Playground

That dataset is part of Pro. See plans or choose a free dataset.

• Order-line export with customer lookup

  A shop's order lines exported from an order system, with the customer master table used for the region lookup.

• Account updates from several systems

  The same accounts sent by several systems with different update times and priorities.

• Names in mixed Unicode forms

  Names copied from several sources: composed and decomposed accents, no-break and ideographic spaces.

• Orders and customer segments

  Orders with padded or missing customer keys, and a segment table that includes a row with a missing key.

• Monthly meter readings

  Long-format readings per entity and month, where some months were never reported and some report zero.

• Payments and account master

  Payments that reference account numbers, some missing or unknown.

• Versioned order lines

  Order lines re-sent with increasing versions; multi-line orders are common.

• Sales and refunds

  Transactions including very large sales and negative refunds.

• Event dates in several locales

  Event dates written as ISO dates or slash dates, with a per-row locale that is sometimes missing.

• Payments in several currencies and locales

  Amounts written with US or German separators, sometimes in parentheses for negatives, in several currencies.

• Monthly export files

  Three monthly files: one without the optional channel column, one with it, and one with a broken schema.

• Training and test values with gaps

  A numeric feature split into training and test sets, with missing values in both.

• Event timestamps with and without offsets

  Timestamps from devices, some with offsets, some without, some malformed.

• Contact emails

  Emails with padding and mixed case, including dots and plus tags that must be kept.

• Support tickets by channel

  Tickets where agents typed the channel freely: case, spacing and spelling variants.

• Supplier inventory feed

  A supplier feed with impossible quantities, zero prices, malformed numbers and unknown countries.

• CRM and web shop customers

  Customer records from the CRM and the web shop with different ids, email variants and name spellings.

• Members, postcodes and opt-in flags

  Member records with leading-zero identifiers and opt-in flags written many ways.

• Survey answers (wide)

  One row per respondent and one column per question; unanswered questions are missing.

• Site visits by region

  Visits with blank or missing regions and durations that are sometimes not recorded.

• Sensor readings with sentinel codes

  Readings from three sensors where -999 means no reading was taken.

• Suppliers with hand-typed country codes

  Supplier master data where country codes were typed by hand in mixed case and spacing.

• Support tickets with free-typed priorities

  Tickets where agents typed priority in several ways, or left it empty.

• Orders from several regions

  An order export covering EU, US and APAC, where only EU orders belong in the report.

• Order lines and product catalog

  Order lines with hand-typed SKUs, and the product catalog they should match.

• Product price history

  A price list where each product has several dated price rows, some with broken prices.

• Sign-ups with mixed date quality

  Sign-up records exported from a form, with missing, impossible and pre-launch dates.

• Invoice totals with formatting problems

  Invoices from a billing export with missing, malformed and negative totals.

• Visits with partly recorded durations

  Visit records where some durations weren't recorded and some were typed as words.

• Orders with bulk quantities

  Orders including very large quantities and some invalid ones.

• Employees and department codes

  Employees with hand-typed department codes and the department list.

• Expense claims by category

  Expense claims with hand-typed categories and some malformed amounts.

• Quarterly store sales (wide)

  One row per store with a column per quarter; some quarters weren't reported and some are zero.

• Users mixed with test accounts

  A user export where QA test accounts are marked with a flag written in several ways.

• Event deliveries with resends

  Device events, some delivered more than once with timestamps in different offsets.

• US addresses with mixed state formats

  Addresses where the state is a code or a full name in mixed case.

• Payments with currency codes

  Payments where currency codes and amounts were typed by hand.

• Contacts with blank fields

  Contact records where some names and emails are empty or whitespace only.

• Delivery days by carrier

  Delivery durations with a long tail of late parcels and some malformed values.

• Customer phones with placeholders

  Customer records where empty phone fields were filled with placeholders.

• App event log

  Raw app events, several per user per day, with hand-typed user IDs and a few broken dates.

• Customer order dates

  Order dates per customer, typed in mixed case, with some missing and impossible dates.

• Store unit sales by month

  Unit sales rows per store and month, with repeated store-month rows and months beyond Q1.

• Items with retired category codes

  Catalog items still tagged with retired category codes, plus the approved crosswalk to the new categories.

• Hotel bookings and room states

  Bookings typed by front-desk staff, and the room list with each room's current state.

• Payments with export copies

  A payment export where a glitch repeated some rows, mixed with genuine repeat payments.

• Warehouse shipments

  Outbound shipments with hand-entered weights and tracking numbers, some still pending.

• Vendor customer file

  A customer extract from a vendor system with display-style headers and zero-padded IDs.

• Marketing consent answers

  Consent answers collected over several campaigns, where some people were never asked.

• Orders around Q1 2026

  An order extract spanning late 2025 to mid 2026, of which only Q1 2026 belongs in the report.

• Product tags pasted from the web

  Product colour tags copied from a supplier website, carrying invisible Unicode spaces.

• Fuel price quotes

  Per-litre fuel prices from three stations, quoted to four decimals with some over-precise or broken values.

• CRM records with possible duplicates

  CRM contact records where the same person was entered more than once with varying email case.

• Sensor temperature readings

  Readings from four temperature sensors with a -40 to 85 °C range, including hot days and glitches.

• Nightly order batch

  A nightly order batch that overlaps with orders an earlier run already loaded.

• Events with long account IDs

  Product events keyed by 16- to 19-digit account IDs, some beyond floating-point precision.

• Product listings

  Marketplace listings where price is mandatory and weight is optional and often messy.

• Duplicated contacts

  Contacts entered several times, where newer copies sometimes lost the phone number.

• Store files for one load

  Three store extracts for the same load, one of which arrived with the wrong columns.

• Orders and customer statuses

  Orders from customers with different account statuses, plus a few orders for customers that don't exist.

• Customer contact completeness

  A customer extract with many blank, whitespace-only and missing contact fields.

• Ticket statuses as typed

  Support tickets whose status field was free text for years.

• Sales, products and categories

  Sales linked to a product list, which is linked to a category list that lost some codes.

• Overlapping customer exports

  This week's customer export and an older archive that overlaps with it.

• Orders with optional discounts

  Order data for a pricing model, where the discount field is often left empty.

• Customer spend for a partner

  A customer table with personal details, of which a partner may receive only city-level spend.

• Device last-seen dates

  A device inventory extracted on 30 June 2026, with clock errors and devices that went quiet.

• Discount codes and rates

  Discount codes whose rates were sometimes typed as percents instead of fractions.

• Employment contract end dates

  Contracts where N/A marks open-ended contracts and blanks mark dates not yet entered.

• Units sold by product

  Units sold per product across a few categories, with typed category names and some broken counts.

• Invoice numbers by store

  Invoice numbers typed as text in three stores, some zero-padded and some broken.

• Coupon redemptions

  Redemptions of case-sensitive coupon codes, where some codes differ only by case.

• Product attributes as key-value rows

  A product attribute log stored one attribute per row, with repeated updates and extra attributes.

• Contacts and a suppression list

  A contact list and an opt-out suppression list that contains blank rows.

• Postings around a period close

  Ledger postings from offices in several time zones around the March close.

• Club members with common names

  A membership list where common names repeat and some members were entered twice.

• Contact preference checkboxes

  A preference form that stored each channel checkbox in its own column.

• Visits with accented city names

  Visit logs from two systems that encode accented letters differently.

• Ages and an age-band table

  Self-reported ages to be grouped with the approved age-band table.

• Daily sensor sheet

  A spreadsheet-style sensor log with one column per day.

• Pasted-together sales exports

  Sales exports pasted into one sheet, with repeated header rows and a TOTAL row.

• Ticket sources and channel map

  Ticket sources typed by agents and the approved mapping to reporting channels.

• Payments with currency symbols

  Payments where currencies were captured as symbols or codes, including the ambiguous $.

• Game scores

  Point entries per player from a tournament, with typed player names and some invalid entries.

• Tickets by team and status

  Support tickets per team with free-typed statuses, for a team-by-status count table.

• Orders and product names

  Orders that carry the customer's name, and a product list whose name column means something else.

• Residents by district

  A synthetic resident list to be published as counts per district with small-cell suppression.

• Account status history

  Account status changes where some rows lost their timestamp in an export.

• Customer loyalty tiers

  Customer tiers typed by account managers, including an abbreviation for platinum.

• Order values by channel

  Orders from three sales channels with a few missing or broken amounts.

• Customer file saved with a BOM

  A customer CSV saved as UTF-8 with a byte-order mark before the first header.

• Member birth dates with placeholders

  A membership export where unknown birth dates were stored as 1900-01-01 or 0001-01-01.

• Session events across time zones

  Web session events logged by servers in different time zones.

• Auction bids

  Bids on auction lots placed from several time zones, including equal bids.

• Branch sales files without IDs

  Daily sales files from two branches with no row IDs, where identical rows are separate sales.

• Agreement survey

  A three-question agreement survey stored one column per question.

• Orders with missing statuses

  An order export where cancelled orders must be excluded and some statuses are missing.

• Order status history

  Status changes per order, where some orders were opened and later closed.

• Payment feed with resends

  A payment feed delivered at least once, so some payments arrive two or three times.

• Order event log

  Fulfilment events per order with repeated scans and events from several time zones.

• Coded satisfaction answers

  A coded survey question with special codes for don't know and refused, plus its codebook.

• Support ticket queue

  Open support tickets, some unassigned, created from offices in several time zones.

• Product ratings

  Star ratings for kitchen appliances, some products with only a few reviews.

• Order line statuses

  Fulfilment status per order line, to be rolled up to order-level flags.

• Contacts with three phone slots

  A contact form with three phone fields, often filled with the same number twice.

• Monthly store revenue reports

  Q1 revenue reports per store, where some stores haven't sent March yet.

• Orders for a dashboard tile

  A day's orders feeding three headline numbers on a dashboard.

• Invoices and CRM billing emails

  Invoices to be emailed, and a CRM extract where some customers lack a billing email.

• Stock movements

  Stock movements where negative quantities record returns.

• Sales and a fiscal calendar

  Monthly sales to be reported by a fiscal year that starts in April, with the finance calendar table.

• Mixed two- and three-letter country codes

  Customer countries captured as ISO alpha-2 or alpha-3 codes, plus a reference table.

• Marketing touches

  Marketing touches per customer logged by tools in different time zones.

• Churn training data

  Customer rows for a churn model, some with unknown outcomes and messy spend values.

• Daily values for a forecast

  Daily values spanning 2025 and 2026, to be split by time for a forecasting model.

• Dairy stock batches

  Stock batches with expiry dates, for a sellable-stock report as of 30 June 2026.

• Sales and store regions

  Sales by store code and the store table that assigns each store to a region.

• Sign-ups with birth dates

  Sign-ups that must be checked against an adults-only rule on 30 June 2026.

• Expenses by department

  Expense lines where some were never assigned to a department.

• Units with typed month names

  Q1 unit sales where the month was typed in many ways.

• Catalog price versions

  A price history including versions scheduled to start after the reporting date.

• Order export with Windows line endings

  An order file exported on Windows; the last column carries stray carriage returns.

• Pasted part codes

  Part codes typed or pasted from documents, some with typographic dashes, and the parts catalog.

• Orders and a customer crosswalk

  Orders keyed by shop or CRM customer keys, and the crosswalk that maps each key to a master customer.

• Incremental change extract

  Rows changed around the previous run's high-water mark, stamped in several time zones.

• Member numbers from a spreadsheet

  Member numbers and points that passed through a spreadsheet, with float-formatting artefacts.

• Regional files with a renamed column

  Three regional exports; the oldest one still uses the old amount column name.

• Customer phone columns

  Customers with up to three phone numbers in separate columns.

• Repeat sign-ups

  People who signed up more than once, with details filled in on some rows and not others.

• Daily orders and a marketing calendar

  Daily order counts for November and the marketing calendar of known events.

• Web form checkboxes

  Sign-ups from a web form that exports checkboxes as on or nothing.

• Garbled city names

  Customer cities from a pipeline that once read UTF-8 as Windows-1252, and the list of accepted cities.

• Staff directory names

  Staff names with mixed case and accents in two Unicode forms.

• People and their addresses

  People with addresses typed in inconsistent case and padding.

• Accounts with soft deletes

  An accounts table that keeps deleted accounts, marked with a deletion timestamp.

• Lab results with detection limits

  Lab results where values below the detection limit are written <LOD.

• Amounts from two offices

  Amount files from a German and a US office, each using its local number format.

• Orders for a repeat-customer list

  Orders with mixed statuses, to find customers with at least two paid orders.

• Household survey counts

  Survey answers with a required adult count and an optional children count.

• Self-reported timesheets

  Task times typed by staff from memory, heaping on round numbers.

• Invoice cost allocations

  Invoice costs split across departments by percentage.

• Blog post tags

  Tags on blog posts, typed on phones and pasted from web pages.

• Records and CRM emails

  Records from three systems to be matched to CRM customers by email.

• Customer tier history

  A customer tier history table with closed versions and current rows, including a duplicate open row.

• Country order files

  Order files from three countries, each in its own currency without a currency column.

• Exports with different column orders

  Two exports with the same columns in a different order.

• User active flags

  A user export with the active flag stored as text in mixed case.

• Ledger entries with directions

  Ledger lines where the direction column says debit or credit.

• Customer contracts

  Contracts with start dates and optional end dates.

• Apparel stock by size

  Stock counts per SKU and size, with sizes spelled several ways, and the size order table.

• Event dates by source

  Events from three systems, one of which sends US-style slash dates.

• Freezer temperature log

  Temperature readings from freezer sensors, including many readings of exactly zero.

• Audit log across time zones

  Audit events stamped by servers in several time zones.

• International order dates

  Orders from shops in several countries, with dates typed in each country's convention.

• Weekly extract dates

  Rows from three weekly extracts, to be profiled before they are combined.

• Employee hierarchy

  Employees with a reference to their manager in the same table.

• Flights and airports

  Flights with origin and destination airport codes, and the airport table.

• Monthly refunds

  Refunds around March 2026; in some datasets March has no refunds at all.

• Payment amounts near the column limit

  Payment amounts, some too large for the warehouse's decimal(10,2) column.

• Visits and conversion outcomes

  Website visits by marketing channel, with a yes/no conversion outcome.

• Lead acquisition channels

  Leads with the channel they came from, to be encoded for a model.

• Cities in scoring data

  New rows to score with a city feature, and the list of cities seen in training.

• Order history for features

  Customer orders spanning 2025 and 2026, for features as of 2026-01-01.

• Contact handoff

  Contacts with optional fields that are often empty or whitespace-only.

• Store-by-month sheet

  A wide sheet with one column per month and gaps where stores didn't report.

• Order values for KPIs

  Order amounts for a KPI panel that reads metrics as rows.

• Order lines with full-width digits

  Order lines typed on mixed keyboards; some quantities use full-width digits.

• Contacts with optional phones

  Contacts duplicated by email and phone, where many rows have no phone.

• Daily stock snapshots

  Two consecutive full stock snapshots from a warehouse.

• Pipeline load log

  A log of successful loads per source, stamped in different time zones.

• Late-arriving events

  Events with the time they happened and the (sometimes much later) time they were loaded.

• Orders and product status

  Historical order lines and the current product catalog, which marks some products as discontinued.

• Key-value change feed

  A change feed of inserts, updates and deletes with sequence numbers.

• Timesheet submissions

  Timesheet entries with resubmissions, and the staff list that assigns each employee to a team.

• Survey response rates

  Survey answers in wide form with many unanswered questions.

• Quantities with free-text units

  Ingredient lines with quantities and units typed in many ways.

• Form resubmissions

  Form submissions where people sometimes submit the same form again from another time zone.

• Shelf prices

  Shelf prices typed by store staff, some apparently entered in cents.

• Ticket channels and a dictionary

  Support tickets with free-text channels, and the dictionary of spellings already mapped.

• Unordered inventory export

  An inventory export in arbitrary row order, with some blank aisle cells.

• App releases

  Release records with version parts in separate columns.