Discover: mapping a system across a wall of diagramsDiagnose: reading dashboards to locate the problemDesign: wireframes and interface work for the solutionDeliver: shipping the finished productEvolve: continuous improvement after launchAnalytics flowing from raw data into decisions
Model · Pipeline · Trust

Data platforms

A data platform whose numbers two departments can agree on. Most reporting problems are not reporting problems: the same figure is computed three ways in three places, and the argument is about which copy is right rather than what to do.

  • One
    Definition per metric
  • Fresh
    Stated, not assumed
  • Trace
    Back to the source
  • Test
    On every load

Sectors that trust us

What is a data platform?

A data platform is the agreed place where a number is computed once and read everywhere. It is less about storage than about definitions: what counts as an active customer, which date a sale belongs to, and who is allowed to change those answers once other teams depend on them.

Key takeaways

  • One definition: every metric has one definition and one named owner.

  • Freshness shown: the last successful load time sits beside the number itself.

  • Tests on every load: a failed check stops the load and raises an alert.

  • Traceable figures: any number can be followed back to its source rows.

  • Loud failures: a half-finished load is an alert, never a chart that still renders.

What we build

A data platform, not just a warehouse with charts on top.

Each exists so a number can be defended in a meeting: where it came from, when it last moved, and who may change its meaning.

  • Dimensional modelling and marts

    Facts and dimensions that survive a new question, rather than one table per report.

  • Ingestion and transformation pipelines

    Scheduled, retried and idempotent, so a rerun can never double-count a single day.

  • A metric layer with one definition

    One definition per metric, read by every dashboard instead of copied into each one.

  • Data quality tests on every load

    Row counts, nulls and referential checks that fail the load rather than warn about it.

  • Lineage from source to dashboard

    A traceable path from the figure on the screen back to the rows that produced it.

  • Access control by role and row

    Who sees which rows, decided in one place and reviewable on request by an auditor.

Agents on your data

Analytics agents that read the metric layer and report first

Once the definitions and the tests exist, an agent can sit on top of them safely. It watches the metrics your team argues about, explains a move in plain language, and cites the rows it read before anyone opens a dashboard.

Discover: mapping a system across a wall of diagrams
01Definitions

The agent reads the metric layer, so it answers with the agreed definition.

Sectors we model for

One platform, shaped to your sector.

  • Retail

    Stock, margin and sell-through by store, by week and by line.

  • Banking and fintech

    Transactions, exposure and the reports the regulator asks for.

  • Healthcare

    Capacity, waiting times and outcomes, per clinic and per month.

  • Logistics

    Fleet, fuel and on-time delivery, by route and by driver.

  • Real estate

    Occupancy, collections and maintenance cost per building.

  • Public sector

    Service demand and case throughput, per office and per week.

What we build it with

Open tools, your cloud account, nothing rented from us.

  • PostgreSQL
  • dbt
  • Apache Airflow
  • Kafka
  • Snowflake
  • BigQuery
  • Databricks
  • Metabase
  • Power BI

How it works · our method

The SiaSol five-phase method, applied to this service.

  1. 01

    Discover

    Most technology projects fail long before the first line of code — they fail at the question.

  2. 02

    Diagnose

    Every slow process has an owner and a cost.

  3. 03

    Design

    Architecture is a set of decisions you live with.

  4. 04

    Deliver

    Software not in production is just cost.

  5. 05

    Evolve

    The handover is the point, not the end.

Where we work

Close where it counts, bilingual everywhere

  • Egypt: Cairo (HQ) and nationwide delivery

  • Saudi Arabia: Riyadh, with teams on client time zones

  • Gulf and beyond: Arabic and English, one delivery standard

Cost & engagement

Data platforms: what each stage costs and what it includes

Cost follows the number of sources and how messy their history is. The workshop and the first source are quoted apart and calibrate the rest.

  • Fixed · one to two weeks

    Definitions

    The ten contested figures written down, with their owners, edge cases and a source map.

  • Per source · quoted from the workshop

    First source

    One pipeline, modelled and tested, feeding real dashboards with the load time shown.

  • Monthly · dedicated

    Embedded team

    A standing team adding sources and keeping the tests green as the questions change.

FAQ

What to ask about Data platforms.

01Which warehouse do you use?

Postgres for most estates, because the data volume is usually far smaller than people assume and a managed Postgres is cheaper to run and easier to hire for. We move to a columnar warehouse when the measured query cost says to, not before.

02Can you keep our existing dashboards?

Usually yes, and it is often the right first step. The dashboards stop computing their own version of each metric and start reading the metric layer instead, which changes almost nothing visually and settles the argument about whose number is right.

03How do you handle historic data that is wrong?

It is loaded as it is and corrected in a separate, visible layer rather than edited in place. That way the original stays auditable, every correction has a reason attached, and nobody has to trust that a one-off cleanup script did what it claimed.

04What if two departments cannot agree?

Then both definitions are built, named differently, and shown side by side with their owners. That is a better outcome than one side quietly winning, and in practice seeing the two numbers side by side usually ends the disagreement within a month.

05How fresh will the numbers be?

Whatever you decide, and the figure is displayed beside the number so nobody has to guess. Most estates choose hourly for operational screens and daily for finance, because real-time costs more to run and rarely changes a decision made weekly.

06Can a figure be traced back to its source?

Yes, and that is a requirement rather than a feature. Any number on a dashboard can be followed through the metric definition and the model down to the source rows, which is what makes a disagreement resolvable instead of a matter of seniority.

07What do we own at the end?

The warehouse, the repository holding the models and tests, the schedules, the cloud account and the dated handover document. All of it is in your name from the first week rather than migrated across at the end as an administrative step.

Related services

Often paired with this

Learn more

From the journal

Ready to end the argument about the numbers?

Tell us the ten figures your teams argue about in meetings. We will come back with a written definition for each one, and a plan that makes the numbers agree.

Book workshop