The problem

TINE Brumunddal faces a classic operations challenge: fluctuating raw-milk supply, limited processing capacity and a fragmented farm-collection network. One step — the DA-101 pasteurisation line — is the bottleneck, and some products use scarce bottleneck time without earning competitive margins.

Rather than asking “how do we produce more?”, we asked “what is the true constraint, and what creates the most value per bottleneck hour?”

01 — Product mix with linear programming

I led the optimisation model. Production decisions were formulated as a linear program solved with HiGHS via SciPy: maximise total contribution margin subject to line capacity, supply and process constraints.

• Ranked products by contribution margin per bottleneck hour
• Identified low-margin products (e.g. mellommelk) consuming DA-101 capacity
• Reallocated capacity to higher-value products

Result: about +9.7 % daily contribution margin — roughly 29,625 NOK per day — a textbook Theory of Constraints win.

02 — Route optimisation

Milk collection routes were improved with a 2-opt heuristic on real road distances from the Google Routes API, while respecting truck capacity and mandatory driver rest rules. The outcome: shorter total distance, better vehicle utilisation and full regulatory compliance.

03 — Production levelling (Heijunka)

Supply variability creates bullwhip effects downstream. We simulated a Heijunka levelling strategy — smoothing output and diverting surplus milk into storable products — to stabilise production and inventory.

What I learned

• Bottlenecks define system performance — improve the constraint first
• The right product mix beats more volume
• Logistics is a strategic lever, not just a cost line
• Optimisation is valuable only when it becomes an operational decision