I have been roasting on the Bullet R1 v2 for years (2020) and never put much effort into learning how to improve via profile analysis. That changed last year as I started to geek out on roast science. I have been creating some blog posts about how to detect common defects. I am curious to hear how most bullet users are leveraging the data. Are you reviewing your profiles/curves? I know the only real thing that matters is the end result and how the coffee cups. I have seen real improvements by spending more time in the data. Curious about others.
2 Likes
You are providing some really useful data review tips! I’ve been roasting on a Bullet R1 about as long as you, but I’ve never really felt like I’ve gotten the perfect cup. Before the Bullet, I roasted using a GeneCafe, which obviously had way fewer variables to mess with, and I consistently got what I was looking for, which is a City+ light roast with vibrant origin bean character. I favor Central and South American as well as some African beans (Kenya, Tanzania).
Anyway, I’m going to check out more of what you’ve learned as it is specific to the Bullet. One thing I’m thinking about is deviating from my typical 755g batches. I think I might be constraining myself too much by that high of a thermal load for the Bullet to handle. What are you thoughts?
2 Likes
Thanks, that is kind of you to say.
My honest thought on the perfect cup, I stopped chasing it lol. Taste is too subjective for perfect to mean much of anything, and chasing it just made me second-guess roasts that were actually good. These days the analysis works more like a coach for me. It points at where I have room to improve after every roast, the way a good coach has notes after a win, not just after a loss. It catches my common errors and shows me the trend over months. What it can never do is tell me how a particular coffee will show up in the cup. Only tasting does that, which is the part I enjoy most anyway. This is where the real craft comes in!
On your batch size, I think your gut is right. The Bullet is rated to a kilo, but Aillio’s own getting-started guidance calls roughly 250g to 700g the practical range, so 755g has you living just past the edge of the well-tested zone. Roasting, to me, is managing a big ball of heat: how fast you bring it up, how you walk it through Maillard, when you stop feeding more heat to it, and how quickly you pull the heat back out at the end. The bigger the ball, the harder every one of those moves gets. Drying drags, the charge has to run hotter, and hot charges are where tipping and scorch start showing up. A smaller ball is simply easier to steer. I do almost everything at 500g. My plan was always to dial in a profile there and scale up when I needed the volume, but I am pretty happy with 500g roasts.
Since you like Kenyas and Tanzanian peaberries, one more point to account for… those tend to be dense beans, and density tightens the window the same way weight does. A heavy batch of a dense bean is more challenging to manage, which might be part of why 755 has felt like a fight. Not only do you have to worry about how to get the heat into the large quantity of beans, but high density takes on heat slower (all your changes to heat management have a longer reaction time between the change of the lever and the time the instruments really start accurately reporting on that change).
And I would not read too much into the machine change. Switching roaster types resets more than people expect. You did not lose the touch, the physics changed under you. Fluid bed is very different than drum, and drum on an induction bullet is not equal to a gas drum.
2 Likes
Hey Jeffras, I started roasting only a couple of years ago and only since getting the Aillio have I even looked at any graph profiles. Right now I see myself focusing on the ROR specifically and I’ve been trying to zero in on what my inputs translate to in the cup. I’m framing the challenge less in terms of finding the perfect cup and more in terms of what can I do to get x variable.
I have experimented with lots of heat at the beggining then tapering off before first crack and riding it out to my final temp. Today I tried lowering the charge temp significantly ( increase drying stage) and riding out first crack with higher applied heat, this resulted in lower overall ROR and longer roast time. Will be cupping this weekend.
From my current roast style, high heat in the beggining, I’ve gotten mostly bitter notes, some sweetness, some cacao, very little acidity.I personally enjoy a bright and sweet cup but don’t mind a dark chocolate bitterness. My goal is to figure out how I can maximize acidity and sweetness. ( All of this knowing that beans are all unique and have their own potentials, roasting can only bring out the flavors already inside)
What notes do you look for when cupping and how do you try and coax them out?
@americancoffeeroaste unfortunately my motivation for digging into the roast profiles started because I was getting common defects in cupping (grassy, baked, underdeveloped, etc). That really started driving my deep interest into how/why coffee ends up tasting so wildly different in the roast. I wouldn’t say I am looking for anything specific in any roast, I just love to roast different coffee and see what they have to offer, but I am definitely looking to avoid these defects 
Sweet Marias random coffee subscription has been one of my favorite splurge items fo my own roasts. I love getting the 4 pack of random beans, then it is one and done as these beans are usually out of stock for re-order lol.
1 Like
I mainly compare three things: what I tasted, where the curve changed, and what the recipe or machine was doing immediately before that change. I try not to diagnose every RoR wiggle as a defect. First I look at the overall architecture—drying, the approach to first crack, the transition through it, development, and the finish—then ask whether a specific Power, Fan, or Drum action plausibly contributed. When I make a correction, I prefer one small change followed by another roast and cupping rather than stacking several changes at once.
I developed RoastRecipeAdvisor to help with that type of review. It examines a completed Bullet roast in the context of the recipe, brings the IBTS/RoR curves, event markers, and machine actions together, and helps identify the most important recipe change to test next—or determine that the curve does not justify a change. I still treat it as decision support; the cup is what tells me whether the adjustment was useful.
there are just too many things that can affect the coffee and you might never actually be able to figure out why. A lot of what makes the coffee taste good is not really in the user’s control and has at least as much of an impact on the taste.
For example, if you have a defect it could be the green beans themselves and your roast curve is perfectly fine. The environment also has an impact.
There is also user error like mixing up the beans or weighing them wrong, or not paying close enough attention to the roast.
Roast profiles are great for replicating batches. However aside from that, the only other use it might have is diagnosing a gross roasting error and trying not to repeat it on the next batch. It’s a bit of a waste of time to try to attribute incremental profile adjustments to a particular change in taste. I’ve changed fan and power and drum settings all over the place on different batches, but as long as the curve looks pretty close it’s really hard to tell them apart.
In short, there is a limit to how useful data is
I always review past roast profiles by roasting exclusively in overlay mode. I take notes on what I liked and make adjustments the next time. When I start with a new bean, I pick a profile that I think will be similar, overlay that, and go from there, without trying to match it perfectly. Over time I’ve learned what works. But I’ve personally never found it all that useful to compare roasts in general from different beans at different times.