed the loop: Navision against the custom accounting tools, Innova against the financial records, and FishTalk forecasts against actuals in Azure SQL. The result of the whole journey is biomass forecasting from egg to harvest — and production decisions made weeks earlier than before.
FishTalk, Innova, Navision, Maritech, custom APIs and IoT feeds each model the world differently; agreeing what a "batch" or a "site" means was work, not configuration.
IoT readings from pens in a North Atlantic fjord drop out, drift and spike — the automated quality checks earn their keep daily.
Self-service BI only works once every department accepts the same definition of FCR and mortality — alignment took longer than modelling.
Getting Navision, Innova and FishTalk to agree with the actuals — line by line — is the least glamorous project of the four, and the one finance relies on most.
Salmon farming is a business of long biological lead times and short decision windows. With biomass forecasting running from egg to harvest on one governed platform, Arnarlax makes production decisions weeks earlier than before — when they're cheap to make rather than expensive to correct. The efficiency gains land where a farmer feels them: feed cost, mortality rate, and an analytics team spending its capacity on analysis instead of extraction.
The warehouse alone returned an estimated 150% ROI in year one, with payback in under a year. Downstream, self-service BI cut report generation time by 40%, and the enhanced reporting layer with the AI assistant cut reporting time by 50% — including the regulatory and financial reporting that used to consume the quiet weeks. Numbers that hold up, on land and at sea.
Ingestion from six-plus source systems into the warehouse.
The central model — and the home of forecasts-versus-actuals.
Landing zone for raw extracts before quality checks.
Streams sensor data — oxygen, pH, temperature — from the pens.
Unified KPI models behind the self-service one-stop data shop.
Aquaculture KPIs plus automated regulatory and financial reporting.
The conversational front door to the data platform.
Language understanding behind the data assistant.