Industries
Manufacturing
In steel, cement, building materials and ceramics, we bring line, kiln and supply capacity together with demand forecasting in a single plan.
In manufacturing, the cost of a decision rarely shows up where it was made. Congestion on one line delays the next stage. Raw material bought early creates inventory cost, and bought late it stops the line. A wrong demand forecast either fills the warehouse or leaves the dealer empty. In steel, cement, construction chemicals and ceramics we have met this problem in three forms.
Sequential stages: steel
Turning an order into a finished product is not a single step. In flat steel, material passes in sequence through pickling, cold rolling, galvanizing and coating; each stage has its own line, speed and capacity. A stage cannot run before the preceding one completes; once an order is started, every stage must finish within the same period. Optimizing lines one at a time does not improve the total.
Which raw material an order is produced from is not fixed either: a piece in the warehouse, a supply order in transit, or a purchase not yet placed. Solving production planning and raw material planning separately produces inconsistent plans. The models we build treat both as one problem and produce the production, raw material consumption, supplier and financial plans together.
Network and energy: cement
Cement and clinker are produced in mills and kilns across many sites, stocked in silos, sold domestically and for export. Which plant produces which product at what rate, which demand is served from which site, which port is used: these decisions are linked, and taken separately, an improvement in one place creates cost in another.
Once the clinker plan is set, each kiln’s energy need per period is known; then the fuel mix and the purchase plan (which fuel, when, from where) are solved. The mix must stay within calorific, ash, sulphur and volatility ranges, and the fuel must reach the depot before it is needed.
Demand forecasting: building materials
In construction chemicals and ceramics the main problem is demand: thousands of SKUs, hundreds of dealers, seasonality and campaigns. The forecast model is built at product–dealer level and its accuracy is monitored continuously. What starts as an MVP becomes a product the planning team uses in the S&OP meeting.
Scenario work
In all three areas what is actually needed is a set of plans that can be compared. “What if we raise capacity on this line”, “what if we stop buying from this supplier”, “what if we idle this plant for a month”: each is a scenario and should be answerable with numbers.
Our work in this area
Çimsa
Çimsa S&OP: from global network optimization to the fuel procurement decision
We brought Çimsa's production, logistics and sales network across 15 locations into a single mixed-integer model, then built a second model that turns the resulting clinker production plan into a fuel recipe and procurement decision.
- 10% ↓ Inventory cost
- 3% ↓ Manufacturing, warehousing, distribution cost
MMK
MMK: line balancing and order–raw material matching in flat steel production
An optimization system that matches open orders to the right raw material while respecting production line capacity and every stage of the bill of materials. Built across three phases.
- 4,000+ Open sales orders
- ~9 Raw material candidates per order
Kalekim
Kalekim: from MVP to a forecast management system in three years
We started with a machine-learning MVP for two product groups fed from Excel; in year two it became a forecasting process retrained every month for every product, and in year three the Kalekimforecast application that planners run themselves. The forecast is now an input to Kalekim's supply, production and logistics planning.
- 2 groups → all Scope
- 3 Encoding methods
Kale Seramik
Kale Seramik: a demand forecasting PoC we did not continue
An MVP that started on the same roadmap as Kalekim: two product functions, 34 methods, three encoding approaches. The result fell short of the target and the second phase was not started. This page explains what did not work, and why.
- 141 DFUs
- 200+ Product attributes