Is there a free dataset of online furniture orders with reviews? One corpus carries the transactions

Nine relational tables, one join spine. order_id stitches the orders, items, payments and reviews tables together; customer_id carries buyer location attributes; seller_id names who fulfilled each line. Money arrives per line - price and freight_value in BRL - beside payment_type and payment_value on the settlements table, product weight and length for logistics work, and the lifecycle fields (order_status, purchase timestamp, estimated versus actual delivery dates) that make delivery performance computable rather than anecdotal. Eighteen fields are documented in the dictionary; each arrives typed and identically named on every delivery.

one row per order-item line; keys pre-joined, English category labels applied
values illustrative until your sample lands

order_id                      : <32-character hex key>
customer_id                   : <32-character hex key>
seller_id                     : <32-character hex key>
order_status                  : delivered
order_purchase_timestamp      : 2017-11-14 09:41:50
order_estimated_delivery_date : 2017-11-30
order_delivered_customer_date : 2017-11-27 18:12:04
price                         : 349.90      (BRL)
freight_value                 : 41.20       (BRL)
category (en)                 : furniture_decor
product_weight_g              : 18600
payment_type                  : credit_card
review_score                  : 5

Home furnishings isolates in one predicate: the Portuguese product_category_name joins to an English translation table carrying furniture_decor, bed_bath_table and housewares, so the furniture slice filters out of the roughly 100,000-order corpus without touching the rest of the mix. Two design facts matter more than the field count. An order may hold several items and each item may be fulfilled by a distinct seller, so revenue rollups need seller-aware aggregation at the line rather than naive sums over order headers. And the publisher anonymised real commercial data - every store and partner identity replaced with a Game of Thrones great-house name - which is why prototypes can ship without handling personal information.

Three practical notes decide whether the layer fits the job. The comments stay in Portuguese, so sentiment work either runs natively in Portuguese models or budgets a translation pass. Partner names are scrubbed, which protects anonymity but removes brand attribution from complaint text. And the survey attaches to orders rather than shoppers, so repeat-buyer satisfaction needs the customer key joined first. Handled, the layer feeds NPS-style trackers, delivery-experience diagnostics and citable order-level evidence - the difference between quoting a rounded satisfaction percentage and quoting the distribution behind it.

Microdata calibrates execution; aggregates decide how big the opportunity is. The Census API - Monthly Retail Trade Survey (MRTS) publishes monthly national sales, inventories and inventory-to-sales ratios for furniture and home furnishings stores, with history spanning multiple decades per kind of business - the denominator any conversion, basket or growth claim gets measured against. The Census API - Advance Monthly Retail Trade (MARTS) adds advance national estimates covering the most recent reference months, so a quarter-end readout does not wait on revisions.

For structure rather than rhythm, the US Economic Census 2022 - Furniture & Home Furnishings Stores reports establishments, employment, payroll and revenue for NAICS 442 retail and NAICS 337 manufacturing down to counties, metros and places, with prior vintages back to 1997 for trend work. Together the three turn a zip-code-level demand map into a sized plan: local share times national pace times store density. The head-to-head between the transaction layer and the structural layer is worked through in Olist orders vs Economic Census furniture stores.

Transactions tell you what sold in one marketplace; the retail-side records tell you what the shelf looks like elsewhere. The IKEA US Product Catalog - Furniture & Home Furnishings Taxonomy resolves 22 US departments through subcategories to tens of thousands of SKUs, each carrying structured price, package dimensions and per-store availability - assortment ground truth for taxonomy tests, price-band studies and range comparisons against the furniture_decor mix. The IKEA US Homepage & Navigation Hub captures the merchandising layer above it: promotional pricing banners, seasonal sections and rated product cards, the signals a competitor-tracking workflow diffs over time.

The corporate leg closes the loop. Home Depot Investor Relations - SEC Filings & Financials carries consolidated quarterly and annual statements with machine-readable XBRL, so big-box economics - comparable-store dynamics, inventory posture, category commentary - sit in the same stack as marketplace microdata. Historical Brazilian transactions beside current US assortment beside retailer financials: three views of one industry that rarely share a shelf, delivered as typed rows on one contract.

Four, all manageable once named. Window - coverage runs September 2016 through 2018 and never moves, so nothing after 2018 is represented; the upside is perfect reproducibility, since a training run today reruns to identical results next quarter, and the fix for recency is pairing with the moving Census aggregates rather than hoping for fresh rows. Geography and currency - Brazil nationwide at zip-code-prefix level, amounts in BRL; findings transfer as techniques and benchmarks, not as market constants for other markets. Scope - one marketplace whose model bundles small merchants under a single contract, so read it as marketplace behavior rather than e-commerce overall. Definitions - field-definition confidence is graded inferred, reconstructed from documentation rather than confirmed against captured rows, so settle edge semantics in your sample.

One provenance footnote belongs in any deck: the sixth release of the corpus removed a previously published classified variant, so pin the vintage you build on. Datadory stamps extracts at delivery, which keeps the pin auditable months later when a stakeholder asks which rows produced the chart.

API, files, or your warehouse. Daily, weekly, or hourly.

Open the products behind the answer - the Brazilian E-Commerce Public Dataset by Olist record for sample rows, the full field dictionary and coverage chips, and the Kaggle / Olist source profile for provenance. Frame it with the home furnishings data guide, which maps all fifteen pooled records, and the sibling posts on monthly furniture store sales data and IKEA product catalog JSON for the aggregate and assortment layers respectively.

For positioning across the whole slice, the home furnishings data hub holds the scorecard with quality grades side by side, best home furnishings datasets ranks all ten collections head to head, and the vocabulary underneath lives in the glossary entries for Olist product category codes and the retail transaction log. When descriptions stop and real rows need to start, request a sample scoped to your categories and window - the field dictionary travels with it.