On the back label of each of our chocolate bars, there is a pair of handwritten initials. These belong to the bar’s “owner,” the person who developed the roast profile that we use each time we make that chocolate. It’s an old Dandelion tradition to assign a new owner to each new set of beans, and it’s not a small job. But, as I came to realize over the last few months, it’s a wildly interesting and delicious process that uncovers some mystery, and leaves you with more surprises than you could hope to understand.
When I heard we had bought beans from Tanzania, I wanted to develop the roast profile because I hoped to get closer to the process, but also because I have a sentimental attachment to the place. I have some family in Dar es Salaam, and I lived there with my cousin for a short while a couple of years back. I remember it fondly: the dusty air, the open markets and their heaps of beans and mangoes, milk in plastic bags that I bought with the dubious scraps of Swahili I’d gleaned, and the unfinished dirt roads that, under the wheels of a Bajaji, seemed bent on killing you. It was not a romantic time in my life, but I miss it.
I had been in touch with Simran Bindra, the cofounder of Kokoa Kamili, while we were developing the Sourcing Report at Dandelion, and I’d learned all about the way Kokoa Kamili is raising the quality and the price floor of cacao in Tanzania. I had also learned that he and my cousin were good friends, not a huge surprise given the size of the expat community in Tanzania, but a serendipitous one nonetheless. When it came time to develop the roast profile, I dove right in like it was meant to be.

Sorting the first of the beans from Kokoa Kamili.
To begin the first round of tests, I cast the customary wide net of roast times and temperatures to see what flavors came forth across the spectrum. I collected votes from across the company, and we were surprised to find that the two favorites were a whopping 40° apart, at 180° and 220°, both at 12 minutes (for a single kilo at a time). Both of those test batches were lovely and balanced; the former had an unusual light, sparkling acidity, like young raspberries, and the latter a handful of warmer, more chocolatey notes. And so, I split the focus between those two, and narrowed the parameters around them.
Throughout the first few test rounds, I prepared twenty to twenty-five kilos of the beans at a time (with lots of generous help from other Dandelions). We roasted, cracked, and winnowed the beans when the production team was on lunch or had gone home for the day. When the nibs were ready, we brought them upstairs to the row of miniature melangers that we use for test batches. These baby refiners are called Premier Wonder Grinders, and they live on the mezzanine upstairs.
For each test batch, I weighed out 700 grams of nibs and 300 grams sugar, the right ratio for our 70% bars. In order to give the baby melangers a head start on the refining process, we pre-grind the nibs by running them through peanut grinder first. Depending on the roast and the beans’ natural ratio of fat to solid, the nibs turn into anything between a dry and crumbly mess to a wet and goopy sludge. Next, it must be fed slowly, chunk by chunk, into the spinning grinders. We let those run for an hour, until the nibs are broken down into a thick “liquor,” and then we add the sugar.

Weighing the nibs.
In the earliest tests, we control sugar as a variable, adding the same amount to each batch at the same time. Only when we’ve nailed the roast do we play with the timing of the sugar.
The next morning, after about 14 to 17 hours in the baby melangers, it’s time to pull the batches. This involves a fair amount of spatulas, scrapers, silicon trays, gloves, plastic bags, and, if you’re me, spilling. I poured a little bit of each batch into a mold divided into a hundred little cubes, then poured some more into a plastic bag for record keeping in our “chocofile,” and dumped the rest into a tin pan for cooling, destined for making brownies at home. Once cooled, I popped the little cubes into four or five paper cups for a blind tasting, and politely coerced every person I could find in the factory into tasting and voting. Then, I’d waddle down the stairs balancing tiny melangers on either arm to spray them clean with water in the kitchen.
At each of the six rounds of test batches, I couldn’t find that magical, sparkling balance of tart red fruit and warm chocolate tones. I experimented with roasts that were only five degrees or one minute longer, and that perfect balance slipped away. I couldn’t understand it, and I kept going back and tasting the choco-file to make sure I was remembering right. Would we really stick with a roast from the first round? After all that prepping and weighing and roasting and scraping? But sometimes, that’s how it happens. It could take a year and countless tries to get it perfect, but sometimes you nail it right away.
In the end, that’s what we went with. My personal taste skews more towards warm, chocolatey chocolate. I like smoked nuts, heavy caramel, and even a hint of leather here and there. (In my past life, I must have been an old man who smoked a pipe and drank lots of scotch.) But I love this chocolate. It moves from a sparkling, juicy red fruit to a rich brownie batter finish. Like raspberries dipped in fudge, or molten chocolate cake with a strawberry on top.
We just made our first big batch in a 30-kilo melanger on the production floor, and we will temper the first round this week. If all goes well, it’ll be on shelves before the end of June! Stay tuned.
Sounds as though Mollie has a spectacular life. I would love to be able to develop the roast profile of a batch of chocolate. Instead I can’t really complain that I get to eat the chocolate. I am looking forward to trying it soon. Thanks
Hey guys,
I really enjoyed it reading this post. If anyone had the time to answer, I have a burning question I’d like to ask about the chocolate making process. Its regarding the process of ‘conching’ its a reoccurring word used in describing the production of chocolate but your process seems to be omit or do without it. This has me curious. Visually, the coaching process looks similar to the process you use create the cocoa liquor in the melanger, but are they the same thing?
The impression I tend from the other sources I read is that there’s an element of conching used in commercial production of chocolate that involves, reconstituting fats and other soluble components of cocoa, that isn’t relevant in the creation of speciality chocolate. Is it a relevant point of difference or an irrelevant semantic?
Thank you guys, and love your work.
OMG!!
Ate my first Dandelion Bar (Mantuano, Venezuela 2014 harvest 70%) that I bought at the Ferry building on a whim. I would work for you for chocolate alone. Isn’t that the primary food group?
I’m in love!!!
I am testing these beans too and in my first notes I used the word “sparkle” to describe flavors in the early roast. I am so tickled to see that you found the same, and YES, I’m finding the same tantalizing spread of flavors across the roasts. These beans are so special!