From supply chains to supply networks
The word 'chain' says it all: a linear sequence where every link depends on the previous one, and where a break anywhere stops everything. The crises of recent years exposed the model's limits — and the answer is not 'more stock' but 'network': several origins, several suppliers, several possible paths for every ingredient function.
A supply network, however, cannot be decreed by procurement. It is designed at formulation time, because the recipe decides what the product depends on.
Beyond the chain
Thinking in networks means reasoning in functions rather than references: the product doesn't need 'supplier A's starch', it needs a thickening function with a given viscosity and heat stability. Once the function is specified, several raw materials and several origins can fulfil it — and sourcing gains degrees of freedom that logistics alone can never create.
Ingredient dependency, measured
Step one: make dependencies visible. How many recipes rest on this raw material? This supplier? This origin? A well-kept multi-level bill of materials answers in seconds; scattered spreadsheets answer in weeks — usually during the crisis, when it is too late.
The useful indicators are simple: number of SKUs exposed per raw material, share of revenue depending on a single origin, existence of a qualified substitute. Few manufacturers can produce them on demand.
Network optimisation
Once dependencies are mapped, optimisation becomes a multi-criteria trade-off: cost, risk, carbon footprint, lead times, regulatory requirements (EUDR, declared origins). Qualifying a second origin costs trials and regulatory time; not qualifying it costs exposure. The trade-off must be made recipe by recipe, starting with the highest-stakes SKUs.
What AI recommends
On a structured repository, AI turns the map into an action plan: it spots risk concentrations, proposes plausible substitutions from functional properties and your trial history, simulates the cost-nutrition-labelling impact of each option, and ranks the qualifications to launch. Work that used to occupy a crisis cell for weeks becomes a tooled monthly review.
Food Scientist's Corner
Network substitutions are won in the lab: a replacement emulsifier changes the crystallisation kinetics of a spread; gelatine from another origin doesn't share the same bloom; two lecithins behave differently in chocolate work. Every qualification trial — successful or not — documented in the repository becomes a fallback path available to the whole company.
Future outlook
The linear supply chain optimised a stable world; the supply network optimises an uncertain one. The shift is above all a data shift: without a product repository linking recipes, materials, suppliers and origins, the network remains a concept.
crumble-ai provides that foundation: multi-level bills of materials, dependencies visible in one click, what-if scenarios and AI-assisted substitution suggestions. Resilience stops being an exceptional project and becomes a property of the portfolio.
Is your portfolio ready?
See how crumble-ai continuously audits your recipes and blocks non-conformities at the source.
Book a demo