The end of static recipes
Cocoa more than doubled in 2024. Butter, coffee and olive oil followed comparable paths. Against such moves, the static recipe — computed once, locked, revised every three years — is no longer a guarantee of stability: it is a scheduled margin haemorrhage or a silent quality downgrade.
The alternative is not panic reformulation at every spike, but a shift to dynamic formulation: recipes designed to be re-optimised continuously, within defined quality bounds.
Commodity volatility, the new regime
Erratic climate, geopolitical tensions, speculation, regulatory transitions: the causes stack and reinforce each other. Price cycles that used to unfold over years now play out in months. A recipe whose material cost drifts 30% between two annual reviews leaves two equally bad options: absorb it in margin or degrade in silence.
The dynamic recipe
A dynamic recipe is not an unstable recipe. It is a recipe whose degrees of freedom are explicit and bounded: acceptable dosage ranges per ingredient, qualified substitutes with their conditions, untouchable invariants — target sensory profile, claims, regulatory constraints, customer specifications.
Within those bounds, the exact composition can adjust to the cost context, the way a financial portfolio rebalances without changing strategy.
Continuous optimisation
The process becomes a loop: material costs are refreshed continuously from supplier data; the gap between current and target cost is monitored per recipe; when a threshold is crossed, a re-optimisation is triggered, evaluated, validated by R&D and traced. The annual review gives way to permanent steering — with humans deciding, not calculating.
What AI simulates
The combinatorial work belongs to the machine: exploring the thousands of dosage and substitute combinations that respect the bounds, recalculating for each one the cost, nutritional values, Nutri-Score and ingredient list, and presenting the three or four genuinely interesting options with their trade-offs. R&D decides on quantified scenarios instead of rebuilding each option by hand.
Food Scientist's Corner
In chocolate, playing with the cocoa mass/butter ratio within the bounds of the 'chocolate' denomination (Directive 2000/36/EC) offers a real but narrow lever — AI computes it to the decimal point. In ice cream, sugar/glucose/fibre reformulation must preserve freezing point and total solids: two constraints every scenario must respect by construction. In sauces, the starch/tomato/oil trio optimises under viscosity and pH constraints.
Future outlook
The static recipe belongs to a world of stable costs that is not coming back. Manufacturers who tool up for continuous re-optimisation turn volatility into an edge: they protect margins while others endure, without sacrificing the quality their brand stands on.
That is precisely what crumble-ai enables: multi-level recipes recalculated in real time, what-if scenarios on demand, compliance and labelling checked on every variant, every decision traced. The recipe stops being a document: it becomes a living system.
Is your portfolio ready?
See how crumble-ai continuously audits your recipes and blocks non-conformities at the source.
Book a demo