The foundation of any material planning system is the determination of optimized inventory thresholds that can meet targeted service goals at least cost. The safety stock component of these levels is intended to insulate the service provider from uncertainty in the supply chain over a resupply lead time horizon. These uncertainties most commonly manifest in demand variability, number of units and timing, and suppliers’ performance, adherence to lead time and quality.
Planning systems principally attempt to minimize the uncertainty through the inclusion of a broad array of relevant information and highly specialized algorithms. Any ability to increase forecast accuracy or reduce component lead time drives lower levels of investment for the same service goal achievement – a truly worthy goal.
This article explores how three emerging technologies, predictive maintenance, big data analytics, and 3D printing, can become essential enablers towards that goal.
Advances in new technologies are beginning to expand the available data sets, even causing some to suggest that the traditional planning models and tools may no longer be as relevant. The excitement about incorporating big data analytics and real-time performance data from IoT-enabled connected assets is surely justified.
However, as with any evolution, the new enhances the old rather than replacing it. The optimal path forward incorporates learnings from these new data sources within a robust planning model. Effectively breaking the reliance on proxy data in the forecasting process with data provided by the assets themselves. Actionable data that has been screened through big data analytics, highlights unforeseen patterns of failure.
Some have suggested that predictive maintenance will obviate the need for forecasting and material planning. While a lofty goal, with many commensurate benefits, predictive maintenance will become a key part of the forecasting process but not make it obsolete.
In industries where asset outage has crippling costs (airlines, plants, refineries, etc.) the motivation is high to develop capabilities to predict part failure in advance. The business impact of proactively replacing parts during non-operating hours is significant! Avoiding downtime saves time and money and delights the end customer. Achieving greater operational availability out of supported assets is the principal benefit of predictive maintenance.
This process will evolve into a critical part of the forecasting and planning process. The advance notification of demand will net consumption off of forecasted values and provide lead time to pre-position the correct part in the correct location before the maintenance event is performed. There are a number of reasons that additional forecasting and planning logic are still required.
Predictive maintenance focuses exclusively on parts where readings (e.g. temperature, pressure, activity profile etc.), can provide insight into the status of the specific component. While this demand “stream” may represent highly critical parts for an asset's operability, there will always be many more parts for which this process cannot be applied, necessitating other supporting forecasting logic.
Additionally, for those parts where predictive maintenance is relevant, the advance notices of failure will represent only a partial picture of the entire demand profile. There will be random demand occurrences that were not predicted by the sensor readings that must be accounted for elsewhere and incorporated into a safety stock model to ensure parts availability.
While predictive maintenance may help us understand when and where a part is to fail, there may be even greater value in knowing why it is failing. Traditional planning models used a provided or imputed failure rate for a part and an estimated installed base and / or activity profile. While this is vastly superior to forecasting solely based upon demand history it has some troubling shortcomings. The same part applied to different assets or usage or environmental conditions will likely have very different failure rates and reasons for failing in the first place.
IoT-enabled connected assets generate incredible amounts of sensor-sourced information. Volumes of data too great to be comprehended by an analyst without tools to parse out the needles within the haystacks. Thankfully big data tools are doing just that, helping expose causality that is unique to a part or a class of parts, or where multiple factors combine to contribute to a part failure. Patterns in equipment operations or component level failures, impossible to detect without tools, can provide additional insight into future part demands.
As these processes evolve, they replace simple fail rates, universally applied to a part, with matrixed ones that represent the variation in equipment age, how, where and how much the asset is used. New causal relationships can also help refine the forecast and drive down required safety stocks. As with predictive maintenance, big data analytics generate value for a subset of parts but for certain parts and causes of failure will never be harnessed. These analytical models are extremely valuable inputs to forecasting and planning processes without which it is impossible to translate them into value and actionable decisions.
The first two technologies addressed safety stock reduction through increasing forecast accuracy by incorporating new data and causal relationships. Additive manufacturing, where parts are produced on-demand, can have a dramatic impact reducing manufacturing lead times as well but approaches the challenge from another angle.
A growing number of parts are becoming candidates for this process as capabilities have evolved from plastics into metals and composites. For those parts, we may observe radical changes to the supply chain as the transportation overtakes production as the driver of lead time. In this scenario, the planning system prioritizes demand planning above supply planning, which is typically a balanced set of processes. The goal of the planning system then becomes maintaining sufficient stock or productive capacity to provide requested parts within the customer’s allowable wait time. As with the other technologies discussed, not all parts are amenable to an additive manufacturing process and more traditional planning will still be a necessity.
Predictive maintenance, big data and 3D printing enable next plateau forecasting and planning process results by replacing proxy data and assumptions with factual and dynamic inputs as well as algorithmically determined causality. Doing so gracefully eliminating manufacturing lead time for a class of parts.
However, they will only be available for some of the parts some of the time and therefore must be considered as complementary extensions of next generation forecasting and planning processes and not their replacement.