All projectsBusiness intelligence · Decision support

Real project

30+ trade sources that disagreed, compared on one dashboard.

A Power BI intelligence layer that reconciled trade and maritime sources, made discrepancies visible and gave senior analysts a shared global view.

30+sources reconciled
The problem
Customs and trade sources covered different periods, products and definitions. Comparing them by hand was slow and hid which source to trust.
What I built
A Power BI model fed by Snowflake that compares 30+ sources side by side and shows reliability next to the totals.
My role
Built the data model, Snowflake and Power Query transformations, DAX measures and report views. Junior Data Analyst.
Result
30+ sources compared in one place. Over 99% accuracy on validated datasets, measured through source-to-platform audits.
Why it was hard
  • Different source definitions and coverage
  • Crop-year, date and product filters interacting in DAX
  • Large production datasets
Tools
Power BI · DAX · Power Query · SQL · Snowflake

Market share analysis

  • Compared vessel-lineup and customs trade volumes across 6 markets to find reporting discrepancies.
  • Built country rankings and 24-month trend analyses of trade volumes and year-over-year change.
  • Prepared the analytical datasets in SQL and Snowflake, by country, product and trading counterparty.
Analyst working with a data spreadsheet on a laptop
Illustrative photo.

THE PROBLEM

Why the system needed to exist.

Trade sources covered different periods, products and reporting conventions. Comparing them manually was slow, and a polished chart could obscure the underlying reliability differences.

MY ROLE

Where I created leverage.

I shaped the data model, Snowflake and Power Query transformations, DAX logic and visual monitoring experience around the questions management actually needed to answer.

Technical details

SYSTEM FLOW

From friction
to a repeatable flow.

  1. Source datasets
  2. Snowflake queries
  3. Power Query
  4. Semantic model
  5. DAX measures
  6. Reliability checks
  7. Executive views

CONSTRAINTS

The difficult parts.

  • Different source definitions and coverage
  • Complex crop-year, date and product filter contexts
  • Large production datasets
  • The need to show uncertainty, not just totals

DESIGN DECISIONS

How I approached them.

  • Modelled shared dimensions so source comparisons remained consistent under filtering.
  • Separated data reliability signals from the headline business measures.
  • Used cumulative and regression-oriented views where time and crop-year context mattered.
  • Made refresh health, delays and error rates part of the dashboard experience.

OUTCOMES

Project outcomes

30+

sources compared

Across customs and international trade datasets.

Over 99%

validated accuracy

Over 99% accuracy on validated datasets, measured through source-to-platform audits.

Daily

visibility

Monitoring refreshed automatically.

TECHNICAL SURFACE

Power BIDAXPower QuerySQLSnowflakeData modelingData visualization
NEXT SYSTEM / 04

A month-long onboarding process, cut to 3–4 days.