We Logged 75 Production Batches Over Six Years. "Warmer Is Faster" Turned Out to Be Wrong.
By William Stuart, Founder & Head Brewer, Colorado Sake Co.
Nearly every sake resource online repeats some version of the same rule: ferment warmer and the moromi moves faster, ferment colder and it slows down, and that trade-off is the lever a brewer pulls. We have been brewing sake in our RiNo brewery in Denver since September 2018, and for six years we logged every tank. When we finally analyzed the whole corpus, the rule did not hold. Across our batches, warmer early fermentation correlated with slower ferments, not faster ones. This is what the data actually says, including the parts that complicate it.
What's in this study
The dataset: 75 batches, 1,250 readings
This is not a survey of other people's sake. It is our own production log, and it is the entire log, not a flattering slice of it.
- 75 batches spanning December 2019 through 2026, roughly six brewing seasons.
- 1,250 batch-day readings, temperature, ABV, SMV, pH, specific gravity, recorded tank-side.
- 52 addition events logged against those batches.
- Every batch class we run: our main sake line, the sparkling line, shubo starters, kasu, and seltzer.
If the vocabulary here is unfamiliar, moromi, shubo, koji, our sake field guide defines every term from the brewery floor, and how sake is made walks the full production sequence these batches follow.
It includes our failures. It includes a batch that finished at 7.4% when it should have finished near 14%, a ferment that appeared to lose 2.2% ABV between day 16 and day 19 (physically impossible, it was a sampling or entry error), and three genuinely stalled tanks. We left all of it in, because a dataset you have cleaned of your embarrassments cannot tell you anything.
The finding: warmer was slower
We tested the obvious hypothesis: does a warmer early fermentation reach target ABV in fewer days?
Across the full corpus, the correlation between early fermentation temperature and days-to-12%-ABV came out at +0.48. The sign is the important part. A positive correlation here means warmer batches took more days, not fewer. If the folk rule were right, that number should have been solidly negative.
It is not an artifact of mixing eras together, either. We checked within each era separately. In our 2020-era batches the correlation was +0.34, still the wrong direction for the rule. In our modern batches it was −0.07, which is to say: essentially nothing. Temperature was not the lever.
And the modern batches are the ones that finish fastest. They also run cooler early, around 9.1°C. Cooler and faster, at the same time. That is the opposite of what the rule predicts.
Why the rule looked true for so long
Here is the part we find genuinely interesting, and it is a lesson about brewery data rather than about sake.
The rule looked true because of a confound, and the confound is us. When a tank was sluggish, we warmed it up. Warmth was the response to a slow ferment, not the cause of a fast one. So in the logs, heat and slowness travel together, and if you read that relationship from the wrong end, you can talk yourself into either story.
This is a trap any brewer keeping records will fall into, because the intervention gets recorded right next to the problem it was trying to fix. A correlation in a production log is not an experiment. We had six years of numbers that felt like they supported the rule, and they never did.
What actually produced the speed was a change in method, made in 2024, in how we prepare and enzymatically support the mash, work that happens upstream of the tank, in the koji. That change roughly halved our time to 12% ABV, from about 21 days in the 2020 era to about 9 days now. Not heat. Method.
What six years of tightening looks like
The era-over-era numbers are the clearest thing in the dataset:
- Mean fermentation length: 25.8 days (2020 era) → 14.9 days (2024 onward).
- Mean finished ABV: 13.56% → 14.32%. Faster and stronger, which the folk model does not predict.
- Consistency (standard deviation of finished ABV): 1.51 → 0.77. Half as variable.
- Year over year, the spread collapses: ABV standard deviation of 5.37 in 2019, 1.06 in 2020, and 0.55–0.72 across 2024–2026.
- Batches landing in the target band: 50% → 91% → 100% in 2026.
A brewery getting better looks like variance shrinking. That is the signal we would point any young brewery toward: not your best batch, but the distance between your best and your worst. The finished result of that tightening is what goes into the bottle, our flagship American Standard comes off this line.
The exception that proves it is a real effect
There is one place where warm-and-fast does appear to hold, and it is worth noting because it suggests the finding above is not simply us failing to detect a temperature effect.
Our sparkling line is intentionally run warm and fast, and it behaves exactly as you would expect: one 2026 sparkling batch reached 14.5% ABV in six days, peaking at 14.3°C. Warm, quick, on purpose.
So temperature does do something. The point is narrower and more useful than the folk rule: on a cold-fermented sake line, heat is not the speed lever, and reaching for it when a tank is sluggish is treating a symptom. Do not transfer the sparkling result to the main line. We have the logs to show what happens when you try.
What this does not prove
We would rather state the limits ourselves than have them pointed out to us.
This is observational production data, not a controlled experiment. We did not run matched tanks at different temperatures with everything else held constant. Nobody with a working brewery and a payroll does. The confound we identified above is exactly the kind of thing that hides in data like this, and there may be others we have not spotted.
Our modern sample sizes are small, three batches in 2024, five in 2026. The era effect is large enough that we are confident acting on it, and we have changed how we brew because of it. The finer-grained breakdowns are directional at best.
And this is our rice, our water, our altitude, and a Denver brewery's tanks. We would be glad to be contradicted by someone else's numbers. That is rather the point of publishing ours.
Frequently asked questions
Does warmer fermentation make sake faster?
Not on our cold-fermented sake line. Across 75 batches, warmer early fermentation correlated with slower ferments (correlation of +0.48 against days-to-12%-ABV). Our fastest modern batches run cooler, around 9.1°C. The exception is our sparkling line, which is deliberately run warm and fast.
What temperature does sake ferment at?
Our main sake line runs cold, roughly 7–9.5°C in early fermentation. Sake ferments colder and longer than a typical beer fermentation, which is a large part of why it develops the aromatics it does. Our sparkling line is the deliberate exception and peaks warmer.
How long does sake take to ferment?
In our brewery it has ranged widely. Our 2020-era batches averaged 25.8 days; our 2024-onward batches average 14.9 days. The change came from method, not temperature.
If temperature is not the speed lever, what is?
For us it was a change in how we prepare and enzymatically support the mash, introduced in 2024. It roughly halved our time to 12% ABV, raised mean finished ABV from 13.56% to 14.32%, and cut batch-to-batch variability by half.
Is this a peer-reviewed study?
No. It is a working brewery's production log, analyzed honestly and published with its limits stated. It is observational, not a controlled experiment, and our modern sample sizes are small. We are publishing it because almost nothing like it exists in English, and we would rather be argued with than ignored.
Why we are publishing this at all
Almost everything written about sake fermentation in English is a restatement of a restatement. The primary sources are in Japanese, are decades old, or sit in breweries with no reason to publish. So the same handful of rules get copied forward, and "warmer is faster" is one of them. Even the basics get muddled, people are still sorting out whether it's called sake or saki and whether sake is really a rice wine at all.
We are a working American sake brewery with six years of logs, and we found that one of those rules did not survive contact with our data. That seems worth saying out loud. If you are brewing sake, commercially or in a bucket in your kitchen, we would rather you argue with our numbers than inherit a rule nobody checked.