Three Masters students from ZLC have developed new tools to address lot sizing and capacity planning in the pharmaceutical industry, which have the potential significantly to reduce working capital requirements.
By Dr. Rafael Diaz and Dr. Spyridon Lekkakos
Traditionally, strategic focus in the pharmaceutical industry has been on R&D, sales and marketing. However, as markets have matured, research costs soared, approvals processes lengthened, and in some cases price caps introduced, drug companies need to find savings and efficiencies in manufacturing and supply chain. Optimising lot sizes, and their allocation to available capacity across multiple production lines, is an obvious approach, but it is a far from simple problem.
The company under investigation is a sector-leading multinational using 14 different technologies across 31 sites to manufacture in 42 product families – there are over 3,000 skus in the system. Production is typically in three interdependent stages: creating the active substance, combining these with other materials such as fillers to manufacture the pill, potion or other product, and packaging, which with diverse product- and country-specific requirements for labelling, coding, instruction leaflets and the like is itself an extraordinarily complex process. In theory it may be possible to manufacture a given substance or product on any one of a number of lines across the globe: in practice this flexibility may be impaired because, for example, the regulatory authority in a customer country has not approved all of the firm’s facilities.
Substance and product manufacture is by batch, the batch size being determined to give the best yield given technical, quality and cost constraints. Often, several batches of a drug will be run consecutively, creating a ‘campaign’, but even with this simplification, planning across a complex network with inadequate decision support tools is a challenge. The general approach has been to maximise campaign size to create economies of scale in the utilisation of expensive and specialised equipment (and indeed labour). However, this can create on the one hand excesses of inventory or, on the other, reduced flexibility in responding to variations in the product mix.
Pilar Albar Bello, Bruna Fernandes Basile and Fernanda Caropresso initially looked at this problem in terms of a traditional network optimisation and economic lot sizing approach aimed at maximising profit. However, it quickly became apparent that this approach would be ineffective – production costs, including in particular salaries, are to a very large extent fixed, so no degree of rearrangement of production across the network would have any discernible impact on the Profit & Loss account.
Targeting below the line
Is there an alternative approach that might address ‘below the line’ costs on the balance sheet? Albar, Basile and Caropresso looked instead at the potential for reducing working capital requirements by optimising the well-known relationship between batch size and inventory value – providing demand is met, and within network capacity restraints, smaller batches create lower and smoother stock levels. Additionally, smaller batch/campaign sizes should make more production more flexible in response to changing demand.
The usual counter-argument is, of course, that smaller batches imply more frequent shutdowns and changeovers. The approach would therefore have the goal of maximising capacity utilisation, including shutdown and changeover periods, rather than of minimising cost.
The students decided to develop a prototype focused on the initial drug substance production stage. Through mathematical modelling and computer aids, and using real data on demand, inventory, production and regulatory and other constraints, the students have used non-linear mixed integer programming to develop a new, three part optimisation tool. This includes firstly a campaign sizing optimisation module that recommends how many runs of a product should be allocated to which work centre, given availability, capacity, demand and regulatory constraints. Second, a campaign scheduling module translates these recommendations arising from different production orders into a scheduling solution that minimises average cycle inventory. Thirdly, an inventory reporting module plots demand and optimised production, giving visibility of initial, average and maximum inventories in a dashboard presentation. Importantly, it can also generate a new scenario illustrating what would be the effects if production in (and therefore demand from) the next, downstream, production stage were smoothed.
The medication works
Results are impressive – the model shows a reduction of 13.6% in average inventory through rationalising campaign sizes and schedules; 23.8% through smoothing production, or 19.1% by changing drug substance manufacture according to changes in demand from the drug production stage.
Talking through the use of the new tools with the company, it also became clear that the allocation problem was being exacerbated because planners are divided according to product family and by production stage while competing to schedule onto common machines. The company is responding by creating ‘business partner’ roles in charge of consolidating information and running the model.
Sensitivity analysis has also been built in, enabling ‘what-if’ scenarios to be run – for example, to predict the effects of, and the best responses to, events such as line or plant shutdowns, additional capacity, or changes in regulatory approvals. The firm can quickly calculate impacts on current capacity and answer important questions such as the ability to supply new demand, the effects tis would have on capacity utilisation, and the impact on cycle inventory levels. The intention is now to extend use of the new tools to the next, product stage.