BRENNEN

Industries / Food, farming and forestry / Food and beverage processing / Soft drink bottling

Distribution

Distribution is the work of getting drinks from the bottling plant to the shops, restaurants and stalls that sell them: franchised bottlers and their territories, trucks and handcarts, coolers in shops, and the orders that keep them stocked.

6

Milestones achieved, from 1899 to 2017.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1899

    Bottling rights go to Chattanooga

    In 1899 Benjamin Franklin Thomas and Joseph Brown Whitehead, two lawyers in Chattanooga, Tennessee, obtained the rights to bottle Coca-Cola across the United States except Mississippi, where Joseph A. Biedenharn of Vicksburg had held rights since 1894. They sold exclusive territories to large regional "parent bottlers", which in turn franchised thousands of small local bottlers. The Coca-Cola Company kept making the syrup; independent bottlers mixed it, bottled it and delivered it in their own towns.

    georgiaencyclopedia.org

    Credit

    Benjamin F. Thomas and Joseph B. Whitehead; The Coca-Cola Company under Asa Candler

  2. 1940s

    Bottling follows the troops

    During the Second World War Robert W. Woodruff, president of The Coca-Cola Company, pledged that every American serviceman and servicewoman could buy a Coca-Cola for five cents wherever they were stationed. Supplying that pledge left the company with bottling plants around the world when the war ended; its export arm, the Coca-Cola Export Corporation, had been formed in 1930.

    georgiaencyclopedia.org

    Credit

    Robert W. Woodruff; The Coca-Cola Company

  3. 1980

    Exclusive territories made lawful

    The US Soft Drink Interbrand Competition Act of 9 July 1980 says that antitrust law does not make it unlawful for a soft drink trademark owner to give a bottler the sole right to make, distribute and sell its drinks in a defined area, as long as those drinks face "substantial and effective competition" from other drinks. It protected the local territories on which the franchise system and its delivery routes were built.

    law.cornell.edu

    Credit

    US Congress, Public Law 96-308

  4. 1999

    Pushcarts reach small shops

    In 1999 Coca-Cola Sabco, a bottler with 18 plants in East and Southern Africa, opened ten Manual Distribution Centres in Addis Ababa, Ethiopia. Each is a small independent depot that serves about 150 nearby shops by pushcart or bicycle, where trucks are slow or costly and shops can only afford small, frequent deliveries. By November 2008 Ethiopia had 651 centres carrying 83% of the bottler's sales there and Tanzania 412 carrying 93%, and the Coca-Cola system counted more than 2,500 centres in Africa with over 12,000 jobs, according to a review by the Harvard Kennedy School and the International Finance Corporation (IFC). Owners earn a set margin per case, which the review puts at 3% to 5%.

    dash.harvard.edu

    Credit

    Coca-Cola Sabco; review by Jane Nelson, Eriko Ishikawa and Alexis Geaneotes

  5. 2013

    Medicines skip the drinks crates

    ColaLife started from the observation that Coca-Cola reached remote shops that medicines did not, and designed a pack of diarrhoea treatment for children, oral rehydration salts and zinc, to fit between the bottles in Coca-Cola crates. In its Zambian trial, from August 2012 to August 2013, only 4% to 8% of retailers carried the packs in crates. The packs moved through ordinary wholesalers and shops instead, and zinc use among children with diarrhoea in the trial districts rose from under 1% to 46%. The crate-shaped pack was dropped for a cheaper bag that also measures the water.

    Lessonwhat a drinks network lends other goods is its way of trading, with set margins for each wholesaler and shop, not the space in its crates.

    pmc.ncbi.nlm.nih.gov

    Credit

    Simon Berry and Jane Berry, ColaLife; Zambia's Ministry of Health; UNICEF Zambia

  6. 2017

    Coca-Cola hands back bottlers

    On 30 October 2017 The Coca-Cola Company said it had finished returning its company-owned US bottling operations to independent bottlers, nearly 70 of which now ran the plants and trucks. By the company's figures the handover took 60 transactions covering 350 distribution centres, more than 50 plants, more than 55,000 employees and more than 1.3 billion physical cases of volume; the US system had been largely company-owned in 2013.

    Lessonownership of plants and trucks swings between the brand owner and local bottlers; Coca-Cola formed Coca-Cola Enterprises in 1986 to buy bottlers, and 31 years later gave the US operations back.

    d1io3yog0oux5.cloudfront.netgeorgiaencyclopedia.org

    Credit

    The Coca-Cola Company; bottlers including Liberty Coca-Cola Beverages, Reyes Coca-Cola Bottling and Swire Coca-Cola USA

Watch for

  1. Software picks the shop's order

    Coca-Cola Bottlers Malaysia-Singapore-Brunei says Perfect Basket, a feature of its Coke Buddy ordering app used by about 39,000 shops in Malaysia, suggests which drinks and how many each shop should order, from its past orders, the season, the weather and what similar shops buy. It says 83% of participating shops took the suggestions during a campaign, and that those shops grew sales faster than comparable ones, without giving the size of the difference.

    Achieved whenan outside party publishes a comparison of stock-outs, sales and stock held between shops that use the suggestions and similar shops that do not, with the size of the effect.

    pymnts.com

    Claimed by

    Coca-Cola, through Patrick Go of Coca-Cola Bottlers Malaysia-Singapore-Brunei, September 2026

Open

  1. 01

    Open test for cooler photos

    An open set of photos of drinks coolers in small shops, with every bottle and can labelled by product and every empty slot marked, used as a public test of how well cooler-reading software counts stock and gaps from one phone photo, and a baseline model scored on it.

    The brief
    Credit · unclaimed
  2. 02

    Suggested orders tested openly

    A randomized trial in small shops that compares an open suggested-order tool for soft drinks with shopkeepers' own ordering, measuring days with empty cooler slots and the cash tied up in stock.

    The brief
    Credit · unclaimed

Open 01 · the brief

Open test for cooler photos

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open set of photos of drinks coolers in small shops, with every bottle and can labelled by product and every empty slot marked, used as a public test of how well cooler-reading software counts stock and gaps from one phone photo, and a baseline model scored on it.

Why it is still open

Bottlers already read coolers by camera: Coca-Cola Hellenic uses Trax's image recognition to audit stores and reports 97% recognition of products, a company figure, and a GitHub project by two developers says its cooler scanner was sold to a PepsiCo bottler. Their images and accuracy tests are private. The public sets found are small or come from other settings: Roboflow's "Retail Coolers" project shows 216 images, and VCBev-8, posted in November 2026, holds 4,560 images of 8 drink types inside automated vending cabinets. None found covers the small shops of Africa, South Asia or Latin America that depots such as the Manual Distribution Centres serve, or marks empty slots.

bestpractice.aigithub.comuniverse.roboflow.comieee-dataport.org

Built with what exists

Phone cameras; Label Studio, with Meta's Segment Anything to speed up outlining; an open detector such as Ultralytics YOLO; and a small vision-language model fine-tuned on one graphics card to read product names (see the Note "One GPU to fine-tune").

  1. First step

    Ask 20 shopkeepers in one neighbourhood for permission to photograph their coolers twice a week for a month, and label the first 200 photos.

  2. Done looks like

    At least 3,000 cooler photos from at least 300 shops in at least three countries, under an open licence. On 300 photos from shops not used in training, the baseline's count of empty facings is within one of a person's count in at least 90% of photos, and its product names match the labels for at least 95% of items.

Open 02 · the brief

Suggested orders tested openly

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A randomized trial in small shops that compares an open suggested-order tool for soft drinks with shopkeepers' own ordering, measuring days with empty cooler slots and the cash tied up in stock.

Why it is still open

Coca-Cola's bottler in Malaysia reports that shops following its app's suggestions grew sales faster, with no size or method given. Academic trials are under way on nearby questions: MIT's LIFT Lab has designed a field experiment in Mexico in which shopkeepers log sales of two products and get a chatbot's advice, and a trial in Lima, Peru, registered in July 2025 and due to end in May 2026, promotes a digital sourcing app to corner shops. None found tests soft drink suggested orders against a public baseline.

pymnts.comliftlab.mit.edusocialscienceregistry.org

Built with what exists

A simple open forecasting model trained on each shop's past orders, such as a seasonal moving average or LightGBM; a weekly suggestion sent by WhatsApp or text message; weekly cooler photos read by the model from the item above; and a pre-registered analysis plan on the American Economic Association's RCT Registry.

  1. First step

    Agree with one independent wholesaler or distribution centre owner to share six months of order history for 30 shops, and check whether a simple model would have predicted their weekly orders better than repeating last week's order.

  2. Done looks like

    At least 60 shops, half chosen at random to receive suggestions for 12 weeks, with results reported for empty facings of the 10 best-selling drinks, units sold and stock value, and the code and anonymized data published.