Industries / Consumer goods / Furniture / Mattresses
Selling
Selling gets mattresses from the factory to sleepers through furniture and department stores, chains that sell only beds, makers' own shops and websites, and online marketplaces, with trials, warranties, advertising and reviews doing the persuading, because a buyer cannot judge a mattress by looking at it.
7
Milestones achieved, from 1931 to 2026.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1931
Regional makers share one brand
Thirteen independent US mattress makers joined in 1931 as Sleeper, Inc., launched their first shared brand-name mattress, the Perfect Sleeper, in 1932, and later became Serta. By 1999 Serta, Inc. was "owned by and operated for the benefit of its licensees": ten US makers covered 34 territories, none could make Serta beds outside its own, and the board voted the national advertising budget. One licensee, Sleepmaster, paid about 3% of its gross sales as a royalty, its SEC filing says.
LessonA shared brand let regional factories pay for national advertising together while each kept its own territory.
Credit
the 13 founding makers; Serta, Inc
2016
A court rules on reviews
In April 2016 Casper, an online mattress seller, sued three review sites, saying they ranked high in searches for "mattress reviews" without clearly disclosing that they were paid affiliate commissions, a fee for each sale a site refers, by rival brands. In October 2016 a federal court in New York held that the reviews of one of them, Derek Hales of Sleepopolis, were opinions, and that a commission did not make them false advertising under federal law: "The fact that Hales receives a kickback does not make his thoughts any less his own." It let a claim under New York consumer law go forward. All three sites settled, and Casper then lent money to JAKK Media, which used it to buy Sleepopolis, Gizmodo reported in 2017.
LessonUS law asks review sites to disclose that sellers pay them, not to refuse the money.
Credit
Casper Sleep; Derek Hales; US District Court for the Southern District of New York
2018
Too many stores, too close
Mattress Firm, a Houston mattress chain owned since 2016 by Steinhoff International, filed for Chapter 11 on 5 October 2018. Chapter 11 is US court-supervised reorganization, in which a company keeps trading while it cuts its debts. Mattress Firm said it would leave up to 700 stores in markets where it had "too many locations in close proximity to each other", and it came out of court on 11 November 2018 with about 2,600 stores, close to 900 fewer than before, BedTimes reported. Its chief executive, Steve Stagner, said its "unprecedented growth had led to duplicative store locations in many of our markets".
LessonStores bought or opened close together end up competing for the same buyers.
Credit
a failure; Mattress Firm
2022
An online seller goes private
Casper, founded in New York in 2014, sold mattresses online with a 100-night trial, and by the end of 2019 also had 60 of its own stores and 18 retail partners, including Amazon, Costco and Target. Its prospectus showed net losses of $73.4 million in 2017 and $92.1 million in 2018, a year in which $80.7 million of refunds, returns and discounts came off its sales on the way to $357.9 million of net revenue. It listed on the New York Stock Exchange at $12 a share in February 2020, and in January 2022 the investment firm Durational Capital Management bought it for $6.90 a share and took it off the exchange.
LessonCasper lost money in every year its prospectus reported, and fast growth with free trials did not change that.
Credit
Casper Sleep; Durational Capital Management
2025
A maker buys its retailer
In January 2017 Tempur Sealy ended its supply contracts with Mattress Firm, which had brought in about 21% of its 2016 sales, and its shares fell 22.9%; it returned to about 2,500 Mattress Firm stores under a 2019 agreement. It then agreed to buy Mattress Firm, and on 2 July 2024 the US Federal Trade Commission (FTC), which called Tempur Sealy the world's largest mattress maker and Mattress Firm the largest US mattress retailer, sued to stop the deal, saying the combined company could cut rivals' floor space and steer shoppers to its own beds. A federal court in Texas denied the FTC's request on 31 January 2025, the roughly $5 billion deal closed on 5 February 2025, and the company became Somnigroup International on 18 February 2025.
LessonBuying its largest retailer ended a maker's dependence on that retailer and gave it a say over how rival brands reach buyers.
Credit
Tempur Sealy (Somnigroup International); Mattress Firm; US Federal Trade Commission
2026
Courts test sale prices
Online mattress sellers often show a struck-through "was" price, a claimed saving and a countdown timer. On 24 April 2026 Australia's Federal Court ordered two Emma Sleep companies to pay A$15 million in penalties, after a case by the ACCC, the competition regulator: between June 2020 and March 2023 all 74 of their products had been advertised with discounts, 58 of them had not been sold at the struck-through price before, and the court found the conduct came from "a deliberate marketing strategy". In the UK, after a case by the Competition and Markets Authority (CMA), the High Court confirmed on 22 May 2026 an order in which Emma admitted that its countdown timers and "high demand" messages broke consumer law. On 30 July 2026 the same court declined the CMA's request to require that, for every two mattresses sold at the lower price, one had been sold at the "was" price. Britain's advertising regulator, the ASA, also upheld complaints about the savings claims of four online mattress sellers between July 2025 and August 2026, all brought by a rival seller, Simba Sleep.
LessonA struck-through price must be one the seller has actually charged, but the UK court would not turn that rule into a fixed sales ratio.
Credit
Federal Court of Australia and the ACCC; High Court of England and Wales and the CMA; Advertising Standards Authority
2026
An air-bed chain is sold
Sleep Number, founded in Minneapolis in 1987 as Select Comfort, makes beds whose firmness is set by pumping air into chambers, and in 2013 made over 96% of its sales through its own 440 stores, website and phone line. In 2025 its sales fell 16% to $1.41 billion, which it put down to lower store traffic; it lost $132 million and ended the year with 600 stores. It filed for Chapter 11 in New York on 12 June 2026, with Sleep Country Canada making the first bid, which sets the floor for an auction, at $415 million; the court approved the sale on 20 July and it closed on 31 July 2026 at a cash price of $529.5 million.
LessonA seller that sells only through its own stores feels a drop in store visits across the whole business at once.
Credit
Sleep Number; Sleep Country Canada
Watch for
Savings from two purchases
Somnigroup says it expects $70 million of future cost savings from combining with Mattress Firm, and $75 million a year from Leggett & Platt, the spring and component maker it bought on 26 August 2026, up from its first estimate of $50 million.
Achieved whenaudited annual results, or an independent analyst, show combined costs fell by these amounts.
Claimed by
Somnigroup International, August and November 2026
Second-largest sleep retailer
Sleep Country Canada, owned by Fairfax Financial Holdings since October 2024, says that buying Sleep Number makes it the "world's second-largest sleep retailer", with more than 800 stores: its own 307 in Canada and more than 570 Sleep Number stores in the US.
Achieved whenan independent ranking of retailers by mattress sales places the combined company second in the world.
Claimed by
Sleep Country Canada, July 2026
An AI shopping guide
GoodBed, a US mattress-matching site, launched GoodBot, an AI assistant that answers shoppers' follow-up questions using only GoodBed's own data on more than 1,000 evaluated mattresses. GoodBed says this gives "more accurate and dependable responses" than general AI tools, and published no test results.
Achieved whenan independent test scores GoodBot against a general-purpose AI assistant on a fixed set of shopper questions and publishes the results.
Claimed by
GoodBed, June 2026
Open
01
Renamed mattresses, mapped
A public dataset for one country outside the US, such as Australia or the UK, that groups mattresses sold under different names at different retailers into likely "same mattress" groups by matching the specifications on product pages and spec sheets (layers, spring count, height and materials), with each match graded and its evidence shown.
The briefCredit · unclaimed02
Trial fine print, coded
An open, dated dataset of the binding terms behind the trials and warranties of 100 mattress sellers in the US, UK and Australia: trial length, nights before a return is accepted, pickup or restocking fees, exchange-only rules, conditions on the buyer, and whether the warranty is prorated (the buyer pays a growing share of a replacement as the mattress ages). It flags wherever the advertised offer leaves out a condition found in the terms, and takes a new copy every quarter to record changes. No buying or returning is involved.
The briefCredit · unclaimed
Open 01 · the brief
Renamed mattresses, mapped
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public dataset for one country outside the US, such as Australia or the UK, that groups mattresses sold under different names at different retailers into likely "same mattress" groups by matching the specifications on product pages and spec sheets (layers, spring count, height and materials), with each match graded and its evidence shown.
Why it is still open
CHOICE, the Australian consumer group, found in 2017 that one Sealy model, the same inside, was sold as the "Posturepedic Swan" at Snooze, the "Posturepedic Aspire Vogue Deluxe" at Harvey Norman and the "Aberdeen Firm Plush" at David Jones. Which? noted a single UK case in March 2026, a Silentnight model sold direct under one name and by other retailers under another. In the US, GoodBed lists "similar models" for thousands of mattresses but does not say how it matches them or release the data. Searches in November 2026 found retailers' advice to compare spec cards, but no open dataset for Australia or the UK.
Built with what exists
Python and Playwright to collect product pages and PDF spec sheets; an AI language model to copy each specification into a fixed table, which language models can now do cheaply from messy spec text; Gemini Decision 1 (Notes: "A thousand questions, one pass") to answer fixed questions about each candidate pair, such as whether the two have the same spring count; a shared spreadsheet for human review.
First step
Collect one maker's range, such as Sealy in Australia or Silentnight in the UK, from three retailers, and match it by hand from the listed specifications.
Done looks like
1,000 product pages from at least 8 retailers in one country sorted into groups; two people working separately agree with the program on at least 95% of a 200-pair sample; at least 20 matches confirmed in writing by the maker or retailer; data, code and agreement rates published.
Open 02 · the brief
Trial fine print, coded
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, dated dataset of the binding terms behind the trials and warranties of 100 mattress sellers in the US, UK and Australia: trial length, nights before a return is accepted, pickup or restocking fees, exchange-only rules, conditions on the buyer, and whether the warranty is prorated (the buyer pays a growing share of a replacement as the mattress ages). It flags wherever the advertised offer leaves out a condition found in the terms, and takes a new copy every quarter to record changes. No buying or returning is involved.
Why it is still open
The conditions sit in terms pages, not in the headline offer. Casper advertises a 100-night trial, but its terms say returns are not accepted in the first 30 nights, and that mattresses left unopened "too long" may no longer qualify, without saying how long. Searches in November 2026 found review sites that list trial lengths and fees brand by brand, with figures that conflict, but no dataset that codes the full conditions across sellers or tracks how they change.
Built with what exists
Python and Playwright to save terms pages; the Internet Archive's Save Page Now to keep a public dated copy; an AI language model to read each page into 15 fixed fields; a spreadsheet for checks.
First step
Save the trial and warranty pages of 10 large sellers in one country and code the 15 fields by hand.
Done looks like
100 sellers in 3 countries coded on 15 fields; two people agree with the program on at least 95% of fields for a 30-seller sample; four quarterly copies, with every change logged; everything published.