You Lost Control Earlier Than You Think

Coorg, Karnataka. Late January. Ten weeks before Germany.

By the time you start talking about control, most of it is already gone.

Seven drums. Plastic. Unlabelled. Sitting in a row on the floor of a processing shed that smelled of fermentation and damp wood.

185 kilograms of coffee cherries had been floated, cleaned, weighed.

Now they needed dividing. The consultant was already writing on a notepad. He didn’t explain as he went. He expected you to watch and ask later.

I’d arrived thinking processing was where the real craft happened. The intervention. The fermentation chambers, the anaerobic tanks, the extended macerations. The part where a skilled producer left their mark on the coffee.

That belief didn’t survive the first morning.

His rule for a first year: process no more than ten percent of your yield as experimental lots. The rest goes through your existing commercial workflow. Not caution. Arithmetic. You limit the cost of being wrong.

The goal in year one, he said plainly, is not to sell anything. The goal is to send samples to roasters and collect feedback. Most producers don’t believe this when he says it. They come around eventually, usually after an expensive season of beautiful coffee that nobody could explain or repeat.

But before any of that, before the experiments, before the fermentation timelines, before the labelling, the control.

This is the part people skip. It is also the part that makes the rest of the work either meaningful or decorative.

A control sample is a portion of the same lot, same cherries, same float, same cleaning, that receives no special treatment. For naturals, straight to the drying bed. For washed, pulped and processed normally. No experiment. No intervention.

Without it, you have no baseline. You cannot know what your process contributed unless you know what the coffee tasted like without it. The interesting note in your five-day anaerobic might be your fermentation. It might be a characteristic of those specific cherries from that specific hillside that year. Without a control from the same lot, you cannot tell the difference.

You end up with a coffee you can describe but not explain. You can’t replicate it. You haven’t learned anything that compounds.

Coffee doesn’t begin in the cup. It doesn’t begin in the fermentation drum either. The decisions that shape what you’re tasting were made before the fruit was even picked, and most of them weren’t yours to make.

The control sample is not a failsafe. It’s an admission. An acknowledgment that you are working with something that already has a character, and your job is to understand it, not override it.

He set up two controls, one natural, one washed, from the same starting material as the five experimental batches. That left five drums for variables.

The structure:

D1: washed control, pulped and dried the following morning. D2: closed natural, no water, five days. D3: double ferment, five days closed, washed, then returned to water overnight before drying. D4: extended closed fermentation, ten days. D5: natural control, straight to the drying bed. D6: submerged natural, kept in water for approximately five days. D7: one more variable, to be decided.

One variable per drum. Nothing else changes. Same cherries, same drying conditions, same turning schedule, same measurement frequency.

The moment you change two things at once, different cherries and different fermentation time, you lose your ability to draw conclusions. 

You produce results you can describe but not attribute. A single variable per batch is the only way to build knowledge that accumulates rather than just accumulates noise.

Keep fermentation vessels out of direct sunlight. Not because shade is magic, but because direct sun creates temperature swings that destabilise fermentation in ways you can’t track or correct. Consistent conditions matter more than perfect ones.

Label everything immediately. Batch code, process, start date, target end date. And write down anything that deviates from the plan. A drying mat on the floor instead of a raised bed. A fermentation that ran an extra day. These are not failures. Undocumented deviations are failures, because six months later, when a roaster asks about a specific lot, the information is gone and the work loses its value.

Once batches are on the drying beds, the question becomes when to stop.

Digital moisture meters are useful and inconsistent. Readings vary depending on where you probe, ambient humidity, how recently the lot was turned. A number from a meter is a starting point, not a conclusion.

The more reliable complement costs almost nothing.

Pull a kilogram from each lot. Label it. Dry it under exactly the same conditions as the main batch, same bed, same position, same turning schedule. Weigh it every two days. Record the number.

Green coffee going to storage should reach approximately eleven percent moisture. A washed parchment coffee starting at fifty-five percent moisture will weigh roughly 505 grams when it gets there, down from the original kilo. When the bag hits that weight, consistently, across a couple of measurements, you’re done.

A bag. A scale. A notebook. It works.

Storage is where carefully processed coffee quietly falls apart.

During the resting phase, coffee still in parchment, still biologically active, it must be stored in jute bags. Not sealed plastic. Not airtight containers. The coffee is still breathing. Seal it at this stage and you’re creating conditions for mould and unwanted fermentation, undoing weeks of careful work from the inside.

After hulling, the opposite applies. Parchment removed, deterioration accelerates. Now you want hermetic sealing, Ecotact bags, to slow that process. Jute for resting. Ecotact after hulling. The transition point is not optional.

Cool, dark, dry. Bags off the floor and walls. Stack them if needed. Coffee absorbs aromas, don’t store it near anything with a strong smell. If you’re reusing old jute bags, smell them before you fill them.

I spent those few days watching this process run on a farm doing it with real structure for the first time. There were problems, a pH meter producing meaningless numbers because it hadn’t been calibrated, a drying bed shortage solved with black mats and a note in the logbook, drums that needed relabelling because someone wrote on them before the plan was finalised.

None of it was catastrophic. All of it was instructive.

What stayed with me wasn’t the process. It was the discipline behind it. The willingness to spend a full season building documentation before demanding returns. To treat the first year as an investment in understanding rather than a test of the market.

Most producers skip that. The pressure to generate returns is immediate. The payoff from structured learning is slow and deferred and requires the patience to let the coffee tell you what it is before you decide what to do with it.

In April, I would be learning the roasting side of this equation. What I was beginning to understand is that the roasting side is the end of a very long chain of decisions, most of which had already been made by the time the green coffee reached the drum.

The roaster works with what he’s been given. He doesn’t decide what he was given.

That distinction matters more than I’d realised when I booked the course.

Roaster’s Note

∙ Fermentation variability at the farm shows up as profile instability at the roaster: inconsistent development, unpredictable first crack timing

∙ Moisture inconsistency in green coffee means no two roasts from the same lot will behave identically, regardless of profile precision

∙ Control starts at the farm, long before the roaster sees the beans

Practical takeaway: When sourcing green coffee, ask for the processing log. If it doesn’t exist, price accordingly, or don’t buy.

— Roaster on the Road