A report per department, and no shared answer
Arnarlax — Iceland's largest farmer and producer of Atlantic salmon — runs complex aquaculture pipelines spanning fish health, feeding, logistics and finance. Each department had its own systems, its own reports and its own version of the truth.
Preparing those reports by hand ate the time that should have gone into reading them. Historical analysis was scattered and inconsistent between teams, and decisions that ought to have been informed by data were too often made on instinct — not because instinct was preferred, but because the data could not keep up with the business.
The platform that answered it
We designed and built a scalable data platform on Microsoft Azure, integrating fish health, feeding, logistics and finance into one coherent estate: Data Factory orchestrating ingestion, SQL staging and a warehouse behind it, transformation logic in stored procedures, Analysis Services as the modelled layer, and Power BI and Power Apps delivering insight and operational tooling to the people who need them. Document Intelligence took on the work of turning unstructured paperwork into structured data.
Manual reporting gave way to automated, unified consumption, and teams began working from the same numbers with historical context to hand. That platform did its job for years. What changed was everything underneath it.
Two migrations at once
Arnarlax replaced its core source systems and its data platform in the same stretch of time. Mercatus, Business Central and WiseFish took over from FishTalk, Navision and Maritech, while the estate itself moved to Microsoft Fabric.
New sources and a new platform is the riskiest kind of migration: every number changes both where it comes from and where it lives, so a figure that looks wrong could be a broken pipeline or a correct answer from a different system. The plan had to guarantee one thing above all — regular reporting resumes fast, and never silently breaks.
One Fabric platform, medallion all the way
Before
Azure SQL platform
Data Factory orchestration · SQL staging and warehouse · logic in stored procedures · Analysis Services semantic layer · Power BI reports.
After
Microsoft Fabric, unified
Fabric orchestration · OneLake storage · transformations in notebooks across medallion layers · Power BI Semantic Model replacing Analysis Services · the same Power BI reports on new rails.
The semantic model was rebuilt with an explicit old-to-new mapping at table level, so every migrated figure can be traced back to its former home. That is what lets a user open a familiar report and recognise it, rather than being handed a new world and asked to trust it.
Priority-first, phase by phase
Phase 1 delivered the must-haves: development and production Fabric environments as the foundation, ingestion of the new sources — Mercatus, Business Central, WiseFish, EQS and the Power BI API — the new semantic model, and the priority reports migrated across. Eleven Power BI reports moved, and five Finance Excel workbooks were reconnected. Good-to-haves were moved deliberately to later phases rather than allowed to delay day one.
Biology
All Phase 1 biology reports live on Mercatus
The domain closest to the fish — biomass, feeding, production — moved first and completely. Remaining gaps, including source history and newer data, were scoped into Phase 2 rather than allowed to block go-live.
Finance
Business Central live, every report in production
Finance reporting switched to the new ERP without losing its Excel workflows. The five monthly workbooks were reconnected to the new platform instead of being taken away from the team that relies on them.
Sales
Live, with history managed honestly
Sales reports run on WiseFish. Historical figures begin with the new system, so year-to-date KPIs are explicitly flagged until the Maritech history merge lands in Phase 2 — a known limitation, communicated, never a silent error.
People
Onboarding, not just deployment
User onboarding workshops closed the phase. The platform counts as delivered when the business is using it, not when the pipelines turn green.
What Phase 2 adds
Phase 2 extends from there: merging the Maritech history into WiseFish so year-to-date figures span both systems, deeper biology data, and a new board report. Alongside them comes a Data Agent — a governed “chat with your data” experience over the semantic model, so a question can be asked in plain language and answered from the same definitions the reports use.
Team
- —Andriy Zhubryd
- —Mateo Batur
Key objectives
Give the business one source of truth across fish health, feeding, logistics and finance, in place of a report per department.
Move the whole estate onto Microsoft Fabric while the source systems underneath it were being replaced.
Resume regular reporting fast, and never let a figure break silently.
Trace every migrated number back to where it used to live, so people see continuity rather than a new world.
Services
- —Data engineering
- —Platform migration
- —Business intelligence
- —Data analysis
- —Reporting
Technology
Numbers
16
Priority reports and workbooks live in production
3
New source systems onboarded in Phase 1
5
Finance Excel workbooks reconnected, not retired
6
Workstreams across Biology, Finance and Sales
