customers
26,467 rows
- idtext · key
- nametext
- emailtext
- statetext
- created_attime
A hose dataset · vertical
A plant store with a funnel, not just an orders table.
Pageviews, carts, checkouts and the seven in ten that never happen. Orders, shipments, winter holds and the plants that arrive dead.
3,007,210 rows · last row 426 days ago
The receipt
- seed
- 42
- scale
- 200
- start
- 2023-06-01
- fingerprint
- e520474d3a97
The whole dataset grows from this one seed. Everyone who downloads it gets the same rows with the same ids, today and next month, so whatever you build on it keeps working after the data refreshes.
per day, last 90 days
Schema · 10 tables
products
130 rows
- idtext · key
- nametext
- categorytext
- price_usdreal
pageviews
2,744,351 rows
- idtext · key
- customer_idtext
- attime
- pathtext
- referrertext
carts
49,909 rows
- idtext · key
- customer_idtext
- created_attime
cart_items
110,342 rows
- idtext · key
- cart_idtext
- product_idtext
- quantityint
- added_attime
orders
17,238 rows
- idtext · key
- cart_idtext
- customer_idtext
- ordered_attime
- discount_codetext
- total_usdreal
order_items
37,912 rows
- idtext · key
- order_idtext
- product_idtext
- quantityint
- unit_price_usdreal
shipments
17,171 rows
- idtext · key
- order_idtext
- shipped_attime
- delivered_attime
claims
3,684 rows
- idtext · key
- order_item_idtext
- kindtext
- created_attime
discounts
6 rows
- codetext · key
- percent_offreal
- starts_attime
- ends_attime
What the data claims
- In December and January, plants shipped to cold states arrive two days later.
- About one abandoned cart in eight comes back through the recovery email.
- Seven in ten carts are abandoned at checkout.
- Spring outsells winter by more than half again. February sags.
- The top fifth of customers accounts for most of the revenue.
- Three years of daily sales, with weekends and seasons in them, that regenerate byte for byte.
Why this one
Most sample sales databases start at the order. The customer wanted something, paid for it, and the row appeared. Overwater starts three steps earlier. It is a single-brand online house-plant store: live plants, pots, soil and care accessories, sold through a site whose traffic, carts and checkouts are all in the data.
Sessions are not a table. Pageviews are, with a referrer on the entry
view, and you sessionize them with the same 30-minute rule you would
use on a real export. Carts are first-class rows, so abandonment is a
left join, and a recovered cart is an order that lands hours after its
cart with an email-referred pageview in between. Shipments have a
delivered_at that is still null for recent orders. Plants sent to
cold states in winter wait longer, because the sim knows which states
freeze even though the schema only stores a state code.
Spring sells plants and February does not. A few collectors account for a large share of revenue while gift buyers order once and leave. In the spring of the second year a paid-social spike brings in a cohort that looks great on the acquisition chart and worse on the repeat chart. Prices step up a few percent each year. The same monstera costs more in later order lines than in early ones.
The other datasets