Machine Learning for Manufacturing Location Optimization

According to a recent survey of senior leaders at manufacturing companies, over 90% plan to make significant capital investments in facilities in the next two years. While much of the planned investment includes expansion within existing manufacturing footprints, over 60% of respondents intend to pursue either brownfield or greenfield expansion of their capacity. However, before breaking ground on any new location, respondents indicated that they use three factors to weigh decision-making more than anything else – upfront cost, operating costs, and availability of skilled labor. While these three factors all make sense in the abstract, rarely do companies think about how tradeoffs in one or more of these factors can impact the overall cost and payback of the project.

Lotis Blue Consulting recently helped a client take a more strategic and holistic approach to their capacity expansion decisions. Using machine learning, we developed a methodology that compared the upfront construction costs and the steady-state operating costs for a broad range of potential locations (covering every county in the continental United States). The client used the insights to make informed decisions and justify the requested CAPEX budget from their Board.

Client Background:

  • NA industrial manufacturing
  • $1 billion+ revenue
  • Produced both specialized, high-margin and more commoditized, low-margin products
  • Margin on commodity products sensitive to shipping distance

 

Lotis Blue’s Manufacturing Capacity Expansion Optimization Approach

1. Define scenarios

In this first step, we pinpoint exactly what will be produced in the theoretical new plant. Will the plant make all products or a selection of product lines? Do we want to serve all customers or just a subset in a defined region or geography? Once the scope of the plant has been determined, you can begin to understand what your new facility will need (i.e., space, machinery and people). Defining multiple scenarios allows a company to test and quantitatively assess tradeoffs between how they choose to operate in the future.

2. Establish tangible plant parameters

Next, it’s important to translate the potential outputs of your scenarios into tangible characteristics of the manufacturing facility.  Decisions around physical footprint, staffing needs, machinery, and many other factors will impact both CAPEX and OPEX. This step requires a good deal of collaboration with subject matter experts across the company to understand how the choices made in each scenario translate to the operating parameters of a new facility.

For example, working with operations and/or engineering we developed models to translate the volumes in each of the defined scenarios into an estimate of the number of machines required to produce the corresponding volume. From there, we estimated the cost of the process machinery and any supporting equipment. Next, we calculated the physical footprint (sq ft) inside the building envelope and for any required yard space to support safe and efficient operations. Finally, we estimated the number of employees needed – broken down between fixed management roles and variable roles.

 

3. Run the optimization model

With the “plant” parameters defined, we calculated the CAPEX costs and steady-state OPEX costs at various locations across the US. While machinery and balance of plant equipment will be reasonably constant regardless of where you choose to locate new capacity, there can be a huge variation in land and construction costs depending on the county. Using the physical plant parameters (e.g., square footage required for building envelope and property overall) and a proprietary data set that features average land and construction costs by county, we can estimate the cost to buy and build the plant for every potential combination of plant location and scenario.

Similarly, key OPEX drivers like labor, utilities, or freight costs are location dependent. We take a similar approach to calculate steady-state OPEX – leveraging county- or state-level rates for various tiers of labor/expertise and utilities (e.g., electricity/natural gas/water). Finally, using $/unit/mile shipping (provided by internal supply chain or external provides), we can estimate the comparative costs for freight at the various locations. The result is detailed estimates of CAPEX and OPEX that allow us to compare multiple locations’ relative attractiveness.

Analyze the outputs and assessing qualitative tradeoffs

With all the data readily available, we can now make more tactical decisions and assess the benefits and tradeoffs of the various scenario and locations. For example, while certain locations may be desirable from a CAPEX perspective, comparatively high OPEX might make it less suitable for expansion. Finally, leadership can incorporate more qualitative factors into their decision-making process like the need for production redundancy, proximity to existing or future customers, labor environment, and/or local incentives.

When companies are looking to expand manufacturing capacity, they often focus on a few discrete locations in a pre-defined region without considering the global costs (or potential savings) associated with other locations. By looking at a broader range of geographies in early stages of expansion planning a company can uncover locations that can lead to significant upfront savings and/or drive faster payback through improved operations.