# Overscore starter datasets These are real historical observations, not AI-generated data or live feeds. Downloaded and reviewed September 15, 2026. Version 1. No endorsement by the original researchers or repositories is implied. ## Retail sales — 1,716 daily country summaries Original: Chen, D. (2015). Online Retail. UCI Machine Learning Repository. https://doi.org/10.24432/C5BW33 Source: https://archive.ics.uci.edu/dataset/352/online+retail License: Creative Commons Attribution 4.0 International, https://creativecommons.org/licenses/by/4.0/ Original download: https://archive.ics.uci.edu/static/public/352/online+retail.zip File: retail-sales-v1.csv Overscore aggregated all 541,909 source transaction lines into one row per invoice date and country. Coverage: December 1, 2010–December 9, 2011; 38 countries. This is a UK online retailer selling gift items, with many wholesale customers. No product or customer-level detail remains in this prepared file. Dates have not been shifted to appear current. December 2011 is a partial month. - date: original invoice calendar date, timezone not supplied by source. - country: original customer country. - transaction_lines: number of original rows, including cancellations and adjustments. - invoices: distinct invoice numbers within this date/country, not customers. Do not assume this is globally distinct when aggregating across groups. - gross_sales_gbp: sum of quantity × unit price when both are positive. - returns_gbp: absolute value of quantity × unit price when quantity is negative and unit price positive. This is a preparation convention, not a verified accounting classification. - other_adjustments_gbp: all remaining line values, including negative unit-price adjustments. Zero-price lines remain represented in transaction_lines. - net_value_gbp = gross_sales_gbp - returns_gbp + other_adjustments_gbp. It is the source line-value total, not profit or a verified revenue-recognition measure. - units_sold/units_returned: positive/absolute negative quantities matching the gross-sales/returns conditions above. - cancelled_lines: original rows whose invoice number starts with C, not distinct cancelled invoices. No rows were discarded or deduplicated. Negative quantities and unusual values are retained through the aggregates. Net value reconciles to the original file: GBP 9,747,747.934. Original precision is retained rather than silently rounding each transaction. ## City bike rentals — 17,379 recorded hours Original: Fanaee-T, H. (2013). Bike Sharing. UCI Machine Learning Repository. https://doi.org/10.24432/C5W894 Source: https://archive.ics.uci.edu/dataset/275/bike+sharing+dataset License: Creative Commons Attribution 4.0 International, https://creativecommons.org/licenses/by/4.0/ Original download: https://archive.ics.uci.edu/static/public/275/bike+sharing+dataset.zip File: bike-rentals-v1.csv Capital Bikeshare observations for 2011–2012. Overscore retained every hour.csv row, renamed fields, translated weather codes into labels, converted normalized humidity to percent, and omitted redundant index/season/year/month/weekday codes. Some calendar hours are absent. Missing hours are not zero rentals. The source webpage lists 17,389 instances, but the actual hourly file contains 17,379 rows. - date/hour: original recorded day and hour, without a timezone conversion. - working_day/holiday: source 0/1 flags. - weather: source category labels shortened for display. - temperature_index, feels_like_index, windspeed_index: original normalized indices. These are not Celsius or windspeed units. - humidity_percent: original hum × 100. - casual_rentals/member_rentals: source casual/registered rental counts. - total_rentals: sum of those two counts; total across file is 3,292,679. These are rental counts, not unique riders or station-level trips. Observed weather relationships do not establish causation. ## Palmer penguins — 344 observations Original measurements: Kristen Gorman and the Palmer Station Long Term Ecological Research program. Gorman KB, Williams TD, Fraser WR (2014), PLOS ONE 9(3):e90081, https://doi.org/10.1371/journal.pone.0090081 Simplified data package: Horst AM, Hill AP, Gorman KB (2020), palmerpenguins. https://allisonhorst.github.io/palmerpenguins/ License: CC0 1.0 Universal, https://creativecommons.org/publicdomain/zero/1.0/ Source download: https://raw.githubusercontent.com/allisonhorst/palmerpenguins/main/inst/extdata/penguins.csv File: palmer-penguins-v1.csv Overscore retained all 344 rows and the original eight columns. NA markers were converted to empty CSV cells so absent values stay null; 19 cells are missing. No values were imputed. - species: Adelie, Chinstrap, or Gentoo. - island: Biscoe, Dream, or Torgersen. - bill_length_mm, bill_depth_mm, flipper_length_mm: measured lengths in millimeters. - body_mass_g: measured mass in grams. - sex: recorded sex, sometimes missing. - year: 2007–2009. This is a research sample, not a census of penguin populations. Count observed penguins and summarize measurements without treating missing values as zero. ## Reproducibility and reuse Preparation code: packages/hub/scripts/prepare-starter-datasets.py in the Overscore source repository. audit.json beside these files records row counts, checksums, and reconciliation totals. Source data was not generated or augmented. Retail is aggregated; bike and penguin observations are preserved with the transformations above. CC BY 4.0 permits sharing and adaptation, including commercially, with appropriate attribution, a license link, and indication of changes. Keep these source and preparation credits when redistributing the retail or bike files. The Palmer penguin data is CC0; we still acknowledge its researchers. No provider artwork is bundled. The dataset description inside your project includes these credits and this documentation link.