Understand Your Roasts

Every saved Batch comes with more than its Curve. Rostoc computes phase timings and weight loss for each roast, scores the Batch against its Reference Curve, and writes a short analysis of what happened. This page explains what those numbers mean and where to find them.

Phase breakdown and roast stats

Open a Batch and the key numbers sit at the top of the page: input and output weight with the weight-loss percentage between them, charge and drop temperatures, and the timing of each phase — dry end, first crack, and total duration. Duration runs from Charge to Drop. A Batch without both shows ”—”, with a prompt to mark Charge and Drop on the chart. See Batch details for the full page layout.

Two of these deserve a closer read:

DTR (Development Time Ratio) — the share of the roast that happened after first crack, as a percentage of total roast time. Roasters use it as a shorthand for how developed the coffee is likely to taste: a low DTR points toward underdevelopment, a high one toward roastier, more bitter cups.
Weight loss % — the moisture and gases driven off during roasting, as a percentage of the green weight. It tracks with roast level: darker roasts lose more.

Rostoc also works out a Thermal load (AUC) for every roast: the area under the Bean Temperature curve, in °C·minutes. Where DTR and the phase times describe when things happened, AUC describes how much heat the beans absorbed along the way. Two roasts can share a drop temperature and a total time yet still differ here if one ran hotter through the middle — AUC is the number that catches it. It is not among the numbers at the top of the Batch page; you read it in the consistency report.

How to read AUC

Rostoc measures AUC from dry end to drop, above a 100 °C base (it falls back to charge when a Batch has no dry-end marker). Both of those settings are stored with the value, because an AUC figure only means something next to another one measured the same way.

Treat it as a consistency check on your own machine, not an absolute “roast degree”: it integrates one probe’s reading, so it does not translate across different roasters, probe placements, or software. Compared batch to batch on the same setup, a drifting AUC is an early sign that something in your heat application changed even when the milestones look identical. When an imported profile file already carries an AUC value, Rostoc keeps that imported number on the side for reference and always scores against its own computed value instead.

The same stats appear on a Plan detail page for the Plan’s Reference Curve — charge, turning point, and drop temperatures, dry end and first crack times, duration, and DTR — so you can compare a Batch against the target it was roasted from. Hover the small ? markers next to a stat for a short explanation of what it measures.

Consistency score

Rostoc scores how closely a Batch followed its reference roast. The result is an Overall Score and a letter Grade, backed by a metric-by-metric report: DTR, drop temperature, drying and Maillard phase shares, weight loss, charge temperature, turning point, and thermal load (AUC), each compared between the Batch and its reference with the difference shown. Because AUC only compares meaningfully within one machine and probe, the report drops it automatically when the two roasts were measured differently — a large charge-temperature gap, or one Batch missing the dry-end marker — and says why.

The report also looks at the Curve itself: the average and peak Bean Temperature deviation from the reference, how closely the Rate of Rise followed the reference shape, per-phase RoR averages, whether an RoR crash or spike was detected, and the first-crack and total-duration differences. Every metric row has a ? marker explaining what it measures and what a typical range looks like — the score is meant to be understood, not taken on faith.

To open the report, view a Batch and press Cmd/Ctrl + I. Scoring needs a reference roast to compare against — assign one in the Batch edit dialog if the Batch does not have one yet. Use Rescore to recompute after you change the reference.

If the data looks off — for example, the Curves are misaligned rather than genuinely different — the report flags it with a data-quality warning so you do not act on a misleading score.

AI roast analysis

Alongside the score, Rostoc writes a short analysis of the Batch: a summary of the roast, a list of observations, and suggestions for the next one. It is generated automatically after the roast and appears in the Roast Analysis section of the same report — if it shows “Generating analysis…”, give it a moment.

Where the score tells you how far the roast strayed from its reference, the analysis tells you what that means and what to try next. To dig deeper or ask follow-up questions about a Batch, use the AI roast assistant.

Compare roasts

Compare puts two to ten Batches side by side. Their Curves are overlaid on one chart, each in its own colour, and lined up on a milestone you choose: Charge, turning point, dry end, first crack or drop.

On Batches, or in a Plan’s table of Batches, tick the checkbox of each roast you want to compare. A row’s checkbox appears when you point at the row. A bar appears at the bottom of the screen with the number selected.
Click Compare. It works with two to ten Batches ticked.

On the desktop the comparison opens in its own window. In the web app it opens as a page, and you can share its link with someone in your organisation. Batches and each Plan’s table keep their own ticks, and the Clear in the bar, or Escape, clears them. In the comparison, the Plan picker narrows the list of Batches you can add to one Plan, or Any plan; a comparison opened from a Plan starts on that Plan.

Below the chart, Metrics lines up each Batch’s numbers side by side, the milestone times among them, with the difference from the reference Batch of the set. The reference is named in the heading, for example Metrics vs RB-0016, and you can change it there. Roast colour is one of the metrics when you entered a colour reading on the Batch, with its scale named. A reading on another scale than the reference’s is left out and says why; a Batch with no colour reads a dash.

The Repeat score comes last, folded and marked Experimental. It measures repetition, not quality. Show opens it, with what was left out and the weight each measure carries.