Define
Set eligible line, order and calendar-month rules.
↗← ALL WORKEvery number.
INDEPENDENT PROJECT / SQL + PYTHON / 2026 / 02
Every number.
Accounted for.
A trading dashboard built on explicit definitions and reconciled results.
EXPLORE THE STORY ↓£10.00Mpositive invoice value
12 complete months · 19,141 eligible orders
01
THE BUSINESS PROBLEM
Can the business trust the dashboard?
Trading totals and customer analysis use different populations. Make the definitions visible before drawing conclusions.
Dec 2010–Nov 2011. Positive invoice value is not net revenue. Figures come from reconciled SQL and Python outputs.
View monthly data
02From question
WHAT I DID
From question
to useful work.
Build
Aggregate monthly and country metrics in SQL and pandas.
↗Reconcile
Match totals and quantify the impact of data-quality decisions.
↗03One set of numbers.
THE RESULTS
One set of numbers.
Clearer boundaries.
SQL and Python totals match. The UK accounts for 84.06% of invoice value; 16.34% comes from unidentified customers.
United Kingdom
Other markets
Unidentified
Identified
THE BUSINESS TAKEAWAY
Separate trading performance from customer behaviour. Review the duplicate policy before production use.
↗THE EVIDENCE / GO DEEPER
LET’S DISCUSS YOUR PROJECT ↗Methods, context & full project notes ＋
The business question
How is trading changing across months and markets — and can decision-makers trust that every KPI uses the same definition?
The pipeline
Profiled the source, quantified missing customer identities and cancellations, applied a documented exact-row deduplication policy, and selected twelve complete calendar months. Built SQL aggregates in SQLite and checked them against pandas totals.
Findings that change how the dashboard is read
The UK represents 84.06% of positive invoice value. Unidentified customers account for 16.34% of value, so a customer-only view would omit a meaningful part of trading. Retaining exact duplicate rows would add £23,571.39; the difference is quantified rather than silently discarded.
Recommendation
Separate trading KPIs from customer-behaviour reporting, disclose coverage, and review the duplicate policy with the data owner before production use. Monitor market concentration alongside growth.
Definition and scope.
December 2010–November 2011. Positive invoice value excludes cancellations and nonpositive prices or quantities, but includes eligible positive fee/adjustment lines. It is not net revenue, profit, or return-adjusted sales. December 2011 is incomplete and excluded.
Deliverables
Python pipeline, SQL queries, reconciled monthly and country aggregates, data-quality checks, metric dictionary, and a Power BI implementation starter. The interactive chart here reads the actual computed results; the Power BI starter is a specification, not a completed PBIX.
Data: Daqing Chen, Online Retail (2015), UCI. DOI 10.24432/C5BW33 · CC BY 4.0. Independent analysis of historical public data; recommendations have not been tested in a live business.
Al Siddique Bakers ↗
YOUR NEXT MOVE