There have always been uncertainties in supply chains. If we knew what products were going to fail or what parts customers were going to order there would be no need to forecast demand. If suppliers always delivered perfect orders on time, there would be no need to buffer their variability. If both were true there would be no need to carry safety stock as the whole supply chain becomes deterministic, without probabilistic uncertainties. If this were the case, we would have no need for material planning systems.
None of those are realistic attributes of real supply chains! Demand, especially in certain verticals such as commercial aviation, has extreme variability of already low demand volumes. Forecasting is clearly required, but not sufficient. While increasing in precision with specialized algorithms, connected assets, predictive maintenance etc., forecasting can help us gauge the variability of demand. Much more comprehensive planning systems are required to account for it though. Lead time variability introduces uncertainty from the supply side of the equation.
When married with service goals (for example outcome-based Service Level Agreements or warehouse order fill rates) the demand profiles and associated variability can be processed through Multi-Echelon Optimization models to create an array of inventory targets to account for the uncertainty while meeting goals at minimum cost.
During times of extreme uncertainty, such as those introduced because of COVID-19, the ability to model alternate scenarios is critical! These scenarios can reflect radical changes to the assumptions of a steady state supply chain to quickly evaluate alternate response and recovery plans and continue to evolve our response to the pandemic as conditions change.
Advanced planning systems help account for new supply chain realities while honoring our commitment to deliver maximum equipment uptime and parts availability for the end customer.
There are likely financial implications for every stakeholder in the multi-enterprise aftermarket supply chain. For groups planning inventory for aftermarket support, many enterprises will see significant reduction in demand arising from reduced activity profiles of the products they support and associated customer parts orders. As demand is typically correlated to revenue this short term drop in incoming financial resources may eventually impact the amount of cash available for purchase and repair orders and potential reductions in aggregate inventory holdings as well.
Sophisticated optimization algorithms allow for budget constraints, at an enterprise level [or other level including product(s), location(s), customer(s)] to cap investment or spend. This reveals the strategic impact of reducing inventory and cash outlays on the ability to meet the defined service goals. By varying how and where the constraints are applied the service provider can understand how different expense levels and allocations of available funds impacts their customers. Once these outcomes are understood and go-forward plans are defined, these short-term service delivery challenges can be proactively shared with the impacted customers to correctly align expectations and minimize negative consequences.
Demand history has always been a fundamental input to the vast majority of forecasting models. Unfortunately, with COVID-19, yesterday may not be reflective of tomorrow. Some industries, such as medical devices, will experience unforeseen spikes while most other industries may see steep fall off in demand. Reduced utilization of the assets they support including the number of activities a machine performs or the flights a plane is flown, as well as external customer orders can fluctuate with this demand. History cannot predict these shocks to the demand profile caused by the pandemic because they have never happened before!
This can be easily done by reducing the number of historic periods considered in moving average-based models and increasing the alpha values in smoothing techniques. Both changes shift the focus of the model to emphasize recent demand periods over older, less relevant data.
They are best of both worlds models in that they impute a failure rate based on historical failures scaled to historical activity (failures per operational hour) which represents what we can learn from the past. The forecast going forward applies the fail rate to projections of expected asset utilization representing the future – hence the best of both worlds. While this is most relevant to certain verticals, such as airlines, where the activity profiles can be more accurately predicted or even controlled, it can still be applied in simplified fashion to almost any vertical. In this case, factors are created for example, a scale of 100 where 90 represents a 10% reduction in install base activity. These factors can quickly influence forecasting scenarios based upon different expectations of activity growth or decay. Accuracy tracking metrics can automatically highlight which forecast scenario is best borne out by real demand.
Suffice it to say that it is highly likely that using the same history-based forecasting models with the same parameters will lead to significantly overstated forecasts which if unchecked will bloat inventories just as cash is becoming scarcer.
Service already sees more variable lead times than other supply chains because of the uncertainty around the effort to repair parts and the production of many low volume, highly engineered items. These variations in lead time are easily handled by planning systems but the pandemic may have dramatic impacts on the supply side of the plan. Suppliers are being required to reduce staff which reduces productive capacity so orders will either ship incomplete or late. Choose your poison. Internal human resources that manage warehouse operations and record shipments and receipts must also be reduced delaying processing times, further adding to lead time growth and variability.
Longer and / or more variable lead times directly correlate to higher safety stocks. Robust planning tools allow for rapid modeling of percentage changes to lead times across part groups (e.g. by vendor or part category). This will at a macro level help gauge the financial impacts of extended lead times and identify which parts are most effected. At a micro level, modeling individual parts across two candidate suppliers may actually favor buying more expensive parts from a vendor with a shorter lead time.
Typically, we take for granted the physical distribution network and how material passes through it. It is static unless there are planned strategic initiatives to make alterations. The pandemic may impose unplanned modifications to which locations are active in the network and their process capabilities. This can have implications for how material can flow through the network and ultimately to customers. There may be an emphasis to stockpile parts in a very limited number of locations.
Sites may need to be temporarily consolidated or closed based upon staffing constraints, but parts still must get to the customer! Advanced Multi-Echelon Optimization allows us to evaluate how different distribution networks and the material flows within them impacts inventory cost and network efficiency. It will be a valuable tool in quantifying the tradeoffs between different network strategies.
In discussing budget constraints, we focused on how different levels of inventory and spend impact the achievable level of service. The process can also be run the opposite way. If short term reductions to service goals need to be considered new targets can quickly be run through the optimization engine and supply plan to project the reduction in holdings and spend on parts. The changes to the service goals could be universal or only apply to certain parts, products or customers to evaluate where you take your risk with service delivery and how to manage your customer’s expectations.
Maybe now more than ever it is imperative that we focus on the Key Performance Indicators and other metrics that gauge the health of our supply chains. More is at stake than performance reviews and bonuses. These KPIs monitor service to our customers, changes in demand and our ability to predict it, inventory related financials, lead times and supplier performance. The metrics can highlight opportunities to refine the planning model’s parameters or inputs to best prepare us to weather this horrible storm.
In summary, we cannot act as if it is business as usual because little in business is usual today. Every strategic assumption and input to managing a supply is subject to unforeseen change. Tools that can rapidly evaluate the potential dynamics of these factors are mandatory to efficiently understand impacts of changes and how to develop response and recovery plans.