08 · Expert · Professional
Correlation — Variables Trade Against Each Other
Change one and something else always moves with it
Written by Chanho Hong
What this chapter covers
How the seven brewing variables offset and amplify one another; the procedure for predicting a result and then verifying it; which levers control taste and which control texture; and whether to suspect the green coffee, the roast or the barista first.
If the basic level asked "what does each variable determine", the expert level asks "how do the variables entangle with one another". Turn one dial and the effect of another dial changes. Read this give-and-take compensation (Give & Take) and you can predict and design the taste and texture you want.
Variables — Again, but as Relationships
The seven variables (ratio · grind · time · bloom time · bloom water volume · water temperature · ambient temperature) are now viewed not independently but as functions of one another. For example, "grind finer" raises surface area and so raises extraction yield, but at the same time it produces flow rate↓ → contact time↑, so the time variable moves with it. In other words, change one and several follow.
| Variable changed | Direct effect | What moves with it down the chain |
|---|---|---|
| Grind finer | Surface area↑ → extraction yield↑ | Flow rate↓ → contact time↑ → extraction yield↑ again (double action) |
| Water temperature ↑ | Solubility↑ → extraction yield↑ | Shift in acidity tone (malic↑); extraction of bitterness also accelerates |
| Longer ratio (1:17→1:18) | Strength (TDS)↓ | Extraction yield slightly↑ (more water washes out more) |
| Turbulence (agitation)↑ | Extraction yield↑ | Excessive → channelling → local variation in extraction yield↑ |
At the point of extraction, roasting (the availability of the compounds) and the freshness of the roasted coffee act in effect as constants, and the rest are adjusted as variables. The same variable adjustment therefore gives a different result depending on which coffee it is — a correlation always holds on the condition "in this coffee".
① Understanding the double action of grind. "Grind finer" does not only raise extraction yield through surface area↑; it raises it a second time through flow rate↓ → contact time↑. That is why a single click changes more than expected — variables are not independent, they chain.
② Freshness, the hidden constant. Immediately after roasting (before degassing) the coffee holds a great deal of CO₂, so extraction resistance is higher at the same grind. Day 3 and day 14 after roasting give different results from the same recipe, so record the recipe together with the condition "coffee at n days".
③ The two edges of turbulence. Increase agitation (turbulence) and extraction yield rises, but if it is excessive, channelling widens the local variation in extraction yield. The target is not "harder" but "more evenly" — knowing the relationships between variables lets you offset the side effects in advance (Give & Take, next section).
Give & Take Compensation — Prediction · Verification · Control
Give & Take (give-and-take compensation) is a way of thinking in which the effect created by a change in one variable is offset or complemented by another variable. The aim is to make the result "predictable and verifiable even when several variables are moved at once".
Offset in advance, with variable B, the side effect created by raising variable A. Example: grind finer (extraction yield↑, but flow rate↓ and a risk of over-extraction) → lengthen the ratio slightly to lower the strength and reduce contact, bringing it back into balance.
Check whether the prediction held using TDS and extraction yield measurement plus the senses. "Ground finer and lengthened the ratio: extraction yield 20%, TDS 1.35%, acidity retained, sweetness↑" means the prediction succeeded. Causation is only verified if you change one thing at a time.
The balance of acidity, sweetness and bitterness is set with extraction yield (how far extraction has progressed). Too much acidity, raise the yield (finer, hotter, longer); too much bitterness, lower it. Tone is fine-tuned with temperature (high = malic, bright; low = citric, softer).
Body and texture are set with grind (quantity of fines) · filter · strength (TDS). For something heavier: finer, a shorter ratio (TDS↑), a metal filter; for something lighter and cleaner: coarser, a longer ratio, a paper filter. But grind too fine and dryness and astringency attach themselves.
① Hold the taste, reinforce the body only. To avoid touching extraction yield (the taste balance), keep temperature and time as they are and shorten the ratio slightly (TDS↑), or grind marginally finer to increase fines. The taste axis is fixed; only the texture axis moves.
② Grind finer but offset the over-extraction. Grinding finer for detail brings extraction yield↑ and flow rate↓, and with them a risk of over-extraction. Lengthen the ratio slightly to lower strength and contact and offset it, and you gain the detail while avoiding the bitterness.
③ The prediction → verification loop. After the "finer plus longer ratio" adjustment, measure TDS and extraction yield and check whether the prediction (extraction yield 20%, TDS 1.35%, acidity retained, sweetness↑) held. Causation is only verified if you change one thing at a time — this is the essence of Give & Take.
Balancing — Striking the Balance at Three Levels
Even the same "balance" differs according to where it is struck. The three levels — extraction, roasting and variable control — each carry a different time scale and a different cost.
| Level | Tools | What can be changed | Cost · speed |
|---|---|---|---|
| By Extraction | Ratio · grind · time · temperature · water | Extraction yield · strength · tone · texture (day to day) | Immediate · low cost |
| By Roasting | Profile · degree of roast · DTR | Availability of the compounds · the potential balance of acid, sweet and bitter | Slow · per batch |
| By Controlling Variables | Water composition · ambient temperature · equipment | Reproducibility of extraction · baseline conditions | Per setup (semi-permanent) |
When you meet a problem, work from the cheapest and most immediate level upward: if extraction does not fix it → water and environment (variable control) → and if that still does not, roasting. The potential balance already decided in roasting cannot be entirely overturned by extraction (a grassy note from an underdeveloped core, for example, cannot be removed by extraction). Conversely, however good the roasting, poor extraction brings it down.
① Start at the cheapest level. If a single origin is "bright but sour and thin", work through ① extraction (finer, hotter, longer; extraction yield↑) → ② water (hardness↑, alkalinity adjusted) → ③ roasting (development↑ to raise the potential for sweetness and body) in that order, narrowing down the cause.
② What only roasting can do. A grassy note from an underdeveloped core cannot be removed by extraction. If no amount of variable change catches it, the problem was already decided in roasting (development), so let go of the extraction and work on the profile.
③ Reproducibility through variable control (setup). If the cup swings from day to day, what needs fixing is not the individual extraction but the baseline conditions — water composition, ambient temperature, the state of the equipment. Fix the setup level and the effects of extraction and roasting adjustments finally become measurable.
Priority — Green Coffee vs Roasting vs Barista
Quality has a hierarchy. A later stage cannot exceed the ceiling of the stage before it. So when diagnosing a problem, ask "whose responsibility is this — green coffee, roasting or barista?" in order.
Ask "at which stage was this defect already decided?" Green coffee defects (fermented taint, phenol, latent grassiness) cannot be corrected by roasting or extraction. Roasting defects (under-development, baked) cannot be corrected by extraction. Only extraction defects (under / over) can be corrected by the barista on the spot.
Order of diagnosis: ① change the extraction (cheapest and fastest) → if that fails, ② suspect the roasting (verify with a sample) → and if it still fails, ③ conclude it is a limit of the green coffee. Narrow the responsibility down by moving from the bottom upward.
Baristas often waste time trying to fix everything with extraction. An underdeveloped core (roasting) and a fermentation defect (green coffee) are not extraction problems. Know the hierarchy and you quickly separate "what can be fixed" from "what has to be escalated to the stage above".
① A complaint of wateriness and grassiness. If no combination of grind and ratio catches it, it is a signal from roasting (development), not from extraction. Feed it back to the roaster with the cupping data so that the profile can be worked on.
② Fermented taint and phenolic faults. A fermented or medicinal smell that recurs in particular cups may be a green coffee fault. It cannot be corrected by roasting or extraction, so escalate it to the green coffee stage — change the lot or tighten defect sorting.
③ Saving diagnostic time. Knowing the hierarchy prevents the waste of "trying to fix it all with extraction". When you meet a problem, escalate the responsibility in the order extraction → water/environment → roasting → green coffee, narrowing it down and quickly separating what can be fixed from what has to be sent upward.
Questions this chapter answers
- What else changes when I grind finer?
- Surface area rises and so does yield — but flow slows, contact time lengthens and fines increase at the same time. One adjustment moves at least three things. That is why variables have to be changed one at a time, with the rest held fixed, if you want to isolate a cause.
- When a cup falls flat, where do I look first — green, roast or brew?
- Upstream first. Green coffee sets the ceiling on cup quality, roasting turns that potential into actual flavour, and the barista translates it into the cup. A later stage cannot exceed the ceiling of an earlier one, so no amount of adjusting brew variables will create an axis that is not there.
- How do I go about balancing a cup?
- There are three levels: by extraction (yield and strength), by roasting (the width of the spectrum and where it lands), and by controlling individual variables. Decide which level the problem belongs to before adjusting; pick the wrong level and every adjustment moves you further away.
Figures and regulations are verified against a source hierarchy: Tier 1 in-house knowledge base → Tier 2 SCA·WBC and peer-reviewed literature → Tier 3 web. Anything not adequately supported is marked [to verify] in the text.