How we Hone in on Style and Consistency

By Justin Dedini

This post is about how we hone in on roast style and consistency at Pastime and how we use a data driven approach to get there. The framework leans on three pieces of QA equipment: a moisture meter, a color meter, and an accurate scale. The meters have gotten a lot more affordable in the last few years which opens up a level of QC capability that was recently only accessible to companies of much larger scale. Used right, these tools allow for establishing a framework that saves time and coffee by establishing objective benchmarks for a given roast approach. New coffee dial-ins become quicker and more predictable and batch to batch consistency becomes even easier to achieve.  

Dialing in a new coffee at Pastime starts before the first roast. We measure the green for moisture content, then we roast and measure again, we record the weight loss of the batch right out of the cooling tray, take the color on the ground sample, then we cup, adjust, and do it again. The meters don't tell us whether a coffee tastes good. They take the guesswork out of whether a coffee meets our intended targets before we have the opportunity to cup, so we waste less coffee and time getting to the version we want to sell. They also give us a baseline we can hold every coffee against; a consistent stylistic approach across our single origins, and a fixed profile for the blends that require a more consistent roast approach year round.

The three measurements

Each measurement tells you something useful, but also has its limits. Getting a moisture reading on the green tells us how readily a coffee will take on heat in the roaster, which shapes how we approach the curve, and on the green buying side, it tells us how long a coffee can sit in storage before it starts losing it's sparkle. Decent moisture meters read close enough to each other that the information transfers well regardless of the device you use, and if you don't have one yet your green importer should be able to share their readings. However, moisture doesn't tell us anything about flavor.

Roast loss tells us how far we took a coffee, and it needs no equipment beyond a scale you probably already own. To measure it, weigh your starting green coffee and the finished roasted coffee. The difference between the two is the roast loss. The catch is that two coffees roasted to the same loss on two different machines that require different approaches can land in very different places on the cupping table.

Color is the measurement the industry is currently fixated on, but it is the trickiest of the three to share across setups. This makes sense given language around color is how most of us communicate how roasted a coffee is. But using the number a color meter delivers when reading a sample to calibrate with other roasters using different equipment is a bit misguided. Color meters aren't standardized between brands, and two meters meant to read on the same color scale can read the color of a coffee differently. The reading also shifts with grind size, which sounds like a manageable variable, until you account for the specific grinder, the specific burr set, and the particle size distribution that grinder produces. Color meters get misused in a few other ways; some roasters chase the tightest possible outer (whole bean) to inner (ground) color spread, when a tight spread doesn't reliably translate to a better cup. A color reading only means something against a target you've set with your particular equipment, and that target has to come from a flavor decision you've already made. At Pastime, we found that grinding as fine as we can for our color readings because it gives us the most data stability batch to batch. The tradeoff is that fine grinding skews the Agtron scale's naming conventions: if a coffee is ground fine enough, it will read as a light roast, regardless of its roast profile 

Used the way it's intended, color is the most reliable reading we have on how much roast impression a coffee will carry. We dial in the loss target for a given coffee through cupping, then lock in the color reading. That color becomes our anchor for blend component swaps or new single origin offerings. The first color reading after a roast gives us a fast verdict on whether the coffee will hit our target flavor balance, be underdeveloped, or roasty. That's enough to know whether the first batch of a new coffee is something that hit our target, or whether we should load up another bucket and try again. The specific flavor profile still comes from the cupping table, the color reading just gives us a head start on what to expect when we get there.

In practice

For our single origins, we start by matching the loss percentage to the starting moisture percentage; if a green coffee comes in at 9.8% moisture we roast it to 9.8% loss.  The moisture to loss match tells us if we've landed in the ballpark of our targeted roast degree, but it can't tell us how the coffee released water along the way. That depends on water activity (how freely available the water in a bean is), which varies between coffees even at similar moisture readings. For example, we had a coffee come in recently at 11.5% moisture, matched to 11.5% loss it read 135 on the Agtron scale, a touch lighter than our 128 target for single origins. We took it to 12.5% loss to hit 128 and the cup balanced where we wanted it. From there, color on the ground sample is what closes the loop on flavor balance.

Blends follow the same framework, with a wider moisture to loss gap and a different color target for each one. Sure Thing starts at moisture plus 2.5% on the loss side and lands at 108 on the Agtron scale. This target yields a coffee that holds up as drip and pour over but still pulls a bright but balanced shot for customers who want a little more roast impression in the cup. 4:30am gets pushed further, with a color target of 95. The cup we're after has the sugars converted into notes of dark chocolate and brown sugar molasses, the acidity dialed back, the roast forward. It's a coffee developed mainly for espresso use that also brews well as drip for those who prefer a roast forward flavor profile that doesn't push into full on dark roast territory.

When to drop the roast

Loss and color are both measurements we take after the roast cools, which means the roaster-operator needs a different cue to know when to drop the coffee. We use post crack time: the time between first crack and the drop. This is the cleanest measurable signal we've found for ending a roast consistently. The operator watches the clock from first crack, drops at the target, and the loss and color land where they're supposed to. The only caveat is: this only works if the machine starts each batch from the same thermal state, which is what a between-batch protocol gives you. Without a consistent one, post crack time drifts with the heat history of the roaster, and your target on the clock stops meaning anything. Scott Rao coined the term, and his writing on BBP is worth reading directly.

Our roast targets

Here's what our roast targets look like operator side, with two of our blends and two of our single origins pulled in as examples. These are our numbers on our equipment, a Loring S15, and they won't drop cleanly onto another roastery's setup. The method, though, is equipment agnostic. The framework (track your green moisture, set a loss target, read color to confirm) works on any roaster, against whatever flavor balance you're after. There's a blank version of the template available to download if you want to use it for your own coffees.

Consistency across the menu

This framework buys us consistency across the menu, even when the coffees on it are roasted to wildly different places. A single origin at 128 and 4:30am at 95 are doing very different things in the cup, but the way we get them there is the same: green moisture as the starting point, loss as the control, color as the confirmation. Customers can taste the difference between coffees without tasting an inconsistency batch to batch in how any one of them was roasted. The bag you open in October won't taste identical to the bag from August, because the coffee itself changes through the season and we don't try to flatten that out. What stays constant is the approach - how far we take a coffee and how we find its flavor balance, even as the coffee itself changes as we move through the seasons.

The barrier to running a measurement driven program used to be the cost of the meters but that's not really true anymore. A moisture meter and a color meter together are probably the most useful additions you could make to your bench, and ours paid for themselves within the first few months. If you want to try it yourself, what follows is the whole approach compressed into a working reference.

Quick guide

• Equipment: a moisture meter, a color meter, and a scale that reads to the hundredth of a pound (without that resolution your loss percentages will be off).
• To test the framework before buying both meters: a scale plus moisture readings from your importer will get you about 80% of the way there. Add the meters when you want to commit further.
• Use post crack time as the operator's drop cue, and lock in a consistent between batch protocol. Without a consistent BBP, post crack time stops being a useful marker.
• Establish your roast degree target using the loss differential, which is your loss percentage above or below the starting green moisture percentage. We've found these ranges correlate with how general consumers identify roast degrees:
  • Loss below starting moisture to matched matched moisture: ultralight or extra light
  • Loss matched to moisture, up to +1%: Nordic light
  • +1% to +2.5%: light
  • +2.5% to +4.5%: medium
  • +4.5% to +6.5%: dark
  • Above +6.5%: French
• Treat this as a loose guide, with the specific target inside each range coming from your own cupping.
• Roast to that target, cup the result, and adjust until the cup lands where you want it. Take a color reading from your ideal version, and lock it in as the color target for that coffee and its use case.
• Color becomes your anchor for stylistic consistency. Each coffee on the menu, single origin or blend, gets its own color target that reflects how you want it to taste. When you add a new coffee, dial it in with the loss differential as your starting frame and color as the final anchor.
• Once a coffee is dialed in, lock in both the loss and color targets. Measure against both batch to batch for as long as that coffee is on the menu.

Thanks for sticking with us through what turned out to be a long, technical post. We hope it gives someone reading this enough of a frame to tighten up an existing roasting operation, or enough confidence to start one of their own. There are many ways to run a roasting program, from measurement driven programs like ours to more craft based ones that rely on the roaster's senses. There's no one right way, the right approach is the one that fits your goals. Wherever you fall on that spectrum, we hope this offers something useful for the way you craft your coffee.