There is a cup I cannot forget. A certain plot, a certain lot. Floral, with clear acidity and a long finish. It was better than I could believe myself. The next year, same farm, same variety, what I thought were the same steps, and that taste never came back. I could not explain what was different. That was the most frustrating part.
I was a researcher before I was a farmer. Settling something I do not understand with "instinct" never sat well with me. So I replace what cannot be seen with what can be measured. So that that cup would not end as an accident. That was the beginning of the Data Bank.
The taste of coffee is made by microbes. But when they are working, and how hard, cannot be seen. Smell and instinct alone cannot keep up. We measure and record five numbers for every lot: pH, Brix (sugar), moisture, water activity and drying days. These five are the "gauges on the outside" of an invisible process.
pH is the speedometer of fermentation. It falls as sugar turns into acid. So you can read how far things have gone. We watch the way it falls and stop just before it reaches the bottom. Watch only the clock and on a hot day you overshoot.
Brix looks like a gauge of sweetness, and is not. This is where measuring changed my mind. We use it to read the progress of fermentation and the ripeness of the fruit. The sweetness of the bean itself is decided by how processing and drying are built (more on that in the research section).
Moisture is the end point of drying, and a letter to the roaster. Around 11%, and even across the lot. If it varies, a roaster cannot roast the same bean consistently.
Water activity is the keeper of shelf life. At the same moisture, a high value means mould and oxidation advance more easily. For a bean that must cross an ocean, it works quietly.
Drying days are the record that turns time itself into a tool of taste. Dry fast for fresh sweetness; slowly for a wine-like weight. Keep the number of days and you can aim for it again next year.
"Even if you take the numbers, in the end you only know the taste by drinking it." That was said for a long time. Recent research is starting to show the opposite. In a study of Panamanian specialty coffee that predicted cup sensory scores from seven physicochemical properties of the brew, titratable acidity, soluble solids (Brix) and protein correlated positively with the sensory score, and, interestingly, pH correlated negatively. In other words, the firmer the acidity (the lower the pH), the better. For natural process coffee the prediction reached practical accuracy, with a coefficient of determination R² of about 0.83 and an RPD of 2.34 (Vega et al., ACS Omega 2025). It has also been reported that machine learning can call sensory notes from the physical properties, defects and size of green beans (Aguila-Rodriguez et al., Processes 2024).
The significance of moisture is concrete too. At the same roasting temperature and time, green beans of different moisture give different colour, density and aroma. So evening out moisture is the precondition for consistent roasting. Moisture can be measured quickly by NIR (near-infrared), with prediction error brought down to an RMSEP of 0.57 to 0.80% (Adnan et al., Foods 2017). In short, pH, Brix and moisture are not "vague numbers"; they are numbers that actually predict quality and roasting results. The research has come quite close to what I wanted to know that day.
Here is our tightest constraint. "We grow only as fast as the Data Bank can capture the data of every lot." If chasing volume would leave the records behind, we would rather stay small. A lot with a gap in its record, good or bad, cannot be used next time. That is a repeat of the cup we could not reproduce. Being able to explain every lot is worth more to us than boasting of scale.
For example, the Laos / Typica Washed finished in January 2026 had its fermentation stopped at pH 3.9; the harvested cherries measured Brix 22.0; and drying took 20 days to bring moisture to 11.0% and water activity to 0.58 (measured 26 January 2026).
The axis of coffee evaluation is moving from origin name to process. Movements like the EU Deforestation Regulation (EUDR), which require proof of how coffee was made at the level of the farm plot, are gaining strength. For ten years in Laos we have grown, processed and dried with our own hands and kept each day in numbers. It is not a flashy accumulation. Measure, record, make again. That is all we have repeated. But it is probably the only road to being able to make that cup once more. The Data Bank is the record of that quiet stubbornness.
The measured values of our lots are listed on the DATA BANK page. For how to read pH, see Managing fermentation by numbers; for moisture and water activity, The science of drying.
New records and lot releases, delivered to your inbox first.
New lots, drying updates and field notes — every Friday, in one short letter.