Replicating a roast

Discuss roast levels and profiles for espresso, equipment for roasting coffee.
User avatar

Postby Almico » May 12, 2019, 10:19 am

For those that do not follow SR on Instagram, he has a recent post regarding roast replication. It is much more geared toward production roasters, where duplicating a coffees flavor is a highly desirable ability, but the premiss(es) can spill over to anyone that has accomplished a great roast and wants to do it again. This post makes me very uncomfortable...

The post is here:

It begins..."What is the proper standard for roast replication? I'm frequently surprised by what some roasters consider successful roast replication. For example, using the timing of different roast "phases" is not good enough. Picking one or two arbitrary landmark moments in a roast and calling two roasts replicated if those phase times match is a low bar for replication."

Here are some profiles he cites:





User avatar

Postby MaKoMo » May 12, 2019, 10:32 am

In that post I was reading the idea to develop a numeric measure based on the differences of the RoR signals. That is not a splendid idea as one can have two profiles (in theory) with 100% the same RoR but way different temperature signals. Just offset the temperature signal of a roast by say 10C. The RoR computed from this shifted temperature curve will be 100% identical to that of the original temperature curve, however, the temperature curve is at any time 10C different.

However, the other way around, computing a numeric similarity measure on the temperature signal (as Artisan does) makes more sense. If two temperature signals are (almost) identical, their RoR signals will also be (almost) identical.

User avatar

Postby Almico » replying to MaKoMo » May 12, 2019, 10:55 am

I have seen identical RoR curves with offset milestone times. In other words, TP occurs 20s late on roast 2 but hits all the the rest of the milestones, just 20s late. Same RoR curve, but very different roasts. But I think as long as the RoR aligns with the TP exactly for two roasts, the RoR would have to match up with the BT curve.

I believe Scott's point is meticulous "between batch protocol" (BBP). Same green bean temp, same ambient temp, same roaster warm up procedure, same charge amount to .1g, same charge temp, etc. If all this is done, then matching heat and air settings to the second will generate exactly the same roast curves. But nothing is left to chance.

My issue with this is I do not always roast the same batch sizes. So that has led me to saving all my roasts in "batch size folders". I have 1,2,3,4,5,6,7,8# roast profiles saved in individual folders. When I want to replicate a roast, I use the best roast of that batch size as my background roast. I have a "best roast profile" for each coffee in each batch size normally roasted!

Speaking of which, it would be helpful if Artisan would display the actual heat settings in the background profiles instead of just blue diamonds marking the setting changes.

User avatar
[creative nickname]

Postby [creative nickname] » May 12, 2019, 12:46 pm

At the volumes I roast that kind of replication is neither desirable nor attainable. I am satisfied if I like the taste of multiple batches even if those batches have slightly different flavor notes. But I can see how his advice might be important for large, high-quality roasters who are trying to give customers a consistent experience over time.

I'll add that unless the green coffee is stored under vacuum in a deep freeze, the flavor notes will change over time anyway, and replicating profiles will be even harder as its moisture content naturally evolves over time.
LMWDP #435


Postby EddyQ » May 12, 2019, 3:09 pm

Almico wrote:Here are some profiles he cites:

And with no evidence that anyone actually tasted a difference between the two. I'm having a hard time accepting this as something to be concerned with.

I'm quite certain that even if I could replicate a roast that close, I wouldn't be able to replicate my brew close enough to tell a difference.

User avatar

Postby Almico » May 12, 2019, 5:46 pm

Scott mentions in the initial post. "NB: Set the standard wherever you feel comfortable setting it. But it is objectively true that judging replication by matching two entire curves is a higher standard than simply matching the timing of two or three points in a roast."

I roast the some of the same coffees over and over again. I sell these coffees to my customers. They expect them to taste the same. For the most part, they do. But the "standard deviation" is larger than I would like.

I started making little balsa spoke wheels for model airplanes several years ago. It didn't take me long to realize that making one wheel is easy. Making two that look the same is a lot harder. Making 100 wheels that are completely interchangeable is another matter altogether. It takes a systematic and repeatable process.


User avatar

Postby MaKoMo » May 15, 2019, 11:35 am

I wrote a short post "How close?" on how the numeric similarity measure in Artisan works.


Postby SlackBagCoffee » May 15, 2019, 12:26 pm

It is true that increasing the consistency of your "BBP" (between batch protocol) will increase your consistency, what you need to be concerned about ultimately depends on your goals.

If you need to vary the batch size, you will likely not be able to replicate the same curve in the same roaster, and achieve similar sensory results, as the bean mass changing will most likely give you a different Bean Temperature reading.

The best way to plot out different batch sizes is to understand how doing so changes (experience), and then using it, and blind tasting.

In the end, if you can get 5-6 people to not be able to tri-angulate your roasts in a blind tasting with reliability you're well on your way to replication.
★ Helpful

User avatar

Postby yakster » May 15, 2019, 1:45 pm

MaKoMo wrote:I wrote a short post "How close?" on how the numeric similarity measure in Artisan works.

Nice article and I'm sure this feature will become more popular now.

It's interesting to read this Scott Rao post about how difficult it is to replicate a roast on the same roaster with the same batch size after reading the Rob Hoos article about how to replicate roasts across different roasters. I think this discussion will help focus on what's important in roasting.

New Hoos Article in Roast Magazine (avail to public)

LMWDP # 272

User avatar

Postby Almico » replying to yakster » May 15, 2019, 3:32 pm

Robs intention was not to replicate roasts. The goal was to see if certain milestones were matched, would it mask the specific thermodynamic properties of the particular roasters. He even stated that some of the cuppers could taste a difference in coffees. What they could not do is pin the difference to a particular roaster.