Summary
When a cup comes out sour or bitter, the problem is usually the standard practice, not the coffee or the roaster. Most coffee is brewed with one fixed recipe, yet roasted coffee spans a wide spectrum of solubility. A fixed recipe is the best choice for only about 1 in 5 bags, and grind adjustment alone cannot close the gap for the other four.
The Coffee Recipe Series (CRS) takes a different approach. You measure the bulk density of your beans, which takes a minute with a measuring cylinder and a scale, and start from the recipe matched to that density. If the cup is sour you step to a recipe that extracts more; if it is bitter, one that extracts less.
Each recipe sets every brewing parameter within established industry ranges, ordered from least to most extractive. In practice, the density-matched recipe is usually right first time, and almost always within one change.
The argument rests on three claims, each examined below: roast creates a solubility spectrum, density is a practical measure of where a coffee sits on it, and a taste-led series of recipes is a faster route to a good cup than one recipe plus grind adjustment.
Coffee preparation is built around fixed recipes. For filter coffee, the SCA’s Golden Cup standard recommends a coffee-to-water ratio of 55 g/L ± 10%, the origin of the “golden ratio” of about 1:18 (Gustos Coffee Co.). The same 55 g/L figure, within a 10% margin, is one of the tests a brewer must pass to earn SCA home-brewer certification (Gear Patrol).
Espresso is more prescriptive still. The 2017 Barista Guild of America survey, run with SCA Research staff, found that the average barista uses a 1:2 brew ratio measured by weight, and 51% of respondents pulled shots averaging 25–30 seconds (Fasman, 25 Magazine, Issue 3). The typical shot: 18–20 g in, about 36 g out, 9 bar, 93 °C, pre-infusion, an 18 g basket.
These are good recipes for a coffee in the middle of the roast spectrum. The problem is applying them to every coffee.
In CRS terms, the 1:2 espresso is the recipe for a bulk density of 370 g/L. One step in the series is 20 g/L, a change you can clearly taste, so a single recipe is the best choice only for beans within about ±10 g/L of its density.
Across 9,465 home density measurements, corrected for coarse 1 g scale readings and repeated entries, the distribution has a mean of 390.6 g/L and a standard deviation of 37.2 g/L (How the density range was determined). Only about 18% of coffees fall between 360 and 380 g/L, and about 20% using the data since March 2024.
Put plainly: around 4 in 5 bags of coffee do not suit the fixed recipe.
The standard workflow adjusts one thing, grind size, and adjusts it to hit a target time, not a target taste. When the cup is sour or bitter, the usual conclusion is that the coffee or the roaster is at fault, not the recipe.
Even experts reach that conclusion. On the Coffee Lit podcast, Chris Hendon says his lab is sent really high-quality coffees, yet more often than not they don’t brew them well, and he wonders whether the problem is the coffee, the people or the technology. Asked where he thinks the problem lies, Tony Konecny says he leans towards blaming the roaster in most cases (Coffee Lit, Ep. 17, 3:07). Roasting can never be precise enough to make every coffee suit one recipe, so the recipe has to move instead.
Grind size is the right tool for setting flow and contact time, but a poor tool for moving extraction a long way. In espresso, grind is tied to puck resistance: grind finer to extract more and the shot slows towards choking; grind coarser to extract less and the puck loses resistance and runs fast.
Research supports this limit. Cameron et al. modelled espresso and predicted that extraction yield should rise steadily as grind gets finer, but in experiments yield peaked and then fell at the finest settings, which they attributed to uneven flow through the bed (Cameron et al., Matter, 2020; on modelling uneven extraction in coffee beds, see also Moroney et al., PLOS ONE, 2019). Past a point, grinding finer does not extract more; it extracts less, and less reproducibly.
So a sour shot from a light roast cannot simply be ground finer until it tastes right, and a bitter shot from a dark roast cannot simply be ground coarser. The other parameters, above all how much water you use, have to change too.
Water matters for a second reason. Sensory research at the UC Davis Coffee Center found that brew strength had a bigger effect on how coffee tastes than extraction yield did (Batali et al., Scientific Reports, 2020). Grind can only shift strength as far as it shifts extraction; changing the ratio moves strength directly, and every CRS step changes the ratio.
The usual advice is to change one thing at a time. With brewing variables this is hard to honour, because they are linked: change the dose and the bed depth, flow and time change with it.
CRS keeps the spirit of the advice while fixing its weakness. You change one thing, the recipe number, and every parameter moves together in the same direction: more extraction or less.
Nobody would cook every cut of meat with one recipe. A fillet is perfect after a few minutes on a hot surface; a brisket needs hours at low heat. One “golden recipe” of an hour at 180 °C would ruin both, and everyone understands why: cuts differ in tenderness, so the method must match the meat.
Roasted coffee has an equivalent property: how easily it gives up its soluble compounds. Like tenderness in meat, it varies widely, and it is set largely by roasting.
Green coffee is hard, dense and full of water. Roasting drives off water, generates gas, and expands the bean, so it loses mass while gaining volume, and its density falls.
Studies consistently find bulk density decreasing as roast level increases (Odžaković et al., 2019; Putri et al., 2023). One study of Robusta also found bean hardness falling as roast level rose (Yulianti et al., 2023).
This matches every home barista’s experience. A light roast is hard work in a hand grinder; a very dark roast is brittle enough to crush between your fingers.
Under-extracted coffee tastes sour; over-extracted coffee tastes bitter. This relationship has been central to coffee brewing since Lockhart’s Brewing Control Chart in the 1950s, and current sensory research has mapped in detail how sourness, bitterness and other attributes shift with strength, extraction and roast level (Frost et al., 2020; Guinard et al., 2023).
A dense, light roast resists extraction, so a mid-range recipe leaves it sour. A porous, dark roast extracts quickly, so the same recipe takes it into bitterness. The coffee is not faulty; the recipe is in the wrong place.
| Darkest roast | → | Lightest roast | |
|---|---|---|---|
| Bulk density | Lowest | → | Highest |
| Speed of extraction | Fastest | → | Slowest |
| Extraction the recipe should deliver | Least | → | Most |
| Typical espresso ratio | 1:1 | → | 1:4 |
Darker roasts extract faster, but they also have less left to extract: one study found total extraction yield falling at the darkest roasts, even as the beans became more porous (Lindsey et al., Scientific Reports, 2024). Either way, the recipe should deliver less.
Roast is not two categories, filter and espresso. It is a continuous spectrum, and even the same coffee from the same roaster varies from batch to batch.
To choose a starting recipe, you need to know roughly where a coffee sits on the spectrum. Bulk density is the most practical way to find out at home.
The method is simple: fill a 100 mL measuring cylinder with whole beans, weigh them, and multiply by ten to get grams per litre (how to measure coffee beans). It takes under a minute, needs only a 0.1 g scale and a cylinder costing about US$15, and gives a number, not a judgement.
Roasters measure roast degree by colour, typically with an Agtron spectrophotometer. That works inside a roastery, where the roaster knows the profile, but colour is an imperfect guide for the person brewing.
Two coffees can share a colour but differ in how they were roasted. In one study, slower roasting produced beans with lower hardness and water content than faster roasting at the same nominal roast level (Yulianti et al., 2023). Colour meters are also expensive, and a bag’s label (“light”, “espresso roast”) follows no industry standard.
Density responds to what the roast has actually done to the bean: how much mass it has lost and how far it has expanded. It is not a direct measure of solubility, but it tracks it closely enough to choose a starting point, and anyone can take it.
CRS does not need laboratory precision. Measurements vary with cylinder tolerance and how the beans settle, and a Class A cylinder is accurate to about ±2 g/L at typical densities.
That is well inside the 20 g/L step between noticeably different recipes. The measurement only needs to put you near the right recipe; your taste does the rest.
Roasted coffee bought by home brewers spans about 330 to 450 g/L. That range comes from 9,465 density measurements submitted between August 2021 and October 2026 (How the density range was determined).
The raw entries needed correcting before they could be used. Many came from scales that read only to 1 g, which limits a 100 mL measurement to multiples of 10 g/L, the entry form suggested a default value, and users often typed a run of values to explore recipes. After down-weighting these coarser entries and counting each session once, the key results are:
| Measure | Result |
|---|---|
| Median density | 390 g/L (95% CI 388.5–391) |
| Mean ± standard deviation | 390.6 ± 37.2 g/L |
| Share of sessions within 330–450 g/L | 90% (95% CI 89–91%) |
| Below 330 g/L / above 450 g/L | 4% / 6% |
| Same analysis, 0.1 g scale entries only | Median 390 g/L, 89% within range |
The distribution is broad and flat-topped, not a narrow peak. Even near the centre, each 1 g/L accounts for only about 1% of coffees, which is why no single recipe can cover more than a small share.
The range was set at 90% deliberately. Extending to 95% would add about 30 recipes per method to serve 5% more coffees, at the extremes where measurements are scarcest and least reliable. Coffees outside the range still get the nearest end recipe.
CRS is not a set of personal recipes. Each parameter takes its range from established brewing practice, and that range is mapped linearly across the measured density range.
The industry already works in ranges, not single values. Espresso brew ratios run from about 1:1 (ristretto) to 1:4 (lungo), filter ratios from about 1:13 to 1:20, and the heritage SCAA definition of espresso allows water at 90.5–96.1 °C and brew times of 20–30 seconds. What has been missing is a rule for which point in each range suits which coffee.
First, find where a coffee sits in the density range, from 0 at the lowest density to 1 at the highest:
ρ̂ = (ρ − ρmin) ÷ (ρmax − ρmin), with 0 ≤ ρ̂ ≤ 1
Then set each parameter from that position. Parameters that should rise with density, such as brew water, use the first form; parameters that should fall use the second:
P = Pmin + ρ̂ × (Pmax − Pmin)
P = Pmax − ρ̂ × (Pmax − Pmin)
Here ρ is the measured bulk density (g/L), with ρmin = 330 and ρmax = 450, and Pmin and Pmax are the industry limits for that parameter. Positions outside 0–1 are clamped, so out-of-range coffees receive the end recipe.
A recipe is the full set of parameters for one density: ratio, dose, yield, temperature, basket and cup size for espresso, with the equivalents for each filter method. Every parameter moves in step, so each recipe extracts slightly more than the one before it.
There is one recipe per 1 g/L from 330 to 450 g/L, giving 121 recipes per brew method. The recipe number is the density minus 329, and the 1 g/L step matches the resolution of a 0.1 g scale on a 100 mL cylinder.
The linear mapping is a design choice, not a law of nature. It is the simplest relationship consistent with the evidence that density and extraction difficulty move together, and five years of results suggest it works well. Whether a curve would fit better is an open question, discussed below.
With CRS, taste makes the decisions. Measure your beans, brew the matching recipe, and ask one question: is it sour or bitter?
Sour: the coffee is under-extracted. Press Extract More to move 20 recipes up the series.
Bitter: the coffee is over-extracted. Press Extract Less to move 20 recipes down.
Neither, and it tastes good: you have found your recipe. Use grind size to keep the brew time on target.
A 20-recipe jump is a deliberate, clearly noticeable change in the cup, large enough to judge easily.
Because all coffee is slightly bitter, sourness is usually the easier signal. A useful rule is to extract more until the sourness just disappears. If you genuinely cannot tell, brew one step in each direction and keep the one you prefer.
Think of a golfer’s bag. Instead of trying to reach every coffee with one club, grind size, you change the whole recipe to cover the distance and keep grind for the fine putting.
Of 222 coffees I’ve brewed as espresso using the series, my notes record changing the starting recipe for about 1 in 10, and by more than one step for about 1 in 25. These are my own records, not a blind trial, so treat them as a guide.
The most common reason for a change was bean size. Unusually small beans tended to behave like a darker roast and needed a step less extraction; unusually large beans needed a step more.
Compare that with the fixed recipe, which suits about 1 in 5 coffees, and grind adjustment, which cannot reliably close the gap for the rest. Starting closer means fewer wasted shots, less wasted coffee, and a good cup sooner from every new bag.
Think of the starting recipe as a weather forecast: probabilistic, not certain. A fixed recipe is like forecasting the same weather every day, right only on the days that happen to match. A density-matched recipe uses an observation to do better, and the taste check handles the rest. For coffees close to 370 g/L, CRS and the fixed recipe agree; the further a coffee is from that point, the bigger the advantage of starting from a density-matched recipe.
People who find the site tend to keep using it, coming back as they buy new coffee rather than trying it once. Users describe the results in strong terms:
“Incredible, like some kind of black magic.”
“I just do not understand why it is not common knowledge … now I drink God Shots twice a day almost.”
Testimonials and repeat use are not controlled evidence, but they show the system working for ordinary home baristas, not only its author.
While I was volunteering at Icarus, a trainee barista named John was learning on a new espresso machine with an excellent Bourbon Pointu. He was doing everything as trained, a 1:2 ratio with the grind set for the right time, but he did not like the coffee and suspected the beans.
The beans were small and well developed, and their density placed them towards the soluble end of the series. The density-matched recipe called for a much lower ratio, a larger dose and a bigger cup with more milk. The first attempt was better but still slightly bitter, so one Extract Less step went further in the same direction.
John called the second cup the best coffee he had ever tasted. Nothing had changed about the beans; under the standard recipe they had simply been over-extracted. Two attempts, guided by taste, took the same coffee from disappointing to excellent.
CRS makes strong claims, so its limits should be stated as plainly.
The density data are self-reported. They come from home users with a scale and a cylinder who chose to try a density-based system. They describe enthusiasts’ coffee, not the world’s roasted coffee supply.
The success figures come from use, not a controlled trial. They are counted from the author’s own brewing notes for 222 coffees brewed as espresso, which record when a starting recipe had to be changed. A coffee with no such note is assumed to have worked, so the figures are a best case. They also reflect one person’s palate, and the coffees were not tasted blind. Consumer research has found distinct groups of coffee drinkers with different preferences (Guinard et al., 2023), so other people may need to adjust more often, though the taste-led steps should still bring them to a recipe they like. A blind, controlled comparison against a fixed recipe would make the case much stronger.
The linear mapping is an assumption. Density and extraction difficulty clearly move together, but the exact shape of that relationship has not been measured. There is already evidence it bends at the dark end, where extraction yield falls even as porosity rises (Lindsey et al., 2024). Tasting data might show a curve fits better at one end of the range.
Density is not the only factor. Bean size, processing, variety and roast profile all affect extraction. Density captures much of their combined effect, which is why taste remains the final judge.
Measurement technique matters. Cylinder accuracy, scale resolution and how the beans settle all affect the reading, though typically by less than the step between noticeably different recipes.
None of these undermines the central point. Coffee varies, so the recipe has to vary, and an ordered series adjusted by taste is a simpler way to do that than one recipe plus a grinder.
WDT, RDT, flat burrs, precision baskets and pressure profiling all help, but they refine a shot that is already in the right place. Applied to the wrong recipe, they polish a coffee that was never going to taste its best.
CRS addresses the step that comes first: choosing a recipe that suits the coffee in front of you.
Starting closer is useful, but the bigger benefit of CRS is how quickly it gets you to a cup that tastes right.
Today, around 4 in 5 bags of coffee do not suit the fixed recipe. “Dialling in” usually means adjusting the grind until the shot hits a target time, not until it tastes good. When the result is sour or bitter, the roaster gets the blame, and many people retreat to one familiar blend that happens to work and stop exploring.
CRS changes that. Every recipe already uses the full set of tools: ratio, dose, temperature, basket and cup size. You get the benefit of all of them without needing the knowledge or skill to set them yourself, and you answer only one question: sour or bitter?
The result is dialling in that is fast and reliable: usually right first time, and almost always within one change. Sour and bitter become problems you can fix rather than reasons to blame the roaster, and every new bag of coffee becomes something to explore, not a risk to avoid.
Measure your beans and find your starting recipe →1. Batali, M. E., Ristenpart, W. D., & Guinard, J.-X. (2020). Brew temperature, at fixed brew strength and extraction, has little impact on the sensory profile of drip brew coffee. Scientific Reports, 10, 16450. doi:10.1038/s41598-020-73341-4. Link
2. Cameron, M. I., Morisco, D., Hofstetter, D., Uman, E., Wilkinson, J., Kennedy, Z. C., Fontenot, S. A., Lee, W. T., Hendon, C. H., & Foster, J. M. (2020). Systematically improving espresso: Insights from mathematical modeling and experiment. Matter, 2(3), 631–648. doi:10.1016/j.matt.2019.12.019. Link
3. Fasman, D. (2017). Defining the ever-changing espresso: findings of the 2017 Barista Guild of America Espresso Survey. 25 Magazine, Issue 3, Specialty Coffee Association. Link
4. Frost, S. C., Ristenpart, W. D., & Guinard, J.-X. (2020). Effects of brew strength, brew yield, and roast on the sensory quality of drip brewed coffee. Journal of Food Science, 85(8), 2530–2543. Link
5. Guinard, J.-X., Frost, S., Batali, M., et al. (2023). A new Coffee Brewing Control Chart relating sensory properties and consumer liking to brew strength, extraction yield, and brew ratio. Journal of Food Science, 88(5), 2168–2177. Link
6. Lindsey, Z. R., Williams, J. R., Burgess, J. S., Moore, N. T., & Splichal, P. M. (2024). Caffeine content in filter coffee brews as a function of degree of roast and extraction yield. Scientific Reports, 14. doi:10.1038/s41598-024-80385-3. Link
7. Moroney, K. M., et al. (2019). Analysing extraction uniformity from porous coffee beds using mathematical modelling and computational fluid dynamics approaches. PLOS ONE. Link
8. Odžaković, B., Džinić, N., Jokanović, M., & Grujić, S. (2019). The influence of roasting temperature on the physical properties of Arabica and Robusta coffee. Acta Periodica Technologica, 50. doi:10.2298/APT1950172O. Link
9. Putri, D. P., Andriansyah, R. C. E., Setiyoningrum, F., Yulianti, L. E., & Hidayat, D. D. (2023). Physicochemical properties of Robusta coffee at various roasting levels using different roaster types. BIO Web of Conferences, 69, 03016. doi:10.1051/bioconf/20236903016. Link
10. Yulianti, L. E., Putri, D. P., Azizah, I. F., Witman, S., Karim, M. A., & Rahayuningtyas, A. (2023). Physicochemical properties of ‘Cisalak’ Robusta coffee with hot air based roasting method. BIO Web of Conferences, 69, 03014. doi:10.1051/bioconf/20236903014. Link
11. Mayston, R. (2026). How the 330–450 g/L roast density range was determined. recipeseries.coffee. Link
12. SCA Golden Cup standard, 55 g/L ± 10%. As reported in: Gustos Coffee Co., “How to make better coffee – SCA Golden Cup standards” (Link); and Gear Patrol, “The best coffee makers, according to the SCA”, on SCA home-brewer certification (Link).
13. Coffee Lit, Episode 17: Coffee Storage with Tonx (Tony Konecny and Christopher Hendon), at 3:07. Link
