A personal note on timing: I proposed this in 2024 and kept building after the “no”
In 2024, I submitted a structured Scope of Work to a dairy company proposing an Integrated Dairy Tactical Planning & Milk Valorisation Decision Support System, a single planning layer between commercial demand (forecast, promotions, customer service rules) and operational reality (milk availability, milk collection and composition, line capacities, standardisation constraints, shifts, CIP windows, cold chain, shelf life). The core argument was explicit: in forward dairy, the operational objective is not “optimal production” in the abstract, but a repeatable weekly equilibrium while allocating incoming milk not only as volume but as nutrient carriers (fat/protein/casein) in an economically optimal and operationally executable way.
The company did not engage at the time. And this is the part that matters to me: I treated that rejection as an investment horizon, not a dead end. I continued to develop the concept into a complete, executable scope of work, and I built a team ready to deliver a project of this class, because in dairy, there are ideas you pitch and capabilities you compound until market conditions force adoption.
My writing today extends that original framework to cheese and dairy ingredients, where the economics are even more structural, more interdependent, and far more sensitive to planning quality than most organisations admit.
The central misdiagnosis in cheese & dairy: treating milk as “litres” instead of a constrained resource system
A cheese & dairy plant is not merely a set of process steps. It is a resource allocation system operating under interrelated constraints: milk composition variability, standardisation requirements, co-products and byproducts, capacity bottlenecks, storage/ageing space, inventory policies, and contractual demand.
In the academic and applied operations research tradition, this is described as milk valorisation: the optimal allocation of raw milk to the most profitable product portfolio while respecting all constraints that define feasibility in real-world settings.
The key point is that, structurally, dairy production is interrelated: byproducts from one process are inputs to another, and a change in one product plan creates second- and third-order effects elsewhere in the network. A comprehensive mid-term model is needed precisely because these interrelations make “local optimisation” (plant by plant, product by product, department by department) systematically unreliable.
Cheese is a two-product business, whether you admit it or not: the cheese whey coupling
The moment you talk about cheese, you are implicitly talking about whey.
Whey is produced in large volumes, is nutritionally rich, and has a high environmental impact when poorly managed, making valorisation a central lever for economic and sustainability. A proper framing from the literature: for many cheese types, a “waste index” expressed as a cheese–whey mass ratio can range roughly between 4 and 11.3, underscoring the scale of whey streams relative to the primary product.
This is why “cheese planning” without explicit whey planning is incomplete decision-making. My base work formalises a critical question that most dairies leave unmodelled: if you increase cheese output because whey margins are attractive, what happens to the optimal volumes of the rest of the portfolio? In other words, why economics can and should change the optimal cheese plan, not just monetise its byproduct after the fact.
Stepwise whey valorisation vs integral valorisation
One of the strongest operational insights in the research is the difference between:
- Stepwise valorization: plan “main” dairy products first, then separately decide what to do with whey; versus
- Integral valorization: optimise milk and whey simultaneously as one integrated portfolio problem.
This is not semantic. It changes which products are “optimal” to produce under capacity, demand, and price conditions because dairy flows are interdependent.
The planning level that actually moves value: tactical (mid-term) integrated allocation
A recurring mistake in dairy organisations is confusing operational scheduling (today/tomorrow) with tactical planning(weeks/months) and hoping the former can compensate for the lack of the latter.
The referenced research distinguishes strategic, tactical, and operational levels and argues that major gains for dairy processors can be achieved by strengthening mid-term tactical planning, especially around production planning, while incorporating necessary supply and demand inputs.
That is precisely the architecture I proposed in 2024: a tactical decision layer that produces stable weekly plans and reduces “plan nervousness,” while still enabling daily execution to remain disciplined rather than reactive.
What the Decision Support System must do in cheese & dairy: model reality, not wishes
A serious DSS for dairy valorisation is not a dashboard. It is a decision engine that must be both comprehensive and comprehensible, complete enough to reflect interrelations, but transparent enough that executives, planners, and plants can trust and act on it.
The non-negotiables in the model core
If you want the model to be credible in a cheese + ingredients environment, you need at least five foundations:
- Recipes (process yields) and flow representation
The model must preserve the correct relationships between produced volume, required inputs, and generated byproducts. - Composition balances (fat/protein/casein/dry matter and, when relevant, lactose)
Component pools constrain milk allocation decisions. The research explicitly specifies composition balance constraints so that the total fat/protein/dry matter in inputs equals the total in outputs (main products plus byproducts), and extends this logic to lactose where relevant for whey-derived products. - Capacity constraints across resources, locations, and time
Plants are not infinitely flexible. Capacity, changeovers, and bottlenecks determine feasibility and should be reflected in constraints, not “fixed later” by heroic execution. - Demand constraints with commercial structure (tranches, contracts, and price-volume logic)
Dairy markets are not linear. The modelling approach incorporates tranche logic (volume bands with tranche-dependent prices), which is far closer to commercial reality than flat pricing. - A profit objective that accounts for revenue minus relevant cost drivers
The linear programming dairy valorization model is designed to maximise profit while accounting for production, transport, and purchase costs—exactly the structure needed for a tactical allocation tool.
In short: this is not “analytics.” It is Operations Research applied to a dairy flow network.
Why whey changes the economics more than most plants admit
Beyond the structural coupling, the research quantifies something operational leaders should internalise: explicit whey valorization can materially alter profit outcomes.
In scenario analysis using an Integral Dairy Valorisation Model, the study reports sizeable profit deltas when whey byproducts are explicitly valorised, with reported average monthly percentage differences in profit between cases and a substantial share of profit attributable to whey processing.
The managerial implication is blunt: if your cheese plan does not explicitly co-optimise whey, you are likely running a portfolio that is locally efficient but globally suboptimal.
The part people underestimate: organisational economics and decision rights
Even perfect math fails under poor governance.
Large dairy groups suffer from decoupled objectives across operating companies and functions (production, inventory, distribution, sales), which can lead to “optimal” local behaviour that is destructive to integrated valorisation. The research explicitly frames integrated planning as a response to this organisational fragmentation and emphasises measurement of performance at multiple levels, from solution feasibility to the efficiency of planning actors.
This is why my 2024 proposal was never “just build a model.” It was: build a model and embed it into an operating rhythm planning calendar, freeze/slush rules, escalation logic, and KPI contracts that prevent the plan from being renegotiated daily.
What “ready to implement” means in practice: the scope I kept building after the rejection
When I say I can implement this now, I mean the work is already structured into a delivery architecture that respects how dairies actually operate.
Phase 1 Diagnostic and blueprint (decision engineering, not process theatre)
- As-is decision mapping: where decisions are made, with what data, and with what incentives
- Data readiness: product master, recipes/yields, composition data, capacity parameters, inventory policies, demand constraints
- To-be integrated planning design: governance, roles, weekly cadence, KPI framework
This aligns with the research emphasis on comprehensive yet comprehensible modelling, and on the importance of the model as a communication and early-warning tool, not merely an optimiser.
Phase 2 MVP DSS (tactical model + decision packs)
- LP/MILP core depending on discreteness requirements
- Scenario library (base, promo spikes, capacity loss, composition shifts)
- Robustness and “plan stability” measurement, recognising uncertainty in key inputs such as milk supply and composition
Phase 3 Pilot, adoption, and industrialisation
- Pilot on a defined portfolio slice (e.g., cheese family + whey products + a constrained milk pool)
- Expand across plants/OpCos and product categories once trust and governance are established
- Synchronise tactical planning with short-term scheduling and, where relevant, long-term capacity investment logic
The “investment thesis” behind all of this: capability compounding beats timing
There is a principle I rely on in both strategy and operations: you invest in certain capabilities before they look urgent, because once they become urgent, the organisation pays a premium in chaos.
Dairy is entering a regime in which volatility in demand, competition, regulation, and composition uncertainty makes integrated planning less a “nice-to-have” and more a requirement for stability and margin defence.
That is why I kept building after the rejection: not to win an argument, but to ensure that when the moment arrives when complexity overwhelms manual planning, the solution is not a concept. It is a deployable system with a team behind it.
Closing position
If you want to manage cheese and dairy ingredients like a portfolio business, you need an engine that speaks the language of reality: composition balances, interrelated flows, capacity, demand structure, and economic objectives executed at the tactical planning level and governed like a management system.
That is what I proposed in 2024. It did not convert then. It is implementable in 2026…