Reducing sugar without losing taste
You have been asked to take out 30% of the sugar, remove palm fat, change supplier or drop an ingredient from the list. The formula can be changed; what you cannot do is guess how far you can go before the consumer notices.
Reformulating means finding the point where the saving is not yet paid for in preference. That point exists and it can be measured. And it is almost never where intuition says.
Noticing a change is not the same as minding it
A reduction project turns on two different thresholds, and half the saving usually sits between them. Most projects stop at the first one believing it is the second.
- The detection threshold
- The point where someone starts telling the reduced version from the original. It is crossed early — a good deal earlier than the project meeting tends to assume.
- The rejection threshold
- The point where that difference stops being a curiosity and starts costing you preference. It sits further out, and it is the only one of the two that really limits the formula.
Neither of them is known in advance, and neither carries over from another category: they have to be found in your product, and testing the level you were handed is not enough for that.
Test the whole staircase, not one step
An isolated target can only give two answers, and both are bad: either you reach it and do not know how much more there was, or you miss it and do not know where the limit sat.
Which is why a reformulation is a sequence and not a test. Five decisions, and the first one is taken before the formula is touched.
What you are willing to lose
How much drop in liking or preference is acceptable in exchange for the saving. That is a business decision, not a laboratory one, and without it no result can say whether a step passes.
The staircase, not the target
If the brief says 30%, do not test only 30%: prepare several levels above and below. Testing the target answers yes or no; testing the staircase tells you where the ceiling is.
Where it starts to show
Every step against the original, blind and in balanced order. A discrimination test settles it quickly because it only asks for a binary answer: told apart, or not.
Discrimination testing →Whether it actually matters
The steps where the difference already shows go to consumers. That is where you see whether the perceived difference turns into less liking, the same purchase intent, or nothing at all.
Product test formats →The finalist, outside the test room
The chosen level is tested in real, repeated use, because one sip side by side is nothing like a week in the fridge. What each field setting sees is explained on its own page.
Home use test and CLT →
A product can be told apart from the original and still be liked just as much.
Which is why stopping the reduction the moment a test says the difference shows is expensive: you give up a stretch of saving nobody was going to claim back. The limit is set by the rejection threshold, never by the detection one.
Knowing what moved so you can fix it
Taking out sugar does not only take out sweetness: it changes body, shifts the perception of acidity, unbalances the aroma and sometimes alters colour. What has to be recovered is almost never the sweetness.
Without that reading, the default correction is to add something that brings the sweetness back, and the project inherits a new problem instead of solving the one it had.
A descriptive profile measured before and after on the same scale is what tells you which attribute moved, and by how much. Sensory profiling covers it in full.
What comes out of the project
- The ceiling
- How far down you can go without anyone telling the difference, and how far down you can go knowing it can be told apart and does not cost preference.
- The cost of each step
- The drop in liking or purchase intent that goes with each level, so the saving can be negotiated against it instead of argued blind.
- What to compensate
- Which attribute moved, in which direction and by how much. That is what steers the next iteration of the formula instead of leaving it to trial and error.
- The test at home
- Whether the chosen level still works once it stops being a sample and becomes the product sitting in the fridge.
One case told in full: Product relaunch with reformulation →
What do you need to reduce?
Tell us the product, which ingredient is going down or coming out, and how much you have been asked for. We will tell you how many steps it takes, which test each one needs and in what order, so you do not pay twice for the same data.