Small, reproducible Python tools that show how we approach data cleaning, mapping and reconciliation. Every example uses fictional records. These are demonstrations, not completed customer projects.
Trims whitespace, normalises three explicitly supported date formats and removes exact duplicate normalised rows. Invalid dates and amounts are flagged for review, never silently guessed.
Deliverables shown: cleaned CSV and JSON audit. Checks: 3 automated tests in the original demo.
Maps named source columns to a declared output schema. Retains every row, reports missing values and exact duplicates, and excludes unmapped fields only by explicit design.
Deliverables shown: mapped CSV and JSON audit. Checks: 5 automated tests in the original demo.
Matches inventory SKUs against a catalogue using case-insensitive, trimmed keys. Flags duplicates, missing matches and blank keys; preserves ambiguous records for manual review.
Deliverables shown: reconciliation CSV and JSON audit. Checks: 6 automated tests in the original demo.
Audits a fictional six-row CSV for blank fields, repeated rows and formula-like values without modifying the source. Reuses Demo 01's input; a negative amount is conservatively flagged, not treated as a confirmed error.
Deliverable: JSON profiling report. Checks: 3 automated tests. Self-initiated demo, not client work.
Read-only inspection of a fictional Excel workbook. Identifies stored formula errors, suspicious references, sheet structure and external-link warnings without changing the file or executing macros.
Deliverable shown: JSON diagnostic report. Checks: 9 automated tests. This is not an automatic formula repair or recalculation tool.
Fictional service-rate table, quantities, discount and illustrative VAT. A standalone Python Decimal calculator independently checks quote totals and rejects unknown rate codes, negative quantities and invalid tax percentages.
Deliverables shown: an editable two-sheet XLSX workbook with lookup and total formulas, plus an independent synthetic JSON quote breakdown. Checks: 7 calculator tests plus workbook structural checks. Fictional examples only; not a client quote or tax advice. Excel/compatible spreadsheet software recalculates formulas when opened.
A fictional four-row, four-column CSV with whitespace, an exact duplicate, a missing field and formula-like text. This bounded tool handles arbitrary named columns, reports issues without exposing cell values in the audit, and keeps duplicates unless removal is explicitly authorised.
Deliverables: synthetic input CSV, cleaned CSV and issue-only JSON audit. Checks: 14 dedicated tests, including 5,000-row boundary and spreadsheet formula neutralisation. This is a demonstration, not client work.
All records are fictional. Download CSV outputs and JSON quality reports to review our actual demonstration files.
Cleanup: CSV · JSON audit
Generic CSV cleanup (duplicates retained): Input CSV · Output CSV · Issue audit
Schema mapping: CSV · JSON audit
Inventory reconciliation: CSV · JSON audit
Read-only CSV profiling: JSON report
Read-only XLSX preflight: JSON diagnostic example
Quotation logic: Editable Excel workbook (synthetic) · JSON quote breakdown (fictional)
Illustrative values only, based on synthetic input; not a client result.
| Issue | Handling |
|---|---|
| Exact duplicate | Report and remove only in the explicitly scoped cleanup demo |
| Unrecognised date | Flag for manual review |
| Unknown SKU | Keep record and flag missing catalogue match |
| Spreadsheet-formula-like text | Neutralise on export in the reconciliation demo |
Before quoting, we agree on the file structure, row count, required changes, acceptance checks, deadline and handling of sensitive data. No passwords, payment-card details or regulated datasets should be sent by email. AI-assisted preparation is checked against documented rules and test cases.