Most juice blends at most bars are built on instinct. Someone in a kitchen decides that equal parts sounds fair, and equal parts is what ends up in the bottle. We wanted to do this differently with our Kii Yuzu and White Peach blend, so we ran a structured tasting study across three weeks, logged every ratio, and treated the whole thing like a small sensory science experiment. The results surprised us, specifically how badly a 50/50 split performed, and how narrow the window of genuinely good ratios actually is. This article is a full account of the methodology, the measurements, and the failures, because we think the process is as interesting as the product.
Why yuzu is not just another citrus ¶
Yuzu is not a lemon and it is not a lime. Its aromatic profile sits somewhere between grapefruit, mandarin, and a faintly floral note that food scientists sometimes describe as terpenic. The dominant volatile compounds are limonene and gamma-terpinene, but yuzu carries an unusually high concentration of yuzu lactone, which gives it that almost medicinal edge at high volumes. In practical terms this means yuzu behaves aggressively in a blend. A small increase in its proportion does not produce a proportionally small increase in citrus character — it tends to tip the whole drink into something abrasive. This is the core design problem the Kii study was trying to solve: yuzu is not linear, and any ratio framework that treats it as linear will fail.
How we set up the 22-ratio framework ¶
We started at 10 percent yuzu juice to 90 percent white peach nectar and moved in five-point increments up to 60/40 yuzu-dominant, which gave us eleven primary ratios. We then ran a second pass filling in the gaps at 2.5-point intervals in the range that looked most promising from pass one, which brought the total to 22 distinct blends. Every sample was prepared fresh each morning by our production lead Tomoko Arai using the same cold-press yield from a single batch of Kochi Prefecture yuzu and a single-origin Wakayama white peach nectar. Brix was measured with a handheld refractometer at intake, and each blend was chilled to exactly four degrees Celsius before evaluation. We used blind triangle tests for the main panel and a ranked preference ballot for the secondary group of eight regular customers who had agreed to participate across three separate sessions.
What 50/50 actually tastes like, and why it failed ¶
The 50/50 result is the easiest to explain and the most counterintuitive. At equal parts, neither ingredient wins. What you get instead is a kind of sensory interference: the yuzu lactone note clashes with the high-ester sweetness of ripe white peach, and the result reads as muddled rather than balanced. Four of the eight panelists in session two independently used the word flat. Two said it reminded them of a cleaning product, which is a reasonable description of yuzu lactone at volume when it is not anchored by enough sugar or fat. The Brix on the 50/50 blend was 12.4, which is not low, but the acidity was pulling the perceived sweetness down significantly. A titratable acidity test run on session-three samples put the 50/50 at 1.38 percent citric acid equivalent, the highest of any blend we evaluated.
The range that actually worked, and what the data showed ¶
The viable window turned out to be narrower than anyone on the team expected: roughly 22.5 to 32.5 percent yuzu, with peak preference scores clustering tightly around 27.5 percent. At that ratio the yuzu contributes aroma and brightness without overwhelming the peach's mid-palate body. Panelist preference scores — recorded on a nine-point hedonic scale — averaged 7.4 out of 9 at 27.5 percent yuzu, compared to 5.1 at 50/50 and 4.8 at 60/40. Below 20 percent yuzu the blend was pleasant but directionally plain: peach juice with a vague citrus lift, nothing worth naming. The 27.5 percent formula also had the most stable Brix-to-acidity ratio over 72 hours of refrigeration, which matters for batch production consistency.
Three specific failures worth documenting ¶
The 35/65 blend looked promising in the first pass but collapsed in the second. When we filled in the 32.5 and 37.5 data points it became clear that 35 percent sat in a local dip in preference scores — not bad enough to stand out immediately, but consistently ranked lower than its neighbors. We still do not have a clean explanation for this. The 15/85 blend smelled excellent and scored poorly on taste, which is a useful reminder that aroma evaluation and full-palate evaluation diverge more than you expect. And the 60/40 blend, the most yuzu-forward we tested, produced what Tomoko described in her session notes as a grip on the back of the throat, which is probably the yuzu lactone concentration crossing a perceptual threshold. We stopped testing above 60 percent after that.
What we would measure differently next time ¶
We did not control for the time of day each panelist completed the evaluation, which is a real methodological gap. Taste sensitivity changes across the day, and citrus perception in particular tends to be sharper in the morning. Our sessions ran between 10am and 4pm, which introduced variation we cannot fully account for. We also relied on a small secondary panel of eight, which is enough to identify strong signals but not enough to say anything statistically robust about population preference. A proper sensory study would want at least 30 panelists across multiple sittings. The triangle tests were the more reliable instrument here, and they consistently pointed to the same narrow window, which gives us reasonable confidence the core finding holds even with these limitations in place.
How the final Kii formula was settled ¶
The production formula for Kii landed at 28 percent yuzu and 72 percent white peach, a slight rounding up from the 27.5 percent peak to account for natural variation in yuzu yield across the season. Kochi yuzu from late-harvest November fruit tends to run slightly less acidic than early-season fruit, so the formula will likely be revisited in spring when acidity climbs. Tomoko has proposed a seasonal adjustment protocol: a quick Brix and TA check on each new yuzu intake batch, with a pre-agreed adjustment band of plus or minus two percentage points. The goal is to keep the perceived balance constant even as the raw ingredient shifts. That kind of formula maintenance is unglamorous but it is the actual work of making a consistent product.
The Kii Yuzu and White Peach blend is now on the menu at Clay Crest Zone, and the full ratio log — all 22 entries including Brix readings and panel scores — is something we are happy to share with anyone who wants to dig into the numbers. If you run a juice program and have done similar testing, we would genuinely like to compare notes. The methodology here was imperfect, but the imperfections are documented, and that feels like the right place to start.