One of the main goals of any manufacturing company in Kazakhstan is to meet the needs of its customers and end-users with a healthy balance between service levels, inventory, and costs.
To achieve this, it is important to follow the 7Rs rule.
There are so many external and internal factors influencing production planning processes that humans are no longer able to take them all into account. In the context of increasing task complexity, Kazakh companies are increasingly turning to new, more modern strategies and tools, including artificial intelligence (AI), machine learning (ML), and technologies based on numerical optimization methods. And while AI/ML-based solutions are already in use in Kazakhstan, approaches to optimizing planning processes are worthy of separate discussion. So which technologies will help a company to implement the 7Rs and achieve a balance that takes all inputs into account?
Economic efficiency. According to various estimates, mathematical algorithms for optimizing production processes can reduce production costs by 2% or more by increasing overall efficiency. How? According to international statistics, the use of integrated business planning platforms and optimizers can reduce inventories by 5-20%, improve customer service by 10-15%, and shorten the planning cycle by anything from 30 to 90%. The range is wide, but the outcome depends largely on the maturity of a company’s planning processes.
In today’s world, optimizers are used in almost every industry for a wide variety of tasks. For example:
– FMCG, pharmaceuticals and retail – to plan production, promotions and deliveries;
– Passenger transport and logistics – to optimize supply chains, routes and pricing;
– Extractive industries – to distribute material flows between processing stages;
– Agriculture – for crop and fertilizer planning, and for allocation of machinery across fields.
Heuristics and optimizers. A plan for production, deliveries, shipments, etc. can be calculated using two mathematical algorithms: heuristics and optimizers. What are these algorithms, and what is the difference between them?
A heuristic conducts a linear, step-by-step examination of a version of the production plan as a simple difference between demand, current inventory and previously confirmed production orders, but takes almost no account of the constraints that exist in the supply chain. It is of course possible to specify some of the constraints using additional formulas and algorithms, but not all of them. In other words, heuristics take into account an unlimited, and therefore not always feasible, demand, which is then analyzed and adjusted manually by experts to match the actual capabilities of the supply chain. Often this adjustment is based on expert opinion rather than numbers.
An optimiser instantly calculates a feasible production plan by simultaneously processing many variants across all stages of the logistics and process chains to find the best option. During the calculation, the plan is balanced against the given constraints (e.g. capacity, component availability, etc.) and the target function (e.g. the final plan should be the most marginal). The final solution is feasible and based on real data rather than intuition. For example, capacity utilization cannot be greater than 100%, so volumes outside this range are trimmed and visualized as under-delivery. Manual adjustments to plans are possible, of course, for example in the case of individual management decisions, but they are many times fewer than in the case of heuristics.
Putting theory into practice, heuristics are more likely to be used in companies with a simple linear technological and logistical chain, with a minimum number of constraints and SKUs. If the technological and logistic chain has a high variability and a large number of regularly triggered constraints, it is difficult to manage without an optimizer. A significant amount of manual work will be required to adjust the plans, and the optimal effectiveness of these plans will be dependent upon the expertise of a particular person.
The algorithms are typically implemented on solvers such as Gurobi and IBM CPLEX, which are increasingly embedded in leading Integrated Business Planning (IBP) platforms such as Anaplan, or in Production Planning solutions.
Case studies. Globally, optimizers have been around for a long time, and are widely used. At a cement company, optimizers monitor the operation of processing equipment – kilns and mills – in real time, resulting in significant productivity gains and energy savings. A sunflower oil producer optimized its supply network by improving the location of distribution centers and balancing production with fluctuations in demand, resulting in lower costs and greater efficiency.
A large fertilizer manufacturer added an optimizer to its Integrated Business Planning (IBP) solution. Algorithms now take 5 minutes to calculate optimal sales and delivery plans for the month, and production plans are made down to a daily level. This makes it possible to select the most profitable sales and production areas, achieve 100% capacity utilization, and determine internal consumption requirements.
A large Kazakh manufacturing company has to manage a fairly complex technological chain (multi-process production, 4-5 alternative lines for one product, component variability), as well as «black swans» in the form of sudden government procurement. The balancing of the plan is done manually, is very labor intensive, and is based on the expert opinion of the employees, with the cost of error being very high. In this instance, we advised the management to implement an optimizer, as heuristics will not solve the company’s tasks, and will hinder its processes rather than helping them.
How do you know whether you need an optimizer or a heuristic? When choosing a tool to solve business planning tasks, companies can rely on the following scheme.
In the case of a large number of constraints, a heuristic helps to find one of a range of admissible solutions, while an optimizer helps to find the best possible solution. Depending on the specific problem and requirements, both heuristics and optimizers can be used. It is just important to remember that the requirements in terms of data quality, knowledge, skills and discipline on the part of the planning team are much higher for an optimizer than for a heuristic.
So when is it time to start using mathematical algorithms to optimize production? If your production process involves many variables, if your company faces frequent changes in production conditions and therefore efficiency problems, if you operate in a highly competitive environment and there are fairly strict requirements for quality and delivery times, then optimization methods based on mathematical algorithms will enable you to reach a new level of productivity. For Kazakh companies, optimizers can be a real game changer, raising the efficiency of production planning to a new level.
Author: Anastasia Pyleva, Senior Supply Chain
Management Expert, Planingo
