For years I’ve been promoting the idea of a steady decline in the ROR (rate of rise) during roasting.  Recently I’ve heard of some "educators" publicly discrediting the idea of a steadily declining ROR, so I think it’s time to address this issue. 

What is an ROR curve? 

The ROR is the number of degrees per minute that the bean-pile temperature is increasing at any point during a roast. An ROR curve is a graphical plot of that data throughout a roast batch.  The ROR tends to decline throughout a roast, primarily because the temperature gradient (differential) between the beans and the roasting environment decreases.  Please note that the “real” ROR does not go up early in a roast as the curve shown here implies. The real ROR at the beginning of a roast is quite high and drops precipitously during the first minute; however, because the bean probe that recorded the curve was hot at the charge, and the probe took a few minutes to shed its heat and match the temperature of the beans, the curve incorrectly implies that the ROR began at a low negative number and increased for the first 2:30 of the roast. 

What happens when the ROR “crashes”? 

Somewhere around the beginning of first crack, the beans release a great deal of moisture from their cores in a short period of time. That moisture is cooler than the bean surfaces or the probe, and may cause the ROR curve to drop rapidly.  This is known as an ROR “crash.”

What is “baked” coffee?

Despite what you may have heard about baked coffee being caused by long roast times (I used to believe that after hearing it for years, but ultimately did not find it true at the cupping table), negative RORs, or some strange combination of factors, the main cause of baked coffee is a pronounced ROR crash, or more precisely, a drastic change in the slope of the ROR.  Relative to a well-executed roast, baked coffee seems hollower, flatter, and less sweet.  Of course, there are degrees of ROR crashes and degrees of baking.  A roast may be slightly baked or extremely baked, and the severity of the ROR crash will correlate with the degree of baked flavor.  As brewed coffee or in a cupping, baked roasts will lose more of their tartness, sweetness, and warmth of flavor as they cool.  Green buyers may make the mistake of blaming the green quality when a cup’s acidity and sweetness “fall apart” as it cools, when the real culprit is a baked roast. (That's not to say green quality doesn't affect the retention of sweetness and acidity as coffee cools; it's just not the usual cause.)

The fine print

There are some nuances to consider when deciding whether an ROR curve indicates baked coffee.  

The speed at which your bean probe reads temperatures will affect how “crashy” an ROR curve looks.  For example, the Cropster graph above shows two ROR curves from the same roast batch; the blue ROR curve’s data came from a 3mm-diameter probe and looks slightly baked. The lower, round, orange ROR curve’s data came from a slow 10mm probe and doesn’t show the ROR crash.  Please keep in mind that these two curves are data from the same roast batch, but the smoothing effect from the slower 10mm probe hides the ROR crash. 

Below is a roast from a Probat with a slow, 6mm probe; note that the crash doesn’t look dramatic, but that same roast would look very baked had the probe been faster: 

I’ve come to believe that if a crash happens very late in a roast, for example in the last 30 seconds, the coffee will be less baked than if the crash had happened earlier.  I see many “Nordic” style roasts with a high ROR at first crack followed by a brief crash just before the drop.  Sometimes those batches taste a bit baked and sometimes they don’t. 

How do I know this is true? 

There are many “experts” who will mock the ideas in this post.  And I'm sure many readers are thinking "my RORs crash but MY roasts aren't baked." All I can say is that if you ever master avoiding the crashes, you'll change your mind.  Until then, rather than dismiss the ideas here, I hope you attempt to test them properly.  

I’ve worked on this concept with several hundred clients over the past few years and not one of them has ever wanted to go back to crashing RORs once they had learned to avoid them.  As well, almost all of them can now taste a roast blindly and know whether its ROR curve had crashed, which is strong validation of the idea.  (If looking at curve can tell you whether a roast is baked, and tasting a roast can tell you about the general curve shape, that’s the definition of validation.)  It's more difficult to detect baking in naturals or in coffees like Kenyas with massive amounts of ripe-fruit potential.  I'm sure many readers have tasted Kenyan coffees that crashed hard yet tasted quite sweet and fruity.  In my experience such coffees got even better when roasted without the crash.

Please try not to get confused, as many do, about the relationship of baking, underdevelopment, and roasty-ness.  Crashed roasts are more likely to flick and more likely to be less developed, so many roasters are used to those issues coexisting in a single batch.  That muddies the issue a bit, so please try to isolate and taste roasts that are definitely well developed and crash but don’t flick; train your palate on those roasts vs. roasts with smooth RORs and eventually you'll always know when a roast is baked. 

I recommend that you trust your taste buds, not what I tell you (I mean you're welcome to, but...), and certainly not what the industry teaches in its certification courses — the "baked" samples in those courses are usually roasted with ridiculous methods that also cause roasty flavors, confusing students and not providing a clear demonstration of baked flavor.  Instead I suggest you do the following if you want to learn to identify baked roasts: 

• If your bean-probe diameter is greater than 4mm, please replace it with a smaller probe before forming strong opinions about how “crashy” your RORs are. (If you have a "grounded" probe, >4mm may be fine, but that's an issue to discuss in a post some other time. For now, just check that your probe indicates the turn no later than 1:30.)

• Ensure your probe is fast enough and your software settings are appropriate to provide useful ROR data.

• Suggested settings would include sampling intervals of one second (standard in Cropster), an ROR interval ("delta span" in Artisan) of no more than 15 seconds, and no additional smoothing of curves when using software that offers various smoothing options. Those parameters, plus a fast probe, usually ensure quality data.

• Cup blindly— always.

• Find the correlation between ROR crashes and flatness vs. sweetness in the cup.

• Always compare cupping results to ROR curves after you have tasted the cups and taken notes.

Thanks for reading.  I'd love to hear some dissenting opinions and also your experiences with baked roasts and roast curves.