The quality of coffee is largely decided before roasting and before processing, by which cherries you pick and when and how you pick them. After ten years at LuLaLao Farm in Laos, doing the growing, processing and drying with our own hands, this is what I have come to believe.
Let me state the conclusion first. Harvest quality is not made by matching colours perfectly. It is made by selection that keeps unripe and overripe fruit out of the lot. And a cherry with high sugar does not simply become a sweet bean. There are two old sayings on the farm. One: if you pick only the ripe fruit, quality goes up. The other: the higher a cherry's sugar, the sweeter the bean. Are they true? Let me set what I do on my own farm and what the world's research says side by side, and look at them separately.
It started on a day when a lot with a little unripe fruit mixed in gave a muddy flavour. Since then, I have taken selection at the picking stage seriously.
This part is about my farm, with no research cited. At LuLaLao Farm, the harvest is never finished in a single pass. We walk the same trees again and again and pick only the deeply coloured fruit that comes away with a light pull. Fruit that resists is not ready yet, so we leave it. The cherries we pick are measured for sugar (Brix) the same day (around 20°Brix as a guide). Then we float them in water and remove the light ones. What floats is usually unripe, insect-damaged, or over-fermented. Once the green beans are fully dried, we spread them out, count splits, blackening and unripe beans by hand, and write down the defect rate. There is nothing flashy about it. We simply repeat the same thing, lot after lot. The values we measure are kept in our Data Bank records, and our partner farms follow the same procedure.
From here it is not my farm but peer-reviewed, published data. Let me line up how researchers around the world measured the same questions.
Coffee fruit comes away from the branch more easily as it ripens. In one measurement, the detachment force was 2.5–3.3 N for ripe fruit and 4.3–5.3 N for unripe fruit, so ripe fruit comes off with about half the force of unripe fruit (Zhou et al., Agriculture 2025). The hand sense that fruit which comes off easily is ripe, and fruit that resists is still early, matches the numbers. The effect of selection also shows clearly. When harvest is timed to the peak of the season, one cultivar's defect rate fell from 11% to 4%, and the cup score rose from about 79 to 81 (Huang et al., Foods 2025). Even so, selection never removes everything. Even an optimised mechanical selective harvest removed 92% of ripe fruit but left 8% of unripe fruit, with a 5.23% damage rate (Yu et al., Agriculture 2023). This is exactly why we select not only when picking, but again after picking.
As a cherry ripens, the sucrose in its skin and pulp nearly doubles, but that sugar does not move into the bean (the seed); the bean's sugar stayed almost constant through ripening (Baptistella et al., Plants 2024). Even comparing the progression of colour within fully ripe fruit (from red-orange to deep crimson and wine red), there was no meaningful difference in cup score, all around 82 points (Osorio Pérez et al., Agronomy 2023). In other words, sugar content is better understood as a gauge of ripeness than as the sweetness of the bean. Of course, sugar content is still very important in the fermentation process.
In a study that ran samples through machine learning, density and moisture were the physical properties that best predicted the sensory profile (Gonzalez-Sanchez et al., Processes 2024). Removing the light fruit by floating it in water is not about appearances. International guidance also lists removing unripe, overripe and diseased fruit by water flotation as a food-safety control point (Codex CAC/RCP 69-2009). On reading fermentation in numbers, I also wrote about it in our record on judging fermentation by pH.
Here I return to LuLaLao Farm. Having lined up the research, I found that one of the two old sayings is right and the other is a misconception. That is why I go into the farm again and again and pick selectively. On the other hand, I can no longer say that high sugar makes a sweet bean. I use Brix as a gauge of ripeness.
This season, on the lots where we were strict about selection, we brought the post-drying moisture to 12% and the water activity to 0.58, and the defect rate came to 8%. We have also started measuring density this season.
What we chase is not perfect colour. It is keeping unripe beans out of the lot as far as we possibly can.
The reference values for density and Brix at LuLaLao Farm cannot simply be borrowed from other origins. We have to confirm them with our own beans and decide for ourselves. So we measure and record every lot. To speak of how something was made in numbers is, I think, to make sure that next year's self and our partner farms can reproduce the same judgement. We grow through discernment, not scale. If you are a roaster with an interest in our beans, please also see our information for roasters.
Yes. Selection visibly reduces defects. In a study that timed harvest to the peak of the season, the defect rate fell from 11% to 4% and the cup score rose. Ripe fruit comes off the branch with a light pull, so you can tell it by hand as well.
No. As the fruit ripens, the sugar in the skin and pulp nearly doubles, but that sugar does not move into the bean (the seed). Sugar content is best used as a gauge of ripeness, not as the sweetness of the bean.
Because the fruit that floats is usually unripe, insect-damaged or over-fermented, and tends to be a source of defects and mould. Density and moisture predict the flavour of green beans well, so flotation sorting is a step for quality and safety, not for appearances.
Follow this record. New measurements, stories and lot releases from LAOS — delivered to your inbox first.
New records and lot releases, delivered to your inbox first.
New lots, drying updates and field notes — every Friday, in one short letter.