Skip to content
Uzair Azhar
Menu

Data and licences

The demos on this site run on public data. This page lists every dataset, where it comes from, its licence, and exactly what was changed.

Rules every demo follows

  • Real data is shown with its real dates and values. Nothing is shifted to look recent.
  • Simulated faults and synthetic data are always labelled as such where they appear.
  • People are never identifiable: customer IDs stay anonymous, and faces and licence plates in video are blurred.
  • Organisations are never named next to an automated judgement; stable codes are used instead.
  • Figures on the site come from recorded runs on this server, not from estimates.

Online Retail II

Real data

UCI Machine Learning Repository, DOI 10.24432/C5CG6D

What it is
All transactions of a UK-based, registered, non-store online retailer of gift items between 1 December 2009 and 9 December 2011: 1,067,371 invoice lines with invoice number (prefix C = cancellation), stock code, description, quantity, timestamp, unit price in GBP, anonymised customer ID and country. The company and its customers are not named in the dataset.
Licence and attribution

CC BY 4.0. Chen, D. (2012). Online Retail II [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C5CG6D. Licensed under CC BY 4.0.

Licence as stated on the UCI dataset page. An earlier redistribution (the onlineretail2 R package, 2020) quotes older donor terms restricting use to non-commercial purposes; this site is a non-commercial portfolio demonstration.

What was changed
  • Split into 25 monthly CSV "drops" and a derived product catalogue and customer register so the pipeline demo has realistic sources. Values are not altered.
  • Failure scenarios (corrupted drop, source outage) apply simulated faults to a copy at run time and are labelled as simulated.
How downloads are verified

The data is not stored in the repository. It is downloaded when the server image is built and accepted only if it has exactly 1,067,371 rows, the expected 8 columns, timestamps from 2009-12-01 07:45:00 to 2011-12-09 12:50:00, and a content fingerprint matching the one recorded here.

fingerprint sha256 56dbc9b75d5834fc0a3a451e126601b14cb4cbc3fed774161f864eacddf18e54

Sources
  • Primary: UCI Machine Learning Repository. Excel workbook with two yearly sheets that overlap on 1-9 December 2010.
  • Fallback: pinned mirror. Unmodified redistribution in the onlineretail2 R package (pinned commit). Used when the canonical host is unreachable; accepted only if the content checks pass.
Used by
Data Pipeline Observatory

New datasets are added here when the project that uses them goes live. The record is generated from the repository's artifact manifest, the same file the download verification reads. See also the privacy notice.