Una Marie S. Goksøyr · Data, quality and systems · Norway

I build the systems behind decisions people can trust.

Most organisations have the data. What they lack is a system around it: clear ownership, checks that catch errors early, and answers shaped for the people who need them. That is the work I want to do more of.

See a working example How I work
BackgroundMechanical engineering, master's in safety management, quality roles in maritime, construction and public sector
NowQA and data analytics: connecting CRM, ERP and legacy systems
StrengthSeeing the whole system: process, people, data and controls
ToolsMicrosoft Fabric, PySpark, SQL, Power BI, Next.js, Supabase
Approach

Start with the problem, not the tool

A systems engineering way of working, scaled to the size of the problem.

  1. 1

    Understand the problem and the users

    Who makes the decision, how often, and what goes wrong today? A seller needs a call list, an auditor needs evidence, a manager needs a trend.

  2. 2

    Design the system

    Map the data flow, agree on definitions, decide who owns each part, and place the controls where errors start.

  3. 3

    Build what fits

    A report, a daily list, an alert or a small app. Best practice, but only as much as the problem needs.

  4. 4

    Verify and improve

    Quality gates before data is written, tests before release, and feedback from the people who use it.

Projects

Selected work

Portfolio project · Microsoft Fabric

Solvi: from quote to kilowatt-hour

In progress

Solvi is a fictional Nordic solar and battery installer. All data is generated.

One data system for a company where sales, installation, service and quality all depend on the same facts, but each needs a different answer. Quotes, orders, installations, production, weather and compliance records are joined once, checked once, and then delivered in the form each group can act on.

WhoQuestionWhat they get
SellersWho should I call today?A short daily list, biggest open quotes first
Sales leadWhat happens to what we quote?Win rate and value lost after a yes, by product
Service teamIs it broken, or just cloudy?Installations below what their own weather allowed
Quality and complianceAre we audit-ready?Missing paperwork, open deviations, worst first

Try the working example. Four views, one per user group, built on the same checked data.

Open demo
System choices
  • One definition per number, shared by every view
  • Quality gate: nothing is written if a check fails
  • Weather and snow in the model, so faults are not confused with cloudy weeks
  • Compliance stored as data: requirement, due date, evidence, deviation
  • Everything in Git, with separate test and production
Ontology, a first sketch 
Customer     receives    Quote
Quote        becomes     Order
Order        delivers    Installation
Installation produces    Energy
Site         experiences Weather
Installation needs       ServiceVisit
Microsoft FabricPySparkSQLPower BIGitOntology
Client project · Web app

Freya Hundepensjonat: booking and daily operations

Live

A booking and operations system for a dog boarding kennel in Hareid. A hectic workplace with people and their pets needed a website and an app, with a separate interface for each group that uses it. The goal was fewer interruptions on the phone and more time for the dogs.

WhoWhat they needWhat they get
Dog ownersBook without callingRegister their dog, upload vaccination records, book stays and send messages
The ownerControl without chasingReview and approve bookings, see at a glance what is waiting
StaffKnow the day, hands fullA simple view of who arrives, who goes home and what each dog needs
Details that mattered
  • Customers always see what the next step is
  • The owner sees at a glance what is waiting for approval
  • Regular daycare customers can book a day in a few clicks
  • Planning considers each dog's needs and who gets along
Built with
Next.jsVercelSupabaseResendGitHubAutomated tests

freyahundepensjonat.no

Principles

Best practice, sized to the problem

The right-sized answer

Not every problem needs a platform. Sometimes a checklist or one well-placed alert does more than a dashboard.

Controls where errors start

Checking at the source is cheaper than explaining a wrong number later. Validation runs before data is stored.

Clear ownership and definitions

Every number has one definition and an owner. That is what lets people trust it, and what makes data ready for AI.

Built with the people who use it

Test early with real users. The small details, like always knowing the next step, decide whether a system is used.