BRENNEN

Industries / Food, farming and forestry / Food and beverage processing

Bottled water

The bottled water industry draws water from springs, wells and town supplies, treats it, seals it in bottles, large refillable jugs, sachets or cans, and delivers it to shops, homes and offices.

1622

Water is recorded as being bottled at the Holy Well in Malvern, England, according to Malvern Hills District Council. The council does not call it a first; the president of the Malvern Spa Association calls it the first record of spring water bottled in the UK.

malvernhills.gov.ukvisitthemalverns.org

33

Milestones achieved across 6 domains.

8

To watch for: announced, not yet confirmed outside.

12

Open deliverables. Nobody has shipped them.

In one minute

Where it started

Bottled water began as spa water, sold on the reputation of a named spring: water is recorded as bottled at Malvern's Holy Well in 1622. For three centuries it travelled in glass that was returned and refilled. In 1968 the French bottler Vittel switched to a plastic bottle that was used once, and the rest of the trade followed.

Where it stands

The UN University's water institute estimated in 2023 that the world buys about 350 billion litres of bottled water a year, worth about US$270 billion, with 60% of sales in Asia, Africa and Latin America. In Mexico 76.3% of households bought jug or bottled water to drink in 2017, and in Greater Accra most people drink water from heat-sealed plastic sachets. The rules differ by type: in Europe a natural mineral water must be clean where it leaves the ground and may not be disinfected, and in 2024 and 2025 France found that some of its biggest brands had broken that rule for years.

What is open

Much of the trade is not measured in public. The US Geological Survey found water-use figures for only 159 of 10,192 US bottled water plants, no French register says which treatments each site may use, and brand test results sit in scattered reports. The open jobs are mostly tables, laboratory tests and photo surveys that any careful small team could run: set permits against pumping, test whether a mineral water was filtered, check labels against laboratory results, and count what is still sold after a ban.

Domains

AchievedWatch forOpen

Advances to build with

1 research note

Open in bottled water

3 of 12 open deliverables

Open · Sources

Permits next to actual pumping

A public table, plant by plant, of how much water each bottling plant may take under its permit and how much it reported taking, year by year. Start where regulators publish both figures, such as Ontario, and add the US plants in the USGS inventory that have water-use records. Every row links to the document it came from.

Built with what exists

Ontario's open "Permit to Take Water" data, whose permits carry a "Bottled Water" purpose and a maximum in litres a day, and its "Water Taking" data of daily and annual volumes reported for each permit; the USGS bottling inventory and its water-use table; Python with pandas; and a small AI model to pull figures out of permit documents and company reports (see the Note "The jobs agents repeat").

Credit · unclaimed

Open · Treatment

Register of allowed treatments

A public register that lists, for each of France's 104 natural mineral water and spring water sites, the prefect's order it runs under and the treatments that order allows (filter size, ozone-enriched air, carbon dioxide removal), with a link to each order. Then extend it to other EU countries.

Built with what exists

The prefects' orders, published by each département; a browser agent to find and download them (see the Note "Deleting 80% to start over"); a small AI model to pull the treatment clauses out of long documents (see the Note "The jobs agents repeat"); and a shared spreadsheet with a source column.

Credit · unclaimed

Open · Filling and packing

Sort returned jugs by camera

An open set of labelled photos and a vision model that sorts returned 19- and 20-litre refillable water jugs into refill or reject (cracks, deep scuffs, dirt, algae, objects inside, misshapen), using a phone or a cheap camera, for small refilling plants that check jugs by eye.

Built with what exists

A phone or USB industrial camera, an LED backlight panel, Label Studio for labelling, and an open detection model such as Ultralytics YOLO, or one of Liquid AI's small LFM3 vision models fine-tuned on a single graphics chip (see the Note "One GPU to fine-tune").

Credit · unclaimed

Where it sits