Data management · Advisory and implementation

Metadata comes first.
Then everything else.

AI, data quality, compliance and modern data pipelines all rest on the same foundation: knowing what your data means, where it comes from and who is accountable for it. Tvibaka specialises in metadata management and data contracts, from strategy through to implementation.

Recently: data lineage for a large Nordic bank

Example · Data lineage Verified
asset
loans.total_balance
definition
Total loan balance at reporting date
owner
Risk Management
lineage
Core system → Data warehouse → Regulatory report
data contract
loans-total-balance v2.1 · active
quality
99.4 % completeness
last verified
12 Sep 2026
Focus areas
  • Metadata management
  • Data contracts
  • Data lineage
  • Data quality
  • Data architecture
  • AI in data pipelines
  • GDPR and NIS2
Why metadata first

Data without metadata is just numbers.

Most data projects do not fail on technology. They fail because nobody knows with certainty what the data means or who owns it. So that is where we start. Once the metadata is in order, everything else becomes simpler, faster and safer.

Example · one word, many meanings
A stick or pole used for support when walking
The vertical timber in a door frame, a door post
A character in alphabetic writing, a letter
A light-sensitive cell in the retina, a rod
A ray of sunlight falling through a narrow gap in the clouds, a sunbeam
One of the wooden boards a barrel is built from, a barrel stave
Whimpering, on the verge of tears
The hard shaft of a bird's feather, once used as a pen, a quill

Eight meanings, one word. In language, context settles it. In a database there is no context — only a column called stafur.

Meaning

That we understand each other

What exactly is an active customer? A shared business glossary and data catalogue make sure the same information carries the same meaning everywhere.

Origin

Every number has a history

Where does the number in the report come from? Lineage answers that in seconds rather than weeks, and regulators get the answer they asked for.

Accountability

Contracts instead of guesswork

Data contracts define what the producer of the data promises and the consumer can rely on: schema, meaning, quality, and who responds when something breaks.

Services

Baked twice. That is why it lasts.

A tvibaka is baked twice: first to take its shape, then to turn crisp and keep. We work the same way. Good data advice has to survive implementation, and good implementation needs a clear strategy behind it. We do both, with the same people from beginning to end.

First bake

Strategy and advisory

Metadata strategy

Which metadata actually matters, who maintains it and how it connects to the business. The foundation everything else is built on.

Maturity assessment and roadmap

An objective assessment of where your data management stands, with a prioritised roadmap that leadership can base decisions on.

Data governance

Who owns the data, who is accountable for its quality, and how decisions about it get made.

Architecture and tooling

Target architecture for your data estate, and a realistic assessment of the tools that fit your size, your requirements and your capacity.

Getting ready for AI

AI is never better than the data beneath it. We assess what needs to be in place before you set off.

Second bake

Implementation

Metadata management Core

A data catalogue, business glossary and both business and technical metadata that people actually use, wired into the systems themselves and kept current automatically.

Data contracts

Clear agreements between the teams that produce data and those that consume it: schema, meaning, quality expectations and accountability, defined in code and monitored automatically.

Lineage and data quality

Lineage from source system through to report, with quality rules and monitoring that stand up to regulatory scrutiny.

Modernising data pipelines

We use AI to analyse, document and rewrite legacy data pipelines, faster and with less risk than before, and always under expert review.

Handover and enablement

We do not leave black boxes behind. Your team takes over the solution, understands it and owns it.

Work

Recent work

Our clients often work in environments where confidentiality matters, so we describe the work without naming them.

Financial services

Data lineage at a large Nordic bank

scope
Lineage · Metadata · Implementation
role
Sub-advisor to a metadata platform
market
Sweden · Nordics

The challenge

A complex data estate in a large bank, and a requirement to trace the numbers in reports all the way back to their source systems.

What we did

We implemented automated data lineage on a metadata platform and mapped the flows from core systems through to reporting, working closely with the bank teams on site in Stockholm.

The outcome

Lineage that can be looked up instead of traced by hand, and a firmer footing for reporting and for changes to systems.

Pipelines and AI

Modernising data pipelines with AI

scope
Pipelines · AI · Architecture
role
Architecture and implementation

The challenge

Legacy pipelines, poorly documented and dependent on a handful of key people, slowed down change and raised operational risk.

What we did

We used language models to analyse legacy code, surface business rules and lineage, and move the pipelines onto a modern data platform, with testing and expert review at every step.

The outcome

Documented, traceable pipelines, less key-person risk and a shorter path to new data products.

Approach

How we work

No vast report that ends up in a drawer. We work in clear stages, and every stage delivers something you can use straight away.

  1. Conversation

    We listen first. What needs to be achieved, which requirements apply, and where does it actually hurt?

  2. Assessment

    A tightly scoped maturity assessment that gives a clear picture of where things stand and what matters most.

  3. Roadmap

    Prioritised actions, a realistic timeline and measures that leadership can follow up on.

  4. Implementation

    Short stages that deliver visible results, worked through with your people from day one.

A good start

Metadata assessment

A tightly scoped assessment: what metadata is available and what steps are needed to reach your goals, who is accountable and where the risk sits, together with a roadmap for the next twelve months. A good way to prepare for AI, lineage, NIS2 or a new data platform.

Find out more
What we stand for

Advice that can actually be implemented

We have written the strategies and done the implementation ourselves. That is why we only recommend what we know holds up.

 

AI used responsibly delivers results

We use AI where it saves time and reduces risk, never as decoration and never without human review.

 

International experience, Icelandic context

Experience from demanding international environments, applied to Icelandic realities.

About

One consultant, twenty years of it.

Tvibaka takes on selected engagements, delivered by the person you talk to. No layers in between, and no team of juniors sent in your direction.

Founder and principal consultant

Sigurdur Jonsson

Sigurdur has worked in data management for over two decades, with organisations in Iceland and internationally. The thread running through all of it is the same: data needs meaning, origin and an owner.

He has implemented data lineage at a large Nordic bank, led data strategy and data culture at an Icelandic bank, managed data initiatives for the City of Reykjavik, and worked in information management and product development in an international industrial setting. He currently works in data architecture at a public institution alongside his consulting through Tvibaka.

education
M.Sc. Information Management, Reykjavik University · B.Sc. Computer Science, University of Iceland
focus
Metadata · Data contracts · Lineage · Data strategy
sectors
Finance · Public sector · Manufacturing · Telco · Healthcare
based in
Reykjavik · engagements in Iceland and across the Nordics
Get in touch

Shall we talk about your data?

Professional advice :: Fairness :: Collaboration