A profile is approved in a beaker or a small batch, at ambient temperature, against a target everyone agrees on. It is then dosed into a product that gets heated, dried, extruded, retorted, or all four, and what arrives at the other end is not the thing that was approved.
The usual reading of this is that flavour was lost and the dose needs to go up. That reading can be incomplete: a higher dose may not restore a profile changed by selective loss.
A flavour is a population, not a substance
The single most useful thing to understand about a flavour is that it is not one molecule doing one job. It is a mixture of dozens or hundreds of volatile compounds, present at very different levels, contributing very different things.
Those compounds differ from each other in every property that matters here. They boil at different temperatures. Some are polar and sit in the water phase, others are not and sit in the fat. Some are chemically stable and some are highly reactive.
The character of the flavour is not the sum of them. It is the ratio between them. Change the ratio and you have changed the flavour, even if the total amount present is unchanged.
Heat does not remove them evenly
This is the whole mechanism.
When a product is heated, the volatiles do not leave in proportion to how much of each is there. They leave in order of how easily they leave. The light, low boiling, high impact compounds go first, and those are almost always the top notes: the fresh, green, citrus, aromatic, immediately recognisable part of the profile.
What stays behind is the heavier base: the roasted, cooked, earthy, sometimes slightly burnt end of the mixture.
So the product does not come out of the process tasting like a weaker version of the approved sample. It comes out tasting darker, flatter and more cooked, because the ratio has moved. That is a different flavour, not a diluted one.
Anyone who has smelled the exhaust from a dryer or the vent on a cooker knows exactly where the top notes went.
Heat does not weaken a flavour. It changes its ratio.
Which is why raising the dose fails
If the loss were even, doubling the dose would restore the profile. It is not, so it does not.
Doubling the dose doubles what survives, and the survivors are the heavy notes that were already over represented. A higher dose does not by itself restore the original balance and can overemphasise the notes that survived. It also costs twice as much.
This is the single most common mistake in the category, and it is made in good faith every time, because the instrument that reports the problem is a person saying it tastes weak.
Three escape routes, not one
Volatiles leave by more than one door, and the doors are not equally open in every process.
Evaporation. Heating a wet product is close to a steam distillation. Water leaving carries volatiles with it. Drying and any open cooking step are the heaviest losses in most processes.
Partition. A volatile that prefers fat will sit in the fat phase and stay there, and one that prefers water will leave with the water. This means the same flavour at the same dose behaves differently in a lean broth and a fatty sauce, and it is why a profile validated in one matrix cannot be assumed in another.
Reaction. Many aroma compounds are chemically reactive. They oxidise, and during heating they take part in the same reactions between amino groups and reducing sugars that create the cooked notes in the first place. So heat is simultaneously destroying part of the flavour and creating new flavour, which is why a heavily processed product often has notes nobody added.
A process that dries hard loses by the first route. One with a high fat content is dominated by the second. A long, hot, low moisture process is dominated by the third. Naming which one is in play is most of the work.
And then there is the shelf
What survived processing keeps leaving afterwards, through the packaging, more slowly, and it keeps oxidising while it waits.
So there are two clocks running, not one. Process loss happens in minutes and is the same for every batch. Shelf loss happens over months and depends on the pack, the temperature and the light. A product that is correct on the day of production and wrong at the end of its declared life has a shelf problem, not a dosing problem, and the two need different answers.
The matrix changes the flavour without changing the flavour
Here is the part that catches formulators who have already understood everything above.
How much aroma reaches the nose while eating is not decided only by how much is in the product. It is decided by how readily the matrix releases it. A thicker, more viscous product holds volatiles back. Fat binds the fat soluble ones and releases them slowly. Salt pushes them out of the water phase and makes them more available. Starch and hydrocolloids both raise viscosity and both reduce release.
The practical consequence is uncomfortable. Adding body can change perceived flavour at the same dose. Reducing salt changes taste and may change aroma release; the direction and size of the effect depend on the compounds and the food matrix.
Nobody changed the flavour. The flavour changed.
Three escape routes, and which process opens which
What this makes the job
If all of that is true, then the target cannot be a percentage in a recipe.
Specify at the point of eating. A dose is an input. The profile in the mouth, in the finished product, prepared the way a consumer prepares it, is the thing that has to be held.
Test in the real matrix and the real process. A profile approved in water tells you very little about the same profile in a fatty stew that has been retorted.
Choose the form for the process, not for the price. Liquid, powder, plated on a carrier and encapsulated forms have completely different survival profiles. An encapsulated flavour is protected through the heat and released at eating, and it exists precisely because of everything described above.
Expect the dose to be process specific. Two customers making a nominally identical product on different equipment will need different doses to reach the same profile, and neither of them is doing it wrong.
None of this is achieved by a single ingredient chosen from a list, which is why it is a system rather than a material.
If a profile that was approved on the bench is not what comes out of your process, tell us what you are making and how it is processed.
