Data platform & governance
One data foundation rather than a sum of silos: shared reference data, infrastructure described as code, traced access and compliance that holds over time.
Scattered data is not governed, it is endured.
Data sits across applications, entities and countries. Without a single reference, each domain lives its own truth and nobody can say where a figure came from.
Governance therefore starts with a foundation: one platform, shared reference data, infrastructure and transformations versioned like code. What follows, quality, analysis, forecasting, only holds on top of it.
What we do
Four workstreams that turn a scattered estate into a foundation you can steer.
Data platform
One platform, cloud or hosted in-house depending on sensitivity, where ingestion, storage and transformation live in the same place rather than in as many tools as there are teams.
Shared reference data
Business reference data becomes common to every domain. That is what lets two departments talk about the same customer, the same product and the same site.
Data as Code
Pipelines, schemas and permissions are described, versioned and replayable. A change is read in a commit rather than reconstructed afterwards.
Catalogue and lineage
Every dataset carries its origin and its path. That is what makes a figure explainable and access auditable, the condition for GDPR and AI Act compliance to hold.
What you get
Estate mapping
The survey is delivered as a file, not a slide deck: sources, systems, an owner per domain and gaps ranked by value and risk. What has no owner here will have none afterwards.
The foundation as code
The platform is described and versioned rather than clicked: zones, retention, encryption, audit logs. Data residency becomes a line you can read back, not a decision you reconstruct.
Reference data and survivorship rules
Shared reference data means written rules: business key, confidence levels, human arbitration where it is needed, and the authoritative source field by field.
The check before go-live
Lineage, owners, residency and logs are verified before opening. A domain with no arbitration rule blocks go-live: it would produce a golden record nobody could defend.
Foundations in production
A data software factory augmented by GenAI agents
×2 delivery velocity · 100% Row Level Security coverage
A rising volume of business requests that data delivery could no longer absorb, and a data estate piling up, with no shared semantic model and no end-to-end security.
A structuring semantic model, shared standards and Row Level Security on sensitive domains, with specialised agents across the cycle and skills transfer from the first sprint.
DevOps and DataOps industrialised on Azure
Automated pipelines · Governed cloud platform
Data flows and deployments handled without a shared chain, where each change had to be replayed by hand and nothing guaranteed two environments behaved the same.
A DevOps and DataOps chain built on Azure: industrialised pipelines, environments described as code, deployments made reproducible rather than replayed.
Insights & Perspectives
Data governance in digital transformation
What governing data means day to day: ownership, quality, lineage, and the arbitration between compliance and the pace the business needs.
Principles of a modern data architecture
Designing, structuring and securing the data infrastructure: sharing, built-in security, fewer copies and a common vocabulary.
Data Mesh: the architecture reshaping enterprise data
Rather than centralising everything, data mesh hands the data back to the business domains that produce it, as governed products.
The rest of the data journey
Lay the foundation first
A survey of your estate, shared reference data and governance that holds over time.
Frequently asked questions
With reference data, not with rules. As long as two departments do not talk about the same customer, no quality rule holds: it will be applied to two different versions of the same entity.
Describing pipelines, schemas and permissions in versioned files rather than in interfaces. A change is read in a commit, reviewed and replayed, instead of being reconstructed afterwards.
The ability to explain a figure. Lineage traces where a value came from and what transformed it, which turns a dashboard number into something defensible in front of an auditor.
It depends on the number of domains and their real autonomy. A mesh hands data back to the domains that produce it, which assumes each has an identified owner: without that, it scatters rather than empowers.
By a routing rule written into the infrastructure itself, domain by domain and by sensitivity. It is versioned with the rest of the foundation, so a change can be reviewed and replayed. Residency decided after the fact gets bypassed by the first flow in a hurry.
By tying it to something that already happens, rather than adding a task. A catalogue updated at a quarterly review is wrong by week five, because nobody is paid to maintain it between reviews. Tying it to deployment changes that: a table reaching production without its owner, its description and its classification does not pass the gate. The cost is then paid by whoever creates the table, at the moment they know the answer, instead of by a governance team six months later who will have to guess it.
Treat it as a candidate for deletion, and say so. Data without an owner is data nobody uses enough to answer for, or whose real use sits somewhere other than where you are looking. We list those sets with their volume, their storage cost and their last access, then propose a deadline before removal. In practice half of it disappears without complaint, and the other half finds an owner within days: it is the removal notice that produces the arbitration the ownership request never did.