Industries / Food, farming and forestry / Agriculture and farming / Cotton ginning
Tracking bales
Giving each module and bale a number that follows it from field to gin, warehouse and mill, and proving where the fiber grew.
7
Milestones achieved, from 1994 to 2024.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1994
Receipts go electronic
A warehouse receipt proves who owns a stored bale; banks lend against it and buyers pay for it. In 1989 a record day of 385,599 bales on TELCOT, the electronic market of the Plains Cotton Cooperative Association (PCCA) in Lubbock, Texas, left paper receipts backed up for days, and PCCA piloted electronic receipts for the 1990 crop at a gin in Dodson, Texas. A USDA rule effective 2 May 1994 said electronic cotton receipts have "the same legal effect as" paper ones, held by approved private providers.
Credit
PCCA; USDA
1997
Twelve digits per bale
Since 1997 every US bale's Permanent Bale Identification (PBI) tag has carried a 12-digit number, five digits for the gin (assigned by USDA) and seven for the bale, printed as text and as a Code 128 barcode. The same number goes on the samples cut for USDA classing, so each bale's measured fiber quality stays linked to it until a mill opens it. A gin may not repeat a number within five years, and USDA makes a gin fix duplicates before it will class the gin's cotton.
Credit
the US cotton industry through the National Cotton Council; USDA Agricultural Marketing Service
2004
Brazil tags every bale
The Brazilian Cotton Growers Association (Abrapa) created the Abrapa Identification System (SAI) in 2004: numbered barcode labels, issued only to registered gins and fixed to each bale as it is made, with tear-off sections for the classing samples. In the 2013/14 season, 275 gins, 90% of those active, printed 9,534,263 labels, which the farm magazine Revista Cultivar describes as practically all of Brazil's crop.
Credit
Abrapa
2012
Modules carry radio tags
A module is a block of seed cotton, fiber still on the seed, made in the field to wait for the gin. John Deere's 7760 picker builds round modules wrapped in plastic film, and in 2012 Deere introduced Harvest Identification, which records client, farm, field and harvest time against each module's serial number. Each wrap carries four passive RFID (radio-frequency identification) tags. USDA Agricultural Research Service engineers added scanners at the gin's truck scale and module feeder, and in 2019, at Tanner and Co. Gin in Frogmore, Louisiana, linked each lint bale's PBI back to the module it came from without paper tags.
Credit
John Deere; John Wanjura and colleagues, USDA ARS, Lubbock, with Cotton Incorporated
2022
Uzbek cotton cleared
On 1 March 2022 the International Labour Organization (ILO), which had monitored Uzbekistan's harvest since 2013, reported from about 11,000 interviews that 99% of 2021 pickers worked voluntarily and that Uzbek cotton was "free from systemic child labour and systemic forced labour." The ILO estimates that two million children and half a million adults were taken out of child and forced labor over seven years of reform. On 14 March 2022 the ILO welcomed the Cotton Campaign's lifting of its boycott of Uzbek cotton.
Credit
Government of Uzbekistan; ILO third-party monitoring
2023
Customs tests for origin
The US Uyghur Forced Labor Prevention Act presumes that goods from China's Xinjiang region were made with forced labor and bars them unless the importer shows otherwise by "clear and convincing evidence"; the presumption took effect on 21 June 2022. US Customs and Border Protection (CBP) turned to isotopic testing, which compares the ratios of heavier and lighter atoms in a fiber with those of the places cotton grows. CBP documents obtained by Reuters showed 10 of 37 garments in a May 2023 batch tested "consistent" with Xinjiang cotton; CBP says testing alone is not enough to clear a detained shipment.
Lessonan origin on paper can be checked against the fiber itself.
Credit
US Congress; CBP
2024
Shipping records beat certificates
In April 2024 the investigative group Earthsight used shipment records to trace 816,000 tonnes of cotton exported from 2014 to 2023 by two growers in Bahia, Brazil, SLC Agrícola and the Horita Group, which it linked to land grabbing and forest clearing, to eight Asian factories supplying H&M and Zara. The cotton was certified by Better Cotton, which Earthsight calls the world's largest cotton certification scheme and says has been criticized for not allowing full traceability. Better Cotton said Abrapa's initial checks found no non-compliance and that it would send an independent auditor.
Lessona certificate on a farm does not tell a buyer which bales reached it.
Credit
Earthsight
Watch for
Better Cotton traces bales
Better Cotton says over 23,000 tonnes of physical BCI cotton had been traced from gins to its retailer and brand members, up from 90 tonnes a year earlier, that more than half of BCI cotton entering supply chains is now traceable, and that over 2,500 supplier sites can trade it.
Achieved whenan outside auditor or researcher samples traced products and confirms, from records or fiber tests, that they came from the gins named.
Claimed by
Better Cotton, November 2025
Tracing regenerative bales
The U.S. Cotton Trust Protocol, a US growers' sustainability program, launched a pilot that includes "testing the traceability of regenerative bales in the supply chain," with field visits and satellite images to verify farming practices and a broader rollout tentatively planned for 2026.
Achieved whenthe program publishes how many bales it followed from gin to finished product, and an independent verifier confirms a sample.
Claimed by
U.S. Cotton Trust Protocol, September 2025
Open
01
A public module-to-bale season
An anonymized, openly licensed record of one full gin season that links every round module (its RFID number, field and harvest time from the picker) to the PBI numbers of the bales ginned from it and to those bales' USDA classing results, so anyone can map fiber quality back to fields.
The briefCredit · unclaimed02
Test set for tracing papers
A public test set of 300 made-up but realistic shipment files, each a chain of the papers used to trace cotton from gin to garment (PBI or SAI bale lists, gin invoices, warehouse receipts, yarn-mill lot records, packing lists and bills of lading), with planted faults such as bale numbers reused within five years, bale weights that do not add up and dates out of order; and a scored run of an AI agent that reads each file and lists the faults.
The briefCredit · unclaimed
Open 01 · the brief
A public module-to-bale season
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An anonymized, openly licensed record of one full gin season that links every round module (its RFID number, field and harvest time from the picker) to the PBI numbers of the bales ginned from it and to those bales' USDA classing results, so anyone can map fiber quality back to fields.
Why it is still open
USDA researchers showed the linking method at one Louisiana gin in 2019 and wrote that "no system exists that compiles all of the module ID and harvest data" into one management system. Open-source software already tracks modules from field to gin feeder: RFID Gin Data Management, MIT-licensed, merges Deere harvest files with truck and feeder scans. A 2026 open-source gin-data plugin for the AgGateway ADAPT farm-data framework, made with Cotton Incorporated, defines the fields from module ID to PBI and bale grades but ships only one made-up test row. I found no public season of real data.
Built with what exists
John Deere Harvest Identification exports; RFID scanners at the module feeder; the gin's PBI bale log; USDA classing results, which the bale owner receives for each PBI; the AgGateway ADAPT cotton gin data plugin (MIT license); Python and pandas; Zenodo to publish.
First step
Ask one gin that already scans round modules at its feeder for last season's scan and bale logs, and draft a grower consent form.
Done looks like
At least 2,000 round modules and their bales from one season, published with growers' written consent and field locations coarsened for privacy, plus a measured mapping error: on 20 modules ginned alone with a gap before and after, the share of bales assigned to the right module.
Open 02 · the brief
Test set for tracing papers
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test set of 300 made-up but realistic shipment files, each a chain of the papers used to trace cotton from gin to garment (PBI or SAI bale lists, gin invoices, warehouse receipts, yarn-mill lot records, packing lists and bills of lading), with planted faults such as bale numbers reused within five years, bale weights that do not add up and dates out of order; and a scored run of an AI agent that reads each file and lists the faults.
Why it is still open
Under the Uyghur Forced Labor Prevention Act an importer must prove by "clear and convincing evidence" that goods were not made with forced labor, and the PBI rules give checkable facts, such as no repeated number within five years. Vendors sell AI supply-chain mapping. Alerterra sells 157 made-up supplier records with forced-labor risk fields, written with an AI model, for testing screening tools, with a free sample of 100; they are supplier profiles, not bale-by-bale paper chains, and no scores are published. I found no public benchmark of such document checks and no published accuracy figures.
Built with what exists
Python with the Faker and ReportLab libraries to generate documents; the National Cotton Council's PBI numbering rules and Abrapa's SAI rules; a low-cost AI model run over every file (see the Note "The jobs agents repeat"); Hugging Face Datasets to host the set.
First step
Write the generator for one three-document chain (gin invoice, warehouse receipt, bill of lading) with five kinds of fault.
Done looks like
300 files with 1,000 planted faults, plus 100 clean files, published with an answer key and a baseline agent scored on faults found and false alarms, against a target of 90% of faults found with no more than 1 in 10 clean files flagged.