Skip to content
Davide Scuteri Moretti
Senior Data Architect
Chapter 02
M / 07

Method

A repeatable process for turning organisational questions into governed, reproducible computational systems.

Delivery Stages
1
01 · Scope

Frame the question

Turn a business problem into an answerable computational question.

2
02 · Semantics

Design the Data Objects

Meaning declared before movement — semantic contracts govern computation.

3
03 · Lineage

Make lineage executable

From conceptual diagrams to the actual computational path.

4
04 · Quality

Encode quality as clauses

Quality is computed, not surveyed.

5
05 · Governance

Governance that executes

Rules encoded as code, not stored as prose.

6
06 · Reproducibility

The right to recompute

Every published claim can be reproduced from evidence and versioned logic.

7
07 · Handover

Institution, not project

The platform outlives the delivery team.

01
01 · Scope

Frame the question

Before touching data I clarify the decisions a platform must support, the actors who will rely on it and the acceptable evidence for each claim. Ambiguity here compounds later; explicit scope prevents it.

Activities
  • Stakeholder interviews and decision mapping
  • Definition of critical questions the platform must answer
  • Identification of governing bodies and evidence requirements
  • Assessment of existing sources and their fitness
Deliverables
  • · Scope memorandum
  • · Decision map
  • · Source assessment
02
02 · Semantics

Design the Data Objects

Each Data Object carries an identity, a semantic definition, a governance owner and a contract. These are the fixed points against which pipelines are validated. Semantics precede storage decisions.

Activities
  • Definition of Data Objects and their contracts
  • Domain modelling and versioning strategy
  • Governance ownership assignment
  • Quality clauses per object
Deliverables
  • · Data Object catalogue
  • · Contract schemas
  • · Versioning policy
03
03 · Lineage

Make lineage executable

Lineage is not documentation. It is the executable path from source field to governed object. I model it as a DAG of computational tasks with declared inputs, outputs and effects.

Activities
  • DAG design of computational tasks
  • Provenance capture at every task boundary
  • Encoded transformations, not narrated ones
Deliverables
  • · Executable lineage graph
  • · Task registry
  • · Provenance schema
04
04 · Quality

Encode quality as clauses

Every contract carries clauses — completeness, timeliness, referential integrity, semantic validity. Dispositions are declared per clause: reject, quarantine, warn. Quality becomes an executable property of the platform.

Activities
  • Definition of quality clauses per Data Object
  • Disposition strategy for each clause
  • Instrumentation for continuous evaluation
Deliverables
  • · Quality contract
  • · Disposition policy
  • · Quality telemetry
05
05 · Governance

Governance that executes

Access, privacy and stewardship become executable predicates enforced at the boundary of every Data Object. Documentation is generated from the same source of truth as enforcement.

Activities
  • Encoded access predicates and stewardship rules
  • Privacy-by-design filters and pseudonymisation
  • Audit trails emitted by the platform itself
Deliverables
  • · Executable policy layer
  • · Audit stream
  • · Generated documentation
06
06 · Reproducibility

The right to recompute

Idempotence, deterministic tasks and versioned artefacts allow any output to be recomputed. This is the platform's promise to auditors, scientists and its own future maintainers.

Activities
  • Idempotent task design
  • Versioned code, contracts and artefacts
  • Preserved evidence for every published output
Deliverables
  • · Reproducibility ledger
  • · Versioning strategy
  • · Recompute runbook
07
07 · Handover

Institution, not project

A Data Platform succeeds when the organisation can continue to evolve it. Handover means transferring not just artefacts but the discipline: the contracts, the review rituals, the governance forums.

Activities
  • Team enablement and pairing
  • Establishment of contract review rituals
  • Definition of continuous evolution practices
Deliverables
  • · Enablement plan
  • · Review rituals
  • · Evolution roadmap
Working Principles

Cross-cutting rules I apply at every stage.

P.01

Meaning before movement

No pipeline is written until its output contract exists. Semantics precede storage.

P.02

Evidence before interpretation

The platform preserves raw evidence so any interpretation can be re-examined.

P.03

Encoded, not narrated

Rules that matter are code. Prose in catalogues does not execute.

P.04

Idempotence by default

A task run twice on the same inputs produces the same outputs. This is not optional.

P.05

Governance is a first-class citizen

Access, privacy and stewardship are part of the architecture, not overlaid on it.

P.06

Small, composable tasks

The atomic unit is the task. Complex platforms emerge from their composition, not from monoliths.

Continue