at odds with unevenness

I very recently acquired Scott Rao’s new book, Everything But Espresso, and having turned through its pages a couple of times I am struck by what a timely piece it is. I agree wholeheartedly with the spirit of the book, and to a large extent the details. The prominence Scott places on correct extraction, brew ratios, even extraction, consistency of methodologies, are all bankable principles – and I will concur with James’ verdict, it is an essential acquisition.

Scott also raises for me, some thought provoking though more debatable topics, such as coffee bed geometry, and the merits of post brew inspection of said bed, and brew diagnosis. The tenant of his point is that there are certain actions you can take to affect the end bed architecture, that will improve the evenness of drawdown, which will impact the quality of that portion of the brew (though not necessarily making up for problems with other parts of the brew).

It is hard to argue with his logic.

This train of thought brought me back to rewatching THAT video, and to rereading THAT divisive but very interesting thread on Coffeed.

One of the points that comes up was whether the grinds at the bottom of the cone were overextracted relative to the grinds at the top of the cone. Though this seems self-evident, especially when a lot of the top grinds are out of the equation at an early stage, has anyone shown real evidence of this?


To show some real evidence of this, using an Uber Boiler set to 92C I prepared 8 pour over brews (actually 9 as there was 1 duplicated with a Buono kettle). Using a Chemex, a V60, a Donut dripper, and a Melitta dripper, I prepared 2 500g (brew water weight) / 30g (ground coffee weight) brews with each, the first a continuous center pour with no manual agitation (ala the SqM video), the second using a series of small pours with stirring at the beginning and end (ala the book).

[xr_video id=”387b1ceb70e54b05b027ebad710695cc” size=”md” title=”false” viewOnXR=”false” width=700]

When these brews were completed I immediately retrieved used grinds from the filter papers, split approximately in half, resulting in two samples from each brew (a total of 16 samples) – representing the top grinds and the bottom grinds.

These were spread out on ceramic saucers and dried in a 50C oven for 30mins.

6g samples of dried grinds were weighed, and rebrewed with 100g of boiling water for 10 minutes to extract remaining solubles.

Beverage strength of these rebrews was determined with a Reichert R2mini Refractometer (aka the Extract Mojo refractometer).

For illustration purposes, soluble yields are normalized (ie graphs represent the ratio of top to bottom – with the higher yielding sample always being one).

Comments on the methodology

  • Apart from the V60 center pour / no agitation test, which was performed twice, all of these are n=1, and as such are not statistically significant, or to be considered definitive. I would encourage others to replicate these experiments.
  • 92C was chosen somewhat arbitrarily, as it was the temperature being used by 3FE to brew Hasbean’s Finca La Fany via Chemex. This was the coffee used for the experiment.
  • Ideally a higher number of samples would be taken, as I am certain extraction imbalances are likely to be gradiential rather than binary.
  • Boiling water was used without any regard for brewing convention and taste, instead merely to readily extract a large proportion of the remaining solubles.


I will present the results on a brewer by brewer basis.

I expected the V60 brew to deliver a relatively even extraction, due to the perception that substantial brew water leaves the brew via the sides as well as the bottom, also it seemed to me plausible that the center pour no agitation method could produce an even extraction if the brew water path was through an equidistant bed of coffee along all surfaces.

My results indicated (and this one was done twice – and the results agreed) that the V60 with this methodology produced the greatest discrepancy between top and bottom. I retrieved 4 times more solubles from the grinds at the top of the cone, than the grinds at the bottom of the cone, when rebrewed.

Using the stirring methodology, greatly improved this, to the point that I retrieved only marginally more solubles from the top grinds. The ratio here was 4:3, instead of 4:1, still uneven but greatly improved.

I expected the Chemex on the other hand to produce the most uneven results, with all of the brew water channeling through the point of the cone, with little side drainage.

The first Chemex brew (similar to James’ video) produced a brew greatly more even than the same V60 brew, while not as good as the stirred V60 brew. In the case of the Chemex, I found the stirring I performed to negatively impact the evenness of the extraction.

The Melitta cone, was probably the star performer on the day, producing an unstirred brew slightly more even than the Chemex brew, and only slightly less even than the sitrred V60. In the stirred Melitta brew, however, the two samples were almost indistinguishable, and certainly within the margin of error of the Reichert mini refractometer, and my experimental methodology.

The Donut, a brewer I used a lot prior to dropping it on the floor and breaking it produced the most even unstirred brew. Oddly, however, unlike the other 3 brewers it produced a brew in which more was extracted from the top, than the bottom. When stirred, however, it produced a brew slightly less even, though markedly different in that now the bottom was more extracted.


I would stress that these are n=1 findings (apart from in one instance), and should not be used to definitively call one brewer superior to another. What I would confidently take from this, perhaps, is that cone shaped brew basket geometry produces and uneven extraction in general, and that stirring can impact both positively or negatively the end result.

The V60, which I will confess has been my brewer of choice of late, on two separate brews, one with the Uber and one with a Buono kettle (at home) produced the most uneven extractions. Certainly, in the process of preparing all the brews on 3FE’s Uber, and tasting small samples of each, I remarked that the V60 brew tasted the worst, jangly, hollow even. This would not have been in keeping with my general opinion on the device.

The Donut, I find quite intriguing, as its geometry is the most unique (in this group). It could be most accurately called a truncated cone, or thought of as somewhere between a cone and a cylinder. Some say it is a pointless brewer (ahem), and I would agree with that – as such the bottleneck effect on the path of the brew water would appear to be less dramatic than any of the other brewers (although the choice of one central hole in favour of several dispersed  perhaps diminishes this). There are so many possible factors that might  explain the results I observed, with this brewer that for me to single some out would be close to idle speculation. So I won’t.

The humble Melitta cone held its own on the numbers front at least, and when stirred produced the most even extraction. Its hard for me to say why I hold little affection for the Melitta cone, perhaps not least because I cannot seem to get filter papers for it that don’t taste like sawdust. Certainly though in terms of geometry it would seem to stand a better chance than the cone shaped V60 or Chemex beds at producing an even extraction.

The question, which naturally proposes itself amidst all of this is whether an even extraction is truly the most desirable outcome of brewing some coffee. Peter Giuliano’s steak analogy on the Coffeed thread, suggests an alternative is possible. While the cooking of steak probably has more in common with the roasting of coffee than the brewing of it, and I don’t think anyone advocates uneven roasting, Peter’s point is a fair and reasonable one.

Scott Rao somewhat addresses this in his book, suggesting that because grinders produce a wide range of particle sizes anyway, that the complexity of multiple extraction levels will always be achieved, even in an otherwise “even brew”. It is unclear if Scott is merely eliminating the complexity argument as a potential criticism for his proposed methodologies, or if he believes the complexity of an extraction spectrum is desirable.

For my part, having compared a lot of brews made with sieved grinds and non-sieved grinds, I remain in favour of evenly extracted brews made from grinds all of the same size (or as narrow a size distribution as possible). I think we have acquired the taste for what our grinders can produce, and somewhat for what our brewers can produce. This is what coffee should taste like in our mind, it is the baseline from which we judge everything. We have , I propose, acquired a taste for unevenness of varying degrees. I would propose an experiment where non-coffee drinkers are given a preference test to determine whether they would choose the sifted brew over the non-sifted kind. My money would be on the former.

But I digress.

My bottom line is that I see no rational reason why we should choose an uneven extraction. Let’s not get hung up on defending our corner, our methodologies, by proclaiming unevenness “a feature”, whether due to grinder, or brewer.

It’s not a feature, it’s a bug.


39 thoughts on “at odds with unevenness

    • tibor says:

      James, perhaps a nicely recorded Penny University “Tasting Flight” preparation could serve as a new video? Also a nice way to close the Redchurch street PU project.

  1. Great post. I must confess that I’d been thinking about putting together an experiment that would measure the solubles-loss from various sections of brewed grinds, but hadn’t yet figured out a way to measure that, so cheers on the re-brewing of used grounds element. Brilliant.

    Makes me think of the other related thing that I’d been thinking about: We should all collaborate on stuff like this. We should establish a centralized point where experiments like this are proposed, with coffee-people from all over repeating the same experiment and reporting their findings. We could find a way to help manage/moderate the proposals, but yeah… could be a thing.

    • Ali Asad Lotia says:

      Nick, I was recently discussing something quite similar with a friend who is a barista at Comet Coffee in Ann Arbor, MI (my home town). I would love to get input from baristas on what sort of data they’d like to track and how they would like it presented. I write software for a living and have recently been doing a fair amount of web development, and have been thinking about doing this as an open source project that would help coffee professionals collaborate on sharing recipes, extraction yields etc. It would of course be most effective if people used a single instance rather than running their own instances since the data would become more interesting as the amount aggregated increases.

      • I’ve also toyed with the idea of a tracking system for coffee preparation, and have a working prototype using Django (or TurboGears, I forget). It’s definitely not complete by any means. I’m not in the coffee biz, but fancy it’s use for my personal experimentation, and the ol’ moleskine is getting a little cramped.

  2. Pingback: Read this blog (post) | jimseven

  3. Mind = blown. Fabulous post David, and a fabulous scientific mind to examine this all.

    Throws another wrench in my V60 brew troubles. Any thoughts on keeping a really low bed on the V60? Seems to be less chance for uneven extraction.

  4. edwin martinez says:

    very nice work.
    the question of whether even extraction is desirable is a tough one.
    i would think the answer lies in how bad you want to taste what is most distinctive in your coffee and whether that is or isn’t a more pleasant experience. very interesting experiment. left me with more questions as i think different brewers are designed to be treated differently. my take away from all this is avoid having too much variance between your most and least dense slurrys and do what ever takes to make that happen.

  5. AndyS says:

    Terrific job David, thank you for doing the work and writing it up. A few comments:

    1. While you certainly deserve all the kudos you are receiving for your methodology, readers may remember that Jim Schulman used the same “rebrew different sections of the spent grounds and measure the brew strength” technique in his espresso study several years ago . The point being, there are very few truly original ideas around.

    2. As I did back then with Jim, I wonder about the accuracy of the rebrewing grounds method. For instance, it is my casual observation that coarser grounds tend to migrate to the top of the filter. Because they are coarser, they tend to extract more slowly and remain somewhat underextracted. When you perform your rebrew technique on the top portion of the puck, these same coarse grounds will still not extract as fast and may give a deceptively low solids reading — which will make them seem LESS underextracted than they really are. So perhaps this technique UNDERESTIMATES THE UNEVENNESS in extraction yield….

    3. Regarding Peter G’s famous steak analogy, when you say “the cooking of steak probably has more in common with the roasting of coffee than the brewing of it, and I don’t think anyone advocates uneven roasting,” I’m not sure that’s true. Preparing a melange blend (ie, a blend of light- and dark-roasted coffees) is a well known trick used to add desirable complexity to an otherwise boring brew.

    4. I followed your entire writeup with enthusiastic interest, but found the conclusion to be not quite justified by the experiment. You presented reasonable n=1 evidence suggesting that stirred extractions might be more even on some brewing devices and less even on others. But aside from your subjective observations with sifted vs unsifted grounds, your final conclusion — that “uneven extractions are a bug, not a feature” — still remains in the realm of educated conjecture. I like your idea of a taste test using inexperienced coffee tasters.

    All comments aside, congratulations and thanks again for a very nice experiment.

    • 1. You are quite right. I don’t know how I hadn’t read Jim’s work before, but he uses the same approach. It’s a very nice read. There may of course be some third party we are not crediting with a prior incarnation of the technique… 😉

      2. I couldn’t say with any confidence that what you describe is truly happening (or not happening). Certainly I think the Donut result disputes that as a cause of confounding the results. That said, I could rule little out, in what is all-in-all a rather complicated environment (hence my hesitance at drawing too many absolutes from the observations). However, I think this is a good casual observation to investigate further…

      3. I don’t know how popular melange roasts are, (aside from home roasters) though I did think of them when writing that sentence. So I take poetic license in the use of the word “anyone”. Please substitute the word “many” for something more accurate. Then again, if it is to be like a steak, you might conclude the beans should be burnt on the outside and raw in the middle – not a melange.

      4. You will see that I headed the last section “discussion”, rather than “conclusion”, and the first paragraph addresses what I would call the evidence based conclusion. The rest does drift off into educated conjecture – wanton opinion slinging even. Happily so I might add, as this is a blog post, meant to be read for enjoyment, and provoke thought, not to be submitted to a peer reviewed journal. The end I hope brought back in where the discussion started, which was with Scott’s book and the previous discussions on coffeed.

      Thanks for the comments Andy. You never fail. 😉

  6. Pingback: “ Grind science ” on Daniel of Arabica

  7. An interesting post. I’m neither a scientist or professionally involved in coffee but I don’t think that matters to my question(?) – what does all this mean for *taste*?

    Whilst I admire your rigour and care in carrying out and describing this experiment I can’t help thinking, as a consumer, if you can’t tie these numbers and methods to the experience of tasting the coffee brewed then ‘so what?’ This could, very well, be a gap in my knowledge – but I think all of this ultimately must boil down to taste, and in particular the experience of the consumer. Otherwise, and to be provocative, doing these kinds of analyses, rightly or wrongly, can be dismissed as irrelevant technical fiddling by ‘geeks’.

    This is not meant to be rude and I’d be very interested to hear your thoughts.

    • Hi Sam. I’m afraid there isn’t a short answer to your question. If I can very basically give you an overview – it would be that most people agree that coffee can be overextracted and underextracted (think if it like overcooked and undercooked). This greatly impacts that flavour. What I was trying to discern here was whether certain parts of the coffee bed are overextracted or underextracted relative to other parts. Hope that helps…

      • Thanks David – a suitably concise response 🙂 I guess I need to do some more reading around what under/over- extraction means to the experience of drinking coffee.

  8. Wonderful read! Reading this makes me question why one should put more effort in V60 techniques when a Melitta produces a more even extracted result with less effort. (apart from the sawdust filters, did you try filtropa white?)

    Is this because V60 promises a more “advanced” result because of the theory of ridges?

    Is our perception of quality diffused by fancy stuff like V60, Buono kettle, Chemex? Getting all obsessed with technique (“gettin’ all mister miyagi” as Nick called it on coffeed) dreaming of the perfect “10” result, while ending up with 6’s and 7’s, while a simple device as a one hole dripper or the Clever dripper might give you 8,5’s all the time.

    Especially when talking to customers, a relatively simple device with ease of use and repeatable results is necessary.

  9. Rasmus Helgebostad says:

    Nice work. Some comments:

    1) For replication purposes, how documented were the agitation, pouring and grinding? Did you follow a certain protocol? (Stir back and forth only, in 8s, only in the top layer, pour x gs every y seconds, grind profile?)

    2) I think my main concern with agitation isn’t weither it can achieve a more even extraction. I’m more worried as to how reproducable your recipe becomes.

    3) I think using the buono is the only fair match for the stirring. Straight from the boiler is just too hard to distribute correctly.

    4) Did you measure the extraction of the different brews? I would think that a higher extraction would mean a more even extraction.

    5) I think there’s an important lesson to learn here when it comes to refractometers as well. Just like when you’re brewing espresso, uneven extraction can yield a good average extraction – but with a completely undrinkable cup as a result. (Think uneven coffee ground distribution/unaligned tamp) Never forget that there are more factors to a brew’s balance than average extraction. Your palette is still your most important brew measuring device.

    • 1. Initial agitation was until I deemed the slurry equally wetted. Grinding for brewing (ie the 30g) was weighed post grinding on a scales with 1 resolution. Weighing for rebrewing (post drying) was done on a scales with 0.1g resoltuion. The end stir was as per the video, and was only the top layer, I would estimate the spoon went no deeper than 1/3 way below the surface. When topping up the brew, my efforts were to maintain a more or less consistent bed depth, so letting it drain by no more than say 1 cm, before bringing it back up to the high water line. Grind profile was of a Mahlkonig Guatemala with relatively new burrs (except for the Buono repeat at home – which was from a Mahlkonig Tanzania).

      2. I agree – I think that is a challenge. French press?

      3. There is certainly a tactile quality to the kettle approach. I don’t know if it can achieve what stirring does though.

      4. I did. They were relatively consistent between brewers using the same method (20% +/- 1% extraction). The stirred brews all read higher than the unstirred brews though – they were not all more even though.

      5. Yes, I think average extraction as we might routinely measure by refractometer cannot be the only consideration.

      Good points.

  10. al_bongo says:

    Great post. Appreciate the attempt to measure the differences in extraction rather than just t theorising.

    For me it’s about understanding the process. Once you know what’s happening you can concentrate on the taste. I want to make a great brew every time and it’s this kind of approach that means a good tasting cup more achievable.

    Good stuff.

  11. Dave Hollinden says:

    We have a gold filter basket that came with our drip coffee maker that has a shape similar to a Mellita cone (conical in the upper portion, but forming a line at the bottom rather than a point). I’ve been setting this basket over a mug, putting in a paper filter, and brewing my coffee essentially the same way I would with a Mellita. I wonder if this might represent another option, with the possibility of the brew leaving the sides as well as the bottom.

    a thought…


  12. Pingback: Hario V60 Handfilter - Kaffee-Netz - Die Community rund ums Thema Kaffee

  13. Oliver says:

    Super interesting experiment! Bravo I’m left with many more questions though. Would love to see this experimentation go on with a group of cuppers and general coffee drinking public. I wonder how this experiment plays out with taste.

  14. Tumi Ferrer says:

    Hi David.

    A fine experiment, and I agree: Scott Rao’s book is essential. It has changed my thinking on the elements of brewing, especially with the concept of agitation.

    I agree with Rasmus on the kettle thing; it’s much easier preinfusing with it than right from the boiler, I would think.

    As I surfed the online tutorials on v60 brewing, I noticed many of them agreed on making that hole in the coffee grounds before preinfusing. Perhaps that leads to a more even extraction—more to the bottom than what came in your result. I couldn’t imagine any other purpose for making that hole…

    If I remember correctly, Scott Rao puts quite an emphasis on keeping the volume in the chamber as low as possible while brewing. I can’t reference the book now because I’ve already borrowed it to a friend of mine (never again!). You on the other hand almost fill the chamber (i.e. use the v60 like a Melitta), both in the stirred and unstirred brewing version. Which is not necessarily wrong; as an experiment, where most factors should be under control, these two brewing styles show extremes in both directions.

    I think we all have our individual idea of a mixed method of the stirred and the unstirred. Pouring over with the kettle I find more and more akin to espresso making; keeping a slow-but-steady flow requires the same consentration as tamping the coffee evenly on the portafilter.

    I hope I make some sense, if any. I just want to say that this experiment is very inspiring. I guess people are still realising the massive influence this experiment will have; and many, I think, have had the same idea, but never carried it out.

    Looking forward to seeing new results in the future!

    — Tumi

  15. Great experiment that gets to the heart of several discussions I’ve had over the past few weeks about barista training for manual brew techniques. You leave me with so many questions and a burning desire to spend a day in the lab with a sieve and a group of skilled cuppers.

  16. Pingback: Melitta (Plastik) Handfilter - Kaffee-Netz - Die Community rund ums Thema Kaffee

  17. Mattise says:

    Excellent Post David,

    to pick up quickly on the geometry part of brewer design, it’s been something i’ve been thinking about for a few months now and i’m convinced that what the numbers show in terms of evenness is down to the elongated base of the Melitta style brewer.

    I’m convinced that the longer more barrow-like shape of the extraction bed seems to lend itself to a evenness more consistently rather than the cone-sphere (is there a better name for this) shaped bed of the V60 or Chemex.

    My initial thoughts are that the shorter transit length through the grinds from top of the bed to the egress has an influence on the extraction, much work still to do though.


  18. Pingback: [Krtsch] der Woche XVIII

  19. Pingback: Read this blog (post) at Love that coffee!

  20. Pingback: Unblocked « hermitudinous dot com

  21. Pingback: Handfilter - Hario vs. Melitta - Kaffee-Netz - Die Community rund ums Thema Kaffee

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s