Ask a panel how much they liked a prototype and you will get a number between one and nine. It will be somewhere near the middle, because most prototypes are. You now know your product is mediocre. You do not know why, and you cannot brief a formulator with it.
What JAR scaling does differently
Just-About-Right scaling asks about each attribute separately, and asks the question directionally: is the sweetness too weak, just about right, or too strong? Same for flavor intensity, mouthfeel, aroma, aftertaste.
The output is not a grade. It is a distribution per attribute, and distributions have shapes worth reading. An attribute where most respondents cluster on “just about right” is finished. An attribute where responses pile up on one side has a direction. An attribute where responses split into two piles at opposite ends is the interesting case — it usually means you have two audiences, not one problem.
Penalty analysis: which flaw is actually costing you
Not every deviation matters equally. Penalty analysis pairs each attribute deviation with how much overall liking drops among the people who reported it. That turns a list of imperfections into a ranked list of costs.
This routinely reverses intuition. A prototype can have a large proportion of people saying the color is off, and almost no liking penalty attached to it — while a smaller group reporting a lingering aftertaste is dragging the mean down hard. Without the penalty step, a team optimizes the loud complaint instead of the expensive one.
Why blinding is not optional
Sensory response to a branded product is partly response to the brand. If the panel can see the packaging, you are measuring the sum of the formulation and the expectation, then attributing the whole result to the formulation. Blinded evaluation is what makes the attribute data about the product.
What a usable output looks like
A sensory read that a formulation team can act on has three parts: the attribute profile with directions, the penalty ranking so the work has an order, and a clear recommendation — proceed, optimize, or reformulate — rather than a page of charts and an invitation to interpret.
The recommendation matters because it forces a decision. A panel that ends in “here is the data” tends to end in another meeting.
Where this sits
This is R&D instrumentation. JAR outputs are for changing the product; they are decision signals, not claim substantiation. When the goal is a statement that will appear on a label, the work moves to a prospective consumer study with a prespecified endpoint and an analysis plan agreed in advance.
This note describes curí’s own laboratory practice and is written for brand and regulatory teams. It is not regulatory advice on a specific product. All programs are designed to support structure/function claims permissible under DSHEA; curí does not design studies to support disease claims.