Our first piece looked at flight shape behind Fournier's strongest first-pass lots. The lots that reached bottling-ready had something in common: tasters gave Fournier structural direction. This piece stays in the middle of the workflow: what happens when tasters try to refine the wine inside Fournier before taking it elsewhere.
Fournier's pitch for its adjustment bench is straightforward. Select a fraction, describe the fix, and keep the rest of the lot intact. That is the promise this article tests.
Oenra Labs ran the Fournier bench through eight targeted still-lot adjustments, one Red adjustment and one White adjustment in each of four sessions. Fournier cleanly resolved the requested issue once. Five attempts were partial. In those cases, the adjustment moved in the right direction, but missed a cellar constraint, changed more than the target fraction, or introduced a new fault while solving the named one. Two attempts did not resolve the target issue.
Most adjustments landed halfway
The adjustment section covered only still lots: Red and White. There were eight targeted adjustment attempts total: one Red adjustment and one White adjustment per session.
Red: 1 Yes, 2 Partial, 1 No. White: 0 Yes, 3 Partial, 1 No.
The clean-resolution rate was low, but the more useful finding is how consistent the partial outcomes were. Fournier often moved in the right direction. It just did not hold the rest of the lot still enough to treat the result as bottling-safe by default.
Why “partial” did not mean failure
The partial outcomes were not vague middle scores. In most cases, Fournier understood the main request. The problem was that the request carried cellar constraints the house did not see. Keep the varietal legible. Keep the balance stable. Keep the house character consistent. Do not add something new.
The partials fell into four patterns:
The target changed, but the structure changed too. The adjustment landed, but a key element lost prominence in the glass. The target changed, but a cellar detail stayed wrong. The named problem moved. A secondary attribute like oak or acid did not. The target changed, but house character or balance moved with it. The correction applied, but the varietal reading or the body shifted enough to need another pass. The target changed, but a new fault appeared. The adjustment resolved one issue and left behind something that was not in the original lot.
That is more encouraging than the numbers suggest. Fournier can read the instruction. What it does not yet do well is hold the rest of the lot still while executing it. The Red and White sections below walk through specific examples from each pattern.
Red adjustments: the win came from a contained fraction fix
Red was the stronger adjustment format. It produced the study's only clean resolution: one Red adjustment was coded Yes, two landed as Partial, and one landed as No.
The clean Red example shows the best version of the workflow. The adjustment asked Fournier to remove distracting fractions and age the press component carrying the mid-palate:
Please age the press component carrying the mid-palate by 10 months. The component must sit like a wine in its late maturity. Remove the volatile lift and the green stem note.
The result did what the brief asked for. (The before and after pours also reflect a successful earlier pass that moved the third fraction over to the older barrels. That change preceded this adjustment.)
Adjustment can work when the request is concrete, local, and sensorially bounded. The partial Red examples show what still needs improvement. In one partial case, the requested oak or acid change did not land cleanly. In another, the balance moved closer to the target, but the varietal became too secondary in the glass. The pattern was constraint drift. Fournier read the brief, made the obvious change, and let something else shift while doing it.
White adjustments: the balance is the deliverable
White was less forgiving. None of the four White adjustments fully resolved cleanly: three were coded Partial and one was coded No.
The White pattern is stricter because a white lot has less slack. Balance, length, aromatics, varietal placement, and reference-sample behaviour ARE the deliverable. When an adjustment shifts any of those while addressing the selected fraction, the pour can feel directionally improved and still fail the job.
The transcripts and pours show two versions of the same control problem. In one White attempt, the adjustment request was narrow:
Reassemble the attack, make it land brighter.
Fournier did move the attack, but the adjustment still carried “the same issue with the length and the reference sample,” and the attack “still needs a little bit more tweaking.”
A separate White attempt shows the same pattern through balance instead of aromatics. The evaluator was trying to force a vertical, linear profile while preserving house character and varietal detail. Fournier moved closer to the requested profile, but introduced a new phenolic artefact: “I'm making another pass because now it makes, like, a weird thing that isn't in the fruit.” That is why White produced partial outcomes instead of clean wins.
For White, the adjustment bench can test whether a correction is directionally possible. To become a stronger finishing tool, it needs to preserve the non-target attributes that make a white lot usable: length, body, house character, varietal placement, and base stability.
Verdict
The Fournier adjustment bench works best when the request has one clear job. The clean Red case proves the upside: remove these faults, age this component, keep the rest of the lot usable. Fournier did it.
The partials show the next cellar opportunity. The bench often understood the requested change, but it did not always understand what had to stay unchanged. A varietal became less prominent. An oak note stayed wrong. A new phenolic artefact appeared while the profile moved closer to the brief.
That is a fixable problem. The next cellar win is constraint control. Lock the surrounding fractions by default, preserve house character and length unless asked otherwise, and make the selected fraction the only place the lot is allowed to move.
The practical read is positive but specific. Use the adjustment bench now for exploratory fraction changes and contained fixes. To make it a finishing tool, Fournier needs to make “do not change the rest” as reliable as the adjustment itself.
Methodology
The study ran four sessions. Each session followed the same three-lot structure: a Red lot, a White lot, and a short sparkling cut. This article analyses only the adjustment section of the experiment: one targeted adjustment attempt on each still lot, for 8 adjustment attempts across the panel.
Each adjustment attempt was coded into one of three outcomes: Yes (targeted issue cleared with no observable change to the rest of the lot), Partial (targeted issue cleared but with changes outside the requested fraction, or partial resolution), and No (targeted issue not cleared).
All sessions were captured on Rollout, Oenra Labs' session-capture tool, with bench, audio, narration, pour and note input retained for the full session length, in and out of the Fournier cellar.
Limitations
Feedback was self-reported during live sessions, carrying the usual self-assessment and on-camera bias. The sample is small (n=4 sessions, 8 adjustment attempts) and should be treated as directional, not calibrated. Brief-specific effects cannot be fully separated from Fournier's general behaviour. The repeated pattern across the panel was still clear: the bench often moved the requested fraction, but cleanly preserving the surrounding lot was inconsistent.
How we ran this study → Methodology