Coffee fermentation

Two different jobs

Sourced

The first job is mechanical. Once a cherry is pulped, the seed in its parchment is coated in mucilage, which is too sticky to wash off. Left in a tank, the mucilage is broken down until it rinses away. The Food and Agriculture Organization's guidance on coffee processing says plainly that this is the primary purpose of fermentation in the wet process, and that the shortest fermentation which loosens the mucilage is the best one.

The second job is protective, and the same guidance explains it. Freshly picked coffee is in an unstable, wet state in which spoilage can be held back in only three ways: by encouraging harmless microbes that crowd out the harmful ones, by restricting oxygen, and by keeping the time short. A fermentation does the first two. That is why the guidance treats it as a controlled extension of a risky phase, and still recommends keeping the phase brief.

Flavour is a third thing, and the guidance is cautious about it: fermentation may contribute to quality. The trials below are attempts to find out how.

Which microbes, and why it varies

Sourced

In ordinary practice nobody adds anything. The microbes come in on the fruit, and which of them take over depends on the conditions. A 2019 trial that fermented pulped Catimor under water in Yunnan, China, found the activity led by lactic acid bacteria, with the mucilage steadily releasing nutrients into the water around the coffee. A 2022 study in Brazil that sealed its tanks found something else: a high prevalence of a family of bacteria called Enterobacteriaceae and of filamentous fungi, and few lactic acid bacteria or yeasts. A 2025 study of honey-processed coffee, which ferments on the drying bed with no tank at all, found bacteria more active where little mucilage was left on the parchment and fungi, including yeasts, where a lot was.

Three set-ups, three populations. The FAO guidance gives the general reason: the rate of a fermentation varies with the kinds and numbers of microbes that arrive on the cherries and with the temperature. It is why a fermentation cannot be specified by hours alone, and why a recipe from one farm does not transfer to another.

Five groups that keep turning up

Sourced

Studies that sequence the microbes in a coffee fermentation, in different countries and set-ups, keep finding members of the same few groups. What differs from one to the next is which of them takes over. Each note below says which study it comes from.

Lactic acid bacteria

Bacteria that turn sugar into lactic acid and so lower the pH. Leuconostoc is the genus named most often: it led a traditional open-tank fermentation studied in Nariño, Colombia, in 2019, where lactic acid bacteria made up more than 60 per cent of the bacterial sequences at every sampling, and it led the underwater fermentations in Yunnan. A 2025 Colombian study found Weissella ahead of it.

Yeasts

Fungi that ferment sugar to ethanol and a range of aroma-active alcohols and esters. The genera differ by place: Pichia in Nariño, Hanseniaspora on twenty farms in Cesar, Kazachstania in a sealed fermentation in Minas Gerais. In the Yunnan trial yeast counts were low and stayed almost constant.

Enterobacteria

A large family of bacteria common on plants, in soil and in the gut, including Enterobacter, Erwinia, Pantoea and Klebsiella. They are present at the start of every fermentation described here. In Yunnan their counts rose in the first day and a half and then fell. In a sealed tank in Paraná, Brazil, they made up about 85 per cent of the bacterial sequences in the first day.

Acetic acid bacteria

Bacteria such as Acetobacter and Gluconobacter that oxidise alcohols and sugars to acetic and other acids, and need oxygen to do it. The Colombian study that compared open with sealed tanks found them more abundant in the open ones; the Paraná study found them at under one per cent of sequences in its sealed tank.

Moulds

Filamentous fungi. The Yunnan trial found none in any of its tanks. They belong to the slower processes. A Brazilian study followed one lot of natural coffee from the tree through 22 days on a concrete drying platform and into a cold store. Culturable moulds rose about a hundredfold while the fruit dried, from roughly 1,500 colony-forming units per gram on fruit as picked to about 200,000 on the twentieth day. The species were varied during drying, and Aspergillus predominated in store. There the counts went on rising only in coffee kept in jute sacks, which took up water again in a humid chamber; in plastic bags they fell. The authors report that neither ochratoxin nor aflatoxin was detected in any sample, without showing the data.

Six studies, six communities

Sourced

What each study reports as dominant. Sequencing gives shares of DNA reads, which is not the same as numbers of living cells, and the studies used different methods, so the rows can be compared in kind and not in size.

Two sealed fermentations of arabica in two Brazilian states, one of pulped coffee and one of cherries, were led by different bacteria, and two studies point in different directions on what sealing does to lactic acid bacteria. No list of coffee's fermentation microbes holds everywhere, and a description of one tank is not a description of the process.

A model of succession in a tank

CoffeeHQ explanation

A diagram assembled by CoffeeHQ from the studies above. Each stage was observed in at least one of them; no single study observed all of them in one tank, and the order of the middle stages differs between set-ups.

Inside a fruit that is drying, not soaking

Sourced

Most of this page concerns tanks. A natural coffee has none: the whole fruit goes to the drying surface, and whatever fermentation happens, happens inside the skin while the fruit loses its water. A honey coffee does the same with the skin off and the mucilage left on. Much less was found on either, and what follows is three partial views.

The first is a clock. The Brazilian study that followed one natural lot through drying took the fruit off the tree at about two thirds water. For more than a week on the platform its water activity stayed above 0.85; it was still 0.82 at sixteen days and fell below 0.7 only around the twentieth. By the rule its authors cite, growth of any kind stops only below 0.60, which that fruit reached in its last two days. So for most of three weeks a drying fruit was wet enough to be a fermentation vessel. The study counted moulds and nothing else. They rose until the twentieth day, and its authors suggest they gained ground as drying took away the water their competitors needed. It did not count bacteria or yeasts, measure an acid or taste the coffee.

The second is a comparison of fruit with and without its skin, in tanks and not on a drying bed. The Colombian study of open and sealed tanks fermented whole fruit beside pulped coffee. Whole fruit kept more sugar and made less acid, and was slower to start, which its authors attribute to the skin. In the open tank it also ran hottest, staying above 30 degrees Celsius from the second day in a room at 20. Their conclusion for practice is specific: holding whole fruit in an open container before drying, a step sold under names such as double fermentation or cherry reserve, is not advisable beyond a day, because sound beans and living embryos were lost. The tasters' scores did not show the damage.

The third is the honey study from Yunnan already mentioned, in which the amount of mucilage left on the parchment decided which microbes were most active on the drying bed. It analysed green beans and tasted nothing.

None of the three followed bacteria, yeasts, acids and taste together through the drying of whole fruit. The familiar account, that a natural coffee owes its character to weeks spent inside fermenting fruit, fits the clock above and is demonstrated by nothing read here. This page keeps to calling it the trade's description.

The seed is not a bystander

Sourced

It is easy to picture fermentation as something done to an inert bean. The authors of the 2019 trial read their results otherwise. The concentrations of sugars, organic acids and free amino acids inside the seed changed from stage to stage, in patterns they attribute to the seed's own metabolism responding first to a lack of oxygen under water and then to drying. On that reading, some of what a process changes in the green coffee is the seed's own doing, not something that soaked in from outside. The trial measured the concentrations; the explanation is the authors' inference from them.

That matters for how the subject is usually explained. Flavour compounds made by microbes in the tank are one route; changes the seed makes in itself are another. The trials read do not separate the two.

From a microbe to a taste: four kinds of evidence

CoffeeHQ explanation

Most statements linking fermentation to flavour are one of these four, and they are not equally strong. The examples are from the studies read for this page; the classification is CoffeeHQ's.

A compound was detected

A substance was measured in the tank or in the green bean. A Brazilian laboratory study sterilised pulped coffee in an autoclave, covered it with sterile water and added one culture at a time: three yeasts and six bacteria, none of them a lactic acid bacterium. Over 48 hours malic, lactic and acetic acid appeared only where bacteria had been added, and several volatile compounds only where yeasts had. Citric acid and two other acids were also found in the uninoculated control, so they cannot be credited to a microbe. That shows what a culture can make on a seed that has been heated to 121 degrees Celsius, not on a living one. The coffee was not roasted or tasted.

A correlation was found

Two things varied together. The Minas Gerais study built statistical models linking the abundance of particular genera to the scores of five certified tasters and calls Leuconostoc, Lactobacillus and Pichia potential positive markers. A marker is not a cause: the same conditions could favour both the microbe and the score.

A mechanism was proposed

The authors explain a result by a pathway known from elsewhere. The Nariño study notes that Pichia yeasts are known to produce acetaldehyde, which contributes fruity notes in alcoholic drinks, and then says that for coffee more studies are needed to evaluate its effect on the final product. The Paraná study attributes its acetic acid to bacterial metabolism under sealed conditions and uses the word speculated.

A sensory effect was demonstrated

Coffees that differed in one controlled respect were roasted and scored by trained tasters, and the scores differed. This is the rarest kind. The yeast-inoculation trial described below is one; the Yunnan trial of fermentation length is another. Each is one experiment at one site with a small panel.

Time: what the trials found and what the guidance says

Accounts differ

In the Yunnan trial, of the three things varied (how the mucilage was removed, how long the coffee was fermented, and whether it was soaked afterwards) the length of fermentation changed the green coffee and the cup the most. Coffee left under water to 84 hours, against 36 or 48, was scored by twelve trained panellists as fruitier and more acidic, and as less floral, in a set of scores the authors describe as close to each other. In the Brazilian sealed-tank study, coffee fermented for 48 or 72 hours was described by two certified tasters as more complex and intense than coffee fermented for 24, with no difference in sweetness or cleanness.

Both findings point the same way, and neither is a rule. The Yunnan authors say the strength of the effect would depend on variety and place; their trial ran in cold weather, and the guidance names temperature as one of the things that set a fermentation's pace. Neither study pushed on to the point where a fermentation fails. And the guidance written to keep coffee safe says the opposite of longer: its code of practice gives twelve to thirty-six hours and warns that the step must be watched so the coffee does not turn sour.

The two positions are answers to different questions. One asks what makes coffee more interesting to a taster; the other asks what keeps a wet crop from spoiling. A producer who lengthens a fermentation is trading against the second to gain on the first, and how far that can go at a given farm is found by trial, not read off a table.

Choosing the microbes

Sourced

The alternative to taking whatever arrives is to add a chosen culture, as brewers and winemakers do. A 2021 trial by a research group in Australia inoculated wet fermentations with two yeasts, separately and together, and compared them with an ordinary spontaneous one. It was done in a laboratory, on five-kilogram batches of Bourbon coffee flown in from a farm. The added yeasts took over, reaching populations about a hundred times those in the control. They used the mucilage's sugars faster and more completely, and the fermented green beans held two to three times the concentration of several classes of compound: glycerol, alcohols, aldehydes, esters and organic acids. Roasting reduced those greatly, but they stayed higher than in the control.

Three certified tasters then scored the coffees. The inoculated ones scored higher for flavour, aroma and acidity. On sweetness, balance, cleanness, uniformity, aftertaste and overall score there was no statistically significant difference. So a chosen yeast changed the chemistry a great deal and the cup by a detectable amount, on a panel of three. That is evidence that the microbes matter. It is thin evidence for any particular culture on a label.

What a producer can change

CoffeeHQ explanation

Each of these was varied, or named as a cause, in a document read for this page. None has a correct setting.

Time

The one variable with a tasting result behind it in two trials: longer changed the cup. The safe upper limit is not given by either.

Temperature

The guidance names ambient temperature as one of the two things that set how fast a fermentation runs. The trials read logged it but did not vary it.

Water in the tank

Coffee can be fermented under water or without added water. The Yunnan trial fermented under water; the guidance adds that turbid water has been reported to ruin a coffee, and that clean water should be used.

Oxygen

A Colombian study of 2025 fermented the same Castillo coffee, as pulped coffee and as whole fruit, in open and in sealed tanks for up to 192 hours in a room held at 20 degrees Celsius. The open tanks ran 5 to 18 degrees warmer than the room, and whole fruit in them stayed above 30 degrees from 48 hours on; the sealed tanks stayed within 4 degrees of the room. Acetic acid bacteria made up a fifth to more than a half of the bacterial reads in the open tanks. The seeds fared differently too. In sealed tanks more than 80 per cent of embryos were still alive at 120 hours; in open tanks viability fell after 48 hours and reached about 25 per cent at 192, and the share of sound beans fell with it. The scores given by three trained tasters did not differ by tank, by time or by whether the fruit had been pulped: all but one sample, which had a phenolic defect, scored between 80.75 and 85 points. The authors conclude that time was the most influential factor for the physical condition of the seed, especially beyond 72 hours, and advise against fermenting whole fruit in an open tank for more than a day.

How much mucilage

A machine can remove all, some or none of it before the coffee ferments or dries. In the Yunnan trial, removing it by machine before a short spell under water gave a cup the authors call comparable to fermenting it off, though less fruity in odour.

Which microbes

Left to chance in most processing. One trial that added chosen yeasts found a large change in chemistry and a smaller one in the cup.

Soaking afterwards

In the Yunnan trial a day's soak in clean water after washing drew compounds out of the beans and brought the green coffee from different lots closer together in composition.

A comparison to run yourself: the word before the taste

CoffeeHQ explanation

A protocol for a reader with two coffees and a friend. It tests the taster, not the fermentation: a study described in the guide to the newer process names found that the information on a label changed what consumers reported. No result is given here, because CoffeeHQ has not run it.

• Does knowing the process change what you taste?

  Change

  Whether the taster knows which cup is the fermented lot. Have someone else brew two coffees from the same roaster, one sold with a fermentation word on the label and one without, and pour them into identical unmarked cups. Taste and write notes. Then repeat on another day with the cups labelled.

  Keep the same

  The brewer, the recipe, the grind, the water and the cup temperature, and the days since roasting as nearly as the two bags allow. Do not look at the bags before the blind round.

  What to notice

  Whether your notes for the same coffee change once you know its process, and in which direction. Two coffees from one roaster still differ in origin and roast, so this cannot tell you what the fermentation did. One taster on two days is an observation and not a result.

When it goes wrong

Sourced

The documents read say less about failure than about success, and what they say is general. The code of practice warns of sour flavours from a fermentation that is not monitored. The FAO guidance notes that cherries held in heaps or sacks before drying ferment hot and fast, in a way it calls uncontrolled, with alarming outcomes sometimes recorded. It also notes that heavy fruit-fly infestations can unbalance a fermentation, and that the acids microbes produce on pulping remnants can damage equipment that is not cleaned.

The trade has a vocabulary for fermentation faults, with words such as sour, vinegary, fermented and, at worst, stinker. CoffeeHQ has read no study that maps those words to a cause, and does not offer one.

What the trials read have not settled

CoffeeHQ explanation

How long, how warm or how acidic a fermentation should be. The sources give a safety range for one and no targets for the others.

That longer fermentation makes coffee fruitier, as a rule. Two trials found a change in that direction, each at one site, and neither tested the limit.

Which compounds made in fermentation survive roasting to be tasted, or which microbe makes which flavour. One trial found fermentation compounds much reduced by roasting but still higher; none read traced a flavour to its source.

What happens inside a drying natural or honey coffee in any detail. Most of the work read concerns tanks, and the three partial views given above are all that was found.

A microbiome of coffee fermentation. Six studies found six communities, and the two that sealed their tanks in Brazil were led by different bacteria. A product or a protocol that promises a particular set of microbes is promising something no study read has shown to transfer between farms.

How much oxygen is in any tank. None of the studies read that use the words anaerobic or semi-anaerobic reports a measurement of it; the words describe the vessel.

What to do next

• Washed process — The method in which fermentation has its original job, with the Yunnan trial in more detail.
• Anaerobic fermentation — The sealed-tank study in full, and what the word on a bag does and does not tell you.
• Experimental processing — The newer names, taken one at a time, with what has and has not been published about each.
• Drying coffee — What comes next, and where the guidance places most of the risk.
• Tasting Lab — A plan for tasting two processes side by side, with what the comparison cannot tell you.