When we started Prodigal, I took a leap of faith on an optical sorter from Alibaba. I don’t think I have to tell you the rest of that story 🤕.  After Mark spent months on Whatsapp with the manufacturer, only to realize the manufacturer didn’t really understand how to use the sorter for coffee, we gave up. The machine was simply too difficult to program. 

Immediately after that, I had the fortune of running into Kacper from Coffee Machines Sale at World of Coffee in Athens. Kacper pulled me to the CMS booth, telling me I’d like what he was about to show me, and that he was going to give me an optical sorter. I was intrigued.

Kacper demonstrated an optical sorter with a machine-learning (AI, if you will) programming system. He passed a “too dark” sample of beans in front of the sensors, pressed a few buttons, and repeated the process with samples that were “just right” and “too light.”  Let’s call it the Goldilocks program. The machine sorted the beans by color pretty well, and a few weeks later, Mark and I were the proud owners of a new optical sorter. Unlike the Alibaba seller, Kacper actually knew how to use the machine and coach us in its use. 

I know several roasters who have purchased a very expensive, popular sorter, only to have a technician spend TWO TO FIVE DAYS at their roastery programming the sorter. Let’s face it, if an expert — who spends all of his time programming sorters — requires 2–5 days to program a machine, the customer has little chance of mastering such a machine anytime soon. Kacper set us up remotely in less than two hours on Facetime.

For those new to sorting, please understand that even with a good set of foundational recipes, the user still needs to be skilled at programming the sorter for new coffees. There is no such thing as “set it and forget it” recipes if you want the sorter to remove quakers with any precision.

Kacper told me “I know if you use this machine, you’ll love it. And if you use it and love it, others will buy it.” I told him I would promise nothing, because I can’t promote something I don’t believe in; my reputation is worth a lot more than a $14,000 machine. 

Kacper was right, the machine is excellent, and he saved me from blowing $30,000 on a machine I would have found frustrating to use. I had no obligation to write this post, other than to pay a debt of gratitude to Kacper and CMS since the machine has been incredible for us. Paolo and I have since purchased a sorter (not free) for Regalia and Multimodal, and I’ve recommended the machine to several clients. 

Let’s discuss some details about sorting, as I think it is both more complicated and more important than most roasters realize. 

Sorting is not easy

As the examples of Alibaba and 2–5 day of initial programming show, sorting is not simple. The machine-learning system is by far the easiest I have seen. Programming a machine by showing it the actual beans to be sorted makes a heck of a lot more sense than trying to learn a complicated system of adjusting front and back cameras, red, green, and blue color settings, throughput speed, and an ultra-confusing bean-size setting I can not put into words.

What, exactly are we sorting? 

In a word, filberts*, but you may know them as quakers. The historical industry standard for a quaker is the obvious, yellowish bean that at a glance stands out in bucket of roasted coffee. If you have never brewed a cup of pure quakers, please collect some, grind them into a cupping bowl, and add that bowl to a blind cupping. You’ll know when you get to that cup🤢. You will learn a lot in that first slurp. And, I’m sorry. 

[*At Thanksgiving dinner last year, I was hand sorting quakers before brewing a pot of coffee. My friend Sharon asked me why I was removing those beans. I explained they were “quakers” and they tasted like peanuts. The next morning Sharon had forgotten the name of those beans, and when I served her coffee, she asked me if I had removed the “filberts.” 😜]

Those yellow quakers are awful. But what about all of those smoother and lighter-colored brown beans? Turns out those are quakers, too. Let’s call those pseudo-quakers for now, or filberts, if you wish.

Not sure if a bean is a pseudo-quaker? Take a potential quaker, smash it on the counter, and smell the crushed bean. If it smells like peanut, it’s a pseudo-quaker. If it smells like fruit and flowers, you just removed what was probably one of the best beans in the bucket. There is risk in over-sorting lighter-colored beans. 

In the photo below, traditional quakers are circled in white, and the red circles represent pseudo-quakers.  Please note I circled only a few representative quakers of each kind.  I did not attempt to circle all of them. 

Below: traditional quakers in green, pseudo-quakers in purple.

Two years ago I embarked on a mission with Mark to train ourselves to detect quakers. We would sort all quakers out of a sample of coffee and portion that sample into several cupping bowls. One bowl would be unsorted, one would have zero pseudo-quakers (fully sorted), and we would add one, two, or three quakers to other bowls. Then we’d scatter those bowls on a cupping table with several other coffees, taste blindly, and score. 

After a few such sessions we got pretty good at detecting pseudo-quakers. Perhaps a little too good, because we now have difficulty enjoying unsorted coffees.

What are quakers? 

Quakers are immature coffee seeds. They can be the result of subpar plant health or nutrition, or picking underripe cherry. Quakers less dense and contain less sugar, protein, and starches than mature seeds do. As a result, quakers brown slower than mature seeds during roasting. Quakers can be difficult to identify and sort in green coffee, but are more obvious after roasting. 

Producers can minimize quaker content by supplying coffee trees with adequate nutrition (starting with great soil health), picking only ripe cherry, and floating the cherries. It is far more efficient and affordable for producers to invest in preventing quakers than it is for roasters to remove quakers through sorting.

How quakers affect cup quality

How do quakers affect a cup of coffee? Have you ever described a coffee, perhaps from Brazil, as nutty? Quakers. Experienced an astringent cupping bowl? Quakers. Does a coffee have early onset of fade or baggy flavors in some cups but not all? Quakers. Quakers tend to be astringent, bitter, peanutty, and often grassy, grainy, or vegetal, depending on development level. 

It’s difficult to quantify the impact of quakers on cup quality, but I’d estimate that if a cupping bowl contains one yellow quaker, the score will drop about one point. If 10—40% of a sample is pseudo-quakers (by weight), cup score will drop by 1/4—3/4 points. While the various industry scoring systems offer guidance for scoring quakers, not all quakers are created equal. Some quakers are barely noticeable in the cup, others single-handedly destroy an entire cup.   

Quakers are not necessarily all bad. After all, people are used to tasting quakery coffee, and a bit of quaker-peanut flavor in an espresso blend meant for milk drinks may work well. 

How many quakers are in coffee? 

After becoming human quaker detectors, we naturally decided to sort as many quakers as possible from our coffee. Until that time, we never quite realized how many quakers of both types were in coffee. Average top-ten coffee from a recent COE competition? 15% quakers. Typical 87-point natural Ethiopian? 40% quakers — on a good day. Average 83-point, mechanically harvested “blender” from Brazil? It would be easier to count the non-quakers. 

Last year we purchased two very expensive coffees from a well-known Colombian producer. The arrival coffees were so quaker-riddled that we had to sort and remove 40% of the coffees’ weight in order to bring the quality up to Prodigal’s standards. Needless to say, the pre-ship sample had hardly any quakers. This is a common bait-and-switch: provide a roaster with a perfectly sorted sample, hook them, and then ship an unsorted or poorly sorted version of the same coffee. It’s as if I you bought a new Toyota Corolla from me, but I deliver you a used car and tell you “but it’s still the Toyota Corolla.”

If a roaster does not use an optical sorter, the most frequent cupping note on your score sheet should literally be “peanut.”  If you’re not noting peanut more than half the time in your cuppings, it means your brain is so used to associating peanut flavor with coffee that it is filtering out the peanut.

I’m sure many readers don’t believe that. But think about this: how often do you note “bitter” in cuppings? Probably not often. Yet every cup of coffee is bitter. Remember your very first cup of coffee?  Bitter, right? So how did bitter go from the most prominent flavor you noticed to something you rarely notice? Simply put, if something is always present, your brain learns to partially ignore it.

As the bar for clean coffee gets raised over time, more roasters will sort their coffee. Once third-wave coffee drinkers get used to well-sorted coffee, serving peanut-flavored coffee will not go unnoticed. 

I challenge the reader to try the quaker-detection training mentioned above a few times of over the course of a couple of weeks. Then, buy some Prodigal and a coffee of the same origin and process from some other third-wave roaster that doesn’t sort. Cup them blindly. Look for the peanut contrast. Once you learn to see it, you cannot unsee it. 

How much should you sort? 

Heavy sorting is expensive. We fight a daily battle to decide the appropriate amount of sorting for each coffee. We hand-sort sample roasts to what we consider a reasonable standard that our optical sorter can easily achieve. If we didn’t sort sample roasts, we’d probably reject far too many coffees, given our quaker sensitivity. But if we sort sample roasts too perfectly, they may achieve cup quality we can’t replicate with the arrival coffee. If you have an optical sorter and sort out 2%-5% by weight of each coffee, you’ll probably remove almost all of the worst quakers, but leave a fair amount of pseudo-quakers in coffees, and the peanut flavor will still be apparent. We sort to a much higher standard than that, but try to avoid extremely high sorting losses where the cost far exceeds the gains in cup quality. There are times when cup-quality improvements are more affordable via sorting, and times when one can purchase better cup quality more affordably by simply buying better green. 

Do you have experience with sorting, or are you curious about it? Please feel free to leave a comment.