The core model
CaffiLab estimates caffeine using a multiplicative model:
C = G × F × E × 1000
Where:
- C = estimated caffeine in milligrams
- G = coffee dose in grams
- F = caffeine fraction (the proportion of the bean's dry weight that is caffeine)
- E = caffeine recovery (what fraction of the available caffeine actually ends up in the cup)
The result is multiplied by 1000 to convert from grams to milligrams.
Caffeine fraction (F)
The caffeine fraction depends on the coffee species:
| Species | Caffeine fraction | Source |
|---|---|---|
| Arabica (Coffea arabica) | 1.2% (0.012) | Farah (2012); Ky et al. (2001) |
| Robusta (Coffea canephora) | 2.2% (0.022) | Farah (2012); Ky et al. (2001) |
These are dry-weight averages across commercially available beans. Published ranges are wider — Arabica can range from 0.8% to 1.4%, and Robusta from 1.7% to 4.0% — but CaffiLab uses the central commercial values as sensible defaults.
Key references:
- Farah, A. (2012). Coffee Constituents. In Y.-F. Chu (Ed.), Coffee: Emerging Health Effects and Disease Prevention. Wiley-Blackwell. Reports typical Arabica caffeine at 1.0–1.3% and Robusta at 1.7–4.0%.
- Ky, C.-L., et al. (2001). "Caffeine, trigonelline, chlorogenic acids and sucrose diversity in wild Coffea arabica L. and C. canephora P. accessions." Food Chemistry, 75(2), 223–230. Measured caffeine across wild accessions, confirming Arabica typically near 1.2%.
- Campa, C., et al. (2005). "Trigonelline and sucrose diversity in wild Coffea species." Food Chemistry, 88(1), 39–43. Extended species-level caffeine measurements.
Blend handling
For blends, CaffiLab computes a weighted caffeine fraction:
F_blend = (arabica% × 0.012 + robusta% × 0.022) / 100
When the user doesn't know the species, CaffiLab uses a fallback hierarchy:
- Package clues (e.g., "specialty single origin" → mostly Arabica; "commercial instant" → Robusta-heavy)
- Price inference (higher prices correlate with Arabica-forward blends)
- Conservative default (100% Arabica if nothing is known)
Chicory adjustment
For Indian filter coffee, chicory is a common additive. Chicory contributes essentially zero caffeine (McCusker et al., 2003), so the effective caffeine fraction is reduced:
F_effective = F × (1 − chicory% / 100)
Caffeine recovery (E)
Caffeine recovery represents how much of the bean's caffeine actually dissolves into the brew. This is the most complex part of the model.
Base recovery by method
Each brew method has a default recovery based on published extraction data:
| Method | Default recovery | Rationale |
|---|---|---|
| Espresso | 0.67 | Short contact, high pressure, fine grind. Andueza et al. (2003) measured 50–80% caffeine extraction in espresso. |
| Pour over | 0.82 | Moderate contact, gravity flow. Gloess et al. (2013) found drip methods typically extract 75–90% of available caffeine. |
| French press | 0.78 | Full immersion, coarse grind. Slightly lower than drip due to coarser grind. |
| Cold brew (12–16h) | 0.80–0.85 | Extended contact compensates for low temperature. Fuller & Rao (2017) showed cold brew approaches hot-brew caffeine levels after 6+ hours. |
| Drip machine | 0.85 | Optimized flow rate and temperature. |
| AeroPress | 0.75 | Short immersion + pressure, highly variable technique. |
| Moka pot | 0.70 | Steam pressure, medium-fine grind. |
| Turkish | 0.88 | Ultra-fine grind, full boiling, unfiltered. Near-complete extraction. |
| Indian filter | 0.72 | Percolation through metal filter, medium-fine grind. |
Adjustment factors
CaffiLab modifies the base recovery with 12 adjustment functions, each applying a small multiplier:
-
Roast level: Light roasts can show higher caffeine in comparable brews (Ludwig et al., 2014). CaffiLab applies a small positive adjustment for light and a small negative for dark.
-
Grind size: Finer grinds increase surface area and extraction. The adjustment is relative to the method's default grind — going finer than default increases recovery, coarser decreases it.
-
Brew time: For hot methods, caffeine extracts rapidly in the first 1–2 minutes, then plateaus (Gloess et al., 2013). CaffiLab uses a logarithmic curve. For cold methods, the model rises quickly early and flattens after ~7 hours (Fuller & Rao, 2017).
-
Temperature: Higher temperatures increase extraction rate. The adjustment is calibrated around the method's default temperature.
-
Brew ratio: The coffee-to-water ratio affects extraction yield. Ratios tighter than the method default slightly increase recovery; wider ratios decrease it.
-
Extraction yield: If the user provides a measured TDS/yield, this is the strongest calibrator — it directly scales recovery.
-
Pressure (espresso only): Around 9 bar is the reference. Deviations apply a modest adjustment.
-
Agitation: Stirring increases contact efficiency for immersion methods.
-
Water minerals: Balanced mineral water extracts more efficiently than very soft or very hard water (Hendon et al., 2014).
-
Freshness: Very fresh beans resist even extraction; stale beans extract less predictably.
-
Filter type: Paper filters produce slightly lower measured caffeine than metal or no-filter methods.
-
Water pH: More acidic water (lower pH) slightly favors caffeine extraction.
Recovery clamping
After all adjustments, recovery is clamped to the range 0.42–0.97. This prevents physically implausible values — you can't extract negative caffeine or more than exists.
Uncertainty model
CaffiLab reports a confidence range, not just a point estimate. The uncertainty percentage starts at a base of 30% and is reduced by:
- Each additional known input (declared, not left at defaults)
- Brew parameters close to method defaults
- Having bean species explicitly set rather than inferred
The uncertainty is clamped between 5% (very high confidence) and 35% (low confidence).
The range is then:
- Lower bound = estimatedMg × (1 − uncertainty%)
- Upper bound = estimatedMg × (1 + uncertainty%)
Limitations
-
Cultivar variation: Even within Arabica, caffeine content varies by variety, growing altitude, and processing method. CaffiLab uses population averages.
-
Brew geometry: The shape and flow dynamics of a pour-over dripper or espresso basket affect extraction in ways the model doesn't capture.
-
Roasting loss: Roasting converts a small fraction of caffeine to methyluric acids. This is accounted for qualitatively (via the roast adjustment) but not with precise mass-balance.
-
No lab validation: CaffiLab's estimates are calibrated against published research, not against HPLC measurements of its own output.
References
- Andueza, S., et al. (2003). "Influence of extraction temperature on the final quality of espresso coffee." Journal of the Science of Food and Agriculture, 83(3), 240–248.
- Farah, A. (2012). Coffee Constituents. In Y.-F. Chu (Ed.), Coffee: Emerging Health Effects and Disease Prevention. Wiley-Blackwell.
- Fuller, M., & Rao, N. Z. (2017). "The Effect of Time, Roasting Temperature, and Grind Size on Caffeine and Chlorogenic Acid Concentrations in Cold Brew Coffee." Scientific Reports, 7, 17979.
- Gloess, A. N., et al. (2013). "Comparison of nine common coffee extraction methods: instrumental and sensory analysis." European Food Research and Technology, 236(4), 607–627.
- Hendon, C. H., et al. (2014). "The Role of Dissolved Cations in Coffee Extraction." Journal of Agricultural and Food Chemistry, 62(21), 4947–4950.
- Ky, C.-L., et al. (2001). "Caffeine, trigonelline, chlorogenic acids and sucrose diversity in wild Coffea arabica L. and C. canephora P. accessions." Food Chemistry, 75(2), 223–230.
- Ludwig, I. A., et al. (2014). "Variations in caffeine and chlorogenic acid contents of coffees: what are we drinking?" Food & Function, 5(8), 1718–1726.
- McCusker, R. R., et al. (2003). "Caffeine content of specialty coffees." Journal of Analytical Toxicology, 27(7), 520–522.