Skip to content
العربية

Service 06 · Data & Analytics

A number nobody trusts is not a number.

Most organisations have more reporting than they can reconcile. The problem is rarely the dashboard; it is that two systems disagree about what a customer is, and nobody decided which one was right. Noble does that work first and the visualisation afterwards.

Trust runs backwards

A forward pipeline shows that data moves, which nobody doubts. What earns trust is going the other way: taking one figure and walking it back to the governed records it came from — and a real figure rarely has one parent.

The estate, by capability family

Seven families, listed as a register: this is work a client needs to be able to point at line by line, and to check against what is already in place.

Data strategy and architecture

  • 01.1Data strategy and target architecture, before any platform is bought
  • 01.2Domain and ownership model — who decides what a term means
  • 01.3Build, buy and consolidate decisions across existing platforms

Data platforms and storage

  • 02.1Warehouse and lakehouse design
  • 02.2Storage layout, partitioning and retention
  • 02.3Environment separation, promotion and repeatable deployment
  • 02.4Cost and performance treated as design constraints

Data engineering and integration

  • 03.1Ingestion from operational systems, batch and streaming
  • 03.2Transformation pipelines with tested logic
  • 03.3Integration across ERP, finance and operational sources
  • 03.4Orchestration, scheduling and dependency management
  • 03.5Reconciliation back to the source of record

Modelling and semantic layers

  • 04.1Dimensional and semantic modelling
  • 04.2A defined metric layer, so one term means one thing
  • 04.3Historical handling and slowly changing data
  • 04.4Models the business can read without a translator

Governance, quality and lineage

  • 05.1Master and reference data management
  • 05.2Quality rules that fail loudly rather than silently
  • 05.3Lineage, so a figure can be traced to its origin
  • 05.4Ownership, stewardship and change control
  • 05.5Classification, access and residency decisions

Business intelligence and analytics

  • 06.1Operational and management reporting
  • 06.2Dashboards and visualisation built on the agreed metric layer
  • 06.3Self-service models built to be used unsupervised
  • 06.4Planning, budgeting and scenario comparison
  • 06.5Operational and near-real-time analytics where the decision needs it
  • 06.6Arabic and English reporting composed for both

Advanced analytics and decision intelligence

  • 07.1Advanced and predictive analytics on the operational record
  • 07.2Segmentation, cohort and behavioural analysis
  • 07.3Feature preparation for machine learning
  • 07.4Serving governed data to applications and to agents
  • 07.5SAP data and analytics relationships — BW, Datasphere and SAC
  • 07.6Interfaces back into the operational systems

What data passes through

  1. 01

    Source

    Where the figure is actually created, which is usually a transaction and not a report.

  2. 02

    Agree

    What a customer, an order and a cost centre each mean — settled once, by people, before any pipeline is written.

  3. 03

    Model

    The transformation, the metric layer and the tests that keep both honest as the source changes.

  4. 04

    Use

    Reporting, planning, a model, or an application — and the reconciliation that lets someone defend the number in a meeting.

Why this comes first

Almost every ambition an organisation has for its technology — prediction, automation, an assistant that can answer questions about the business — resolves into a question about whether its data agrees with itself. That work is unglamorous and it is the difference between a demonstration and a system.

What depends on this