Insight Analytics
Service

Data Warehousing

A dimensional model that answers questions consistently no matter who asks or which tool they ask from.

A warehouse is a business asset, not a technical one. Its job is to encode how your company defines a customer, an order, a period, and a margin — once — so every report inherits the same meaning.

We build Kimball-style dimensional models: conformed dimensions shared across marts, fact tables at a declared grain, and slowly changing dimensions where history genuinely matters. Everything is documented, tested, and version controlled.

The result is that a new report takes days rather than weeks, because the modelling work has already been done and agreed.

Problems this solves

What we usually walk into

  • Every reporting request turns into a bespoke data-gathering project.
  • Historical changes are overwritten, so you can't report as-of a past date.
  • 'Customer' means something different in each department's report.
  • Analysts write six-join queries against transactional tables.

Technologies

What we build with

SnowflakeMicrosoft FabricAzure SynapseSQL ServerdbtPower BI

Industries that benefit

  • Financial services
  • Retail
  • Manufacturing
  • Education
  • Real estate

Implementation process

How the engagement runs

  1. 01

    Business process mapping

    Identify the processes to model — orders, claims, encounters — and declare the grain of each fact.

  2. 02

    Dimension design

    Conformed dimensions with surrogate keys and an explicit strategy for tracking change over time.

  3. 03

    Build and load

    Models built in dbt or native SQL, loaded incrementally with full auditability.

  4. 04

    Semantic layer

    Business-friendly views or a Power BI model so end users never touch raw tables.

  5. 05

    Documentation and governance

    A data dictionary, lineage, and ownership assigned per subject area.

Sample screens

What the finished work looks like

Representative layouts using demonstration data — client work is never shown without written permission.

Mart catalog

Marts

9

Facts

23

Dims

31

Mart catalog

Subject areas, owners, and refresh status at a glance.

As-of reporting

History

5 yrs

SCD2 dims

8

Restates

instant

As-of reporting

Restated prior periods using type 2 dimension history.

Query patterns

Users

180

Queries/day

6.2K

P95

1.4s

Query patterns

How the warehouse is actually being used, by tool.

Illustrative example

How a data warehousing engagement typically plays out

Anonymised scenario · not a verified client record

Commercial real estate owner-operator, 46 properties

Challenge

Occupancy and NOI reporting was rebuilt from the property management system each quarter, and when a property was reclassified the prior year's figures silently changed, making comparisons unreliable.

Solution

A Snowflake dimensional warehouse with a type 2 property dimension, a monthly financial fact at property-and-account grain, and a Power BI semantic layer on top.

Result

Quarterly reporting now runs from the warehouse in minutes, and prior-period comparisons stay stable because reclassification is versioned rather than overwritten.

Illustrative figures

5 yrs

Restatable history

3 wks → 1 day

Quarterly pack

46

Properties conformed

This is a composite illustration of the scope, approach, and range of results this service is designed to deliver. It does not describe a specific named client, and the figures are demonstration values rather than audited outcomes. We're happy to talk through real references under NDA on a call.

Deliverables

What you receive

  • Dimensional model design document
  • Built and loaded fact and dimension tables
  • Slowly-changing-dimension history where required
  • Business-facing semantic layer
  • Searchable data dictionary with lineage

FAQs

Questions we get asked

Kimball or Data Vault?
Kimball for most mid-market companies — it's simpler and analysts can read it. Data Vault when you have many source systems and strict auditability requirements.
How long until we see something?
The first mart is usually live in four to six weeks; later marts land faster because dimensions are already conformed.
Do we need to move to the cloud?
No. SQL Server on-premises is a perfectly good warehouse platform when your volumes and compliance profile point that way.
How do you handle history?
Type 2 dimensions where as-of reporting matters, type 1 where it doesn't. We decide that per attribute with the business, not globally.

Related services

Often scoped alongside this

Data Engineering

Lakehouse, streaming, orchestration, and observability.

View data engineering

SQL Development

Enterprise-grade SQL, stored procedures, tuning, and modeling.

View sql development

Business Intelligence Consulting

Strategic BI advisory, KPI definition, and analytics roadmaps.

View business intelligence consulting

Talk through your Data Warehousing project

A 30-minute call is usually enough to tell you whether this is a two-week fix or a two-month build — and roughly what it costs.

Book a consultation Browse all services