The goal is to increase customer service with reduced inventory. An optimal safety stock policy can be calculated by the application of statistics and probability theory. Inputs include:

• Demand Variability (an output of the demand planning process)
• Supply Variability
• Replenishment Leadtime including the cost and ability to expedite
• Reordering lot size
• Target customer service level

We also implement a process to review the recommended safety stock level of the model versus demonstrated performance. The process can be integrated with the ERP system to enable safety stock levels to be reviewed routinely without repeating time consuming analysis.

The objective of this project was to improve the ability of the production scheduling team to optimise the schedule in a large complex manufacturing plant. Plant constraints and changeover considerations were all held in many peoples heads. The schedule had to be reviewed by approximately 8 people from different departments each week to get a view on schedule feasibility and optimality. Each department had competing objectives requiring a lot of subjective compromise. Even with the high degree of effort applied to review the schedule there were still instances where unknown infeasibilities in the schedule were found out the hard way.

The solution was to build a detailed factory model. The model comprised 145 resources and an attribute based change over matrix was applied to each for both time and cost on each resource. This enabled all the master data required to be broken into logical and manageable blocks.

A graphical schedule board was developed to enable the scheduler to drag and drop production runs with the impact of any scheduling decisions calculated and presented real time. Information available to the scheduler was as follows:

• Resource clashes taking into consideration overhaul plans.
• Predicted change over time and cost for each run based on which products were running on the resources required along with the critical path for each changeover.
• Days of cover before and after each run and whether the production run was inside defined SLOB risk rules.
• Stock out risk. The shipping schedule was added to enable the model to show the consequence of delaying production. For example, delaying production an hour where the output was destined for the local market usually means stock availability in DC an hour later. Delaying production for the export market might mean a missed vessel and a two week out of stock in the DC due to next vessel availability.
• Material constraints of critical materials were included

The last step was to apply a simulated annealing optimisation technique to the model enabling millions of different schedules to be costed and analysed.

The goal was to streamline the material planning process within a team of 5 so they could manage by exception. Challenges of the current MRP system included:

• Raw material inventory was held in multiple locations and the MRP provided no support for purchase versus move decisions. If one location was short of material but present in another location the MRP always recommended to purchase.
• The effect of lot sizing meant that in practice every recommended purchase needed to be reviewed manually prior to conversion to a purchase order. This was necessary to manage obsolescence risk. In the extreme case the MRP would recommend purchase of a lot size when existing inventory and purchase orders already satisfied 99% of the requirement. The residual of the planned purchase beyond the 1% shortage was destined for SLOB.
• The MRP system did not consider the shelf life of materials and would both recommend purchase of materials that would expire prior to consumption and assume stock on hand that was due to expire prior to consumption was not going to be an issue. The later issue was often discovered late
• The reporting of material constraints to the scheduling team was a manual process involving transposing data from the MRP system to an external database.
• There was no adequate days of cover reporting.

A system was built which integrated with the current MRP. Reporting included days of cover before and after each planned purchase and projected inventory value. The model contained logic around when stock held at other locations should be considered. Projected age at consumption was calculated for every planned purchase to manage shelf life. This enabled exception based reporting and application of the 80/20 rule so that the material schedulers attention was directed to the 20% of items that made up 80% of the inventory value. Planned purchases that were within rules were able to be automatically converted to purchase orders without the need for manual review. Generation of the constraints report was also automated.

Most ERP systems don't provide a bottom up inventory forecast. The affect of lead times and offsets mean that it makes no sense to sum the projected inventory for each component as per the ERP system. Typically:

• Material inventory is forecasted to be consumed at a discrete point in time before a production run starts.
• Finished Goods and Work in Progress from production is declared at a discrete point in time after the production run finishes.
• Stock in transit between locations is projected to disappear from the source at a point in time and reappear at the destination a fixed time later.

To accurately forecast inventory we apply a model that integrates with the ERP system and projects inventory at a SKU level taking into account that consumption of materials and generation of manufactured items is not discrete but continuous over the forecast length of production. Robustly projecting inventory is a key component of the supply step in an Integrated Business Management (IBM) process or if you are storage constrained. Note, that to forecast inventory you also need a mechanism to convert the unconstrained MPS to a supply plan that considers plant capacity, labour crewing patterns and material availability.

The reporting of risk around Slow and obsolete stock (SLOB) was a time consuming manual task in a business with approximately 2,500 SKU's. The ERP system provided a report that listed inventory lots that had been in the warehouse for longer than 6 months. Once a month an inventory analyst would take this list and manually look up the forecast to determine stock at risk of obsolescence based on their knowledge of customer shelf life remaining requirements. If the stock was at risk, they would manually add the relevant details to an excel spreadsheet that was subsequently circulated to the sales team. Short comings of the process included:

• The process was painfully manual and time consuming (2 days per month)
• It took at least 6 months for the any risk to appear on the radar. SLOB risk inside this horizon was invisible.
• The report was out of date as soon as clearance plans were added into the sales forecast.
• The risk wasn't prioritised based on financial impact.

The solution was to build a model that took inventory at a lot level and used the sales forecast to automatically project expiry risk based on the set of customer shelf life rules. This data was then made available to the relevant people in the sales team in a format that projected write-off value by month based on the current forecast. The model captured clearance plan assumptions and the write-off projection was updated real time as these were layered into the demand planning process.

We make the analysis of inventory levels routine and improve visibility of days of cover both at a product family and SKU level. This information can be used as an input into the Integrated Business Management process. Many ERP systems do not calculate days of cover. Others divide stock on hand by average demand per day over a fixed constant horizon which doesn't work very well as soon as demand is not constant.

Definitions:

• Days Of Cover (DOC): A forward projection of the number of days stock on hand is projected to exist
• Days In Inventory (DII): A trailing measure of average inventory over average demand over the last 12 months

Planning distribution of finished goods from a production facility to multiple distribution centres isn't always as easy as it could be.

Your ERP system will contain information on stock on hand, scheduled production, and demand at the DC's. However, working out how many shipping containers to book, allocating constrained supply across them, and projecting customer service risk at the end points of your distribution network is often a time consuming process, manually undertaken on spreadsheets. The result of a manual process is unlikely to be optimal for all but the very simplest distribution networks.

We have developed a system which integrates with data from the ERP system. This applies the shipping schedule to formulate a mixed integer program to determine the optimal distribution plan based on the current production schedule. The output of the model includes projected days of cover before and after arrival of each vessel at the DC. This enables the production scheduler to identify the severity of any supply risks as a consequence of their scheduling decisions. For example, the impact of a production run missing a vessel cut off by mere hours may be small or large depending on a number of factors which usually are not properly represented in the typical ERP system or available to the scheduler in the right format. Using our model planning distribution becomes an automated process and provides the right information to the right people at the right time.

We believe it is important to ensure that your planning master data properly reflects demonstrated plant performance. You don't want your capacity plans to suffer from garbage in garbage out. You also don't want to lose sight of your continuous improvement opportunities.

The recommendation is to implement a system around two key metrics:

• Attainment to Schedule to improve plan and schedule feasibility
• Operating Efficiency (availability x performance x quality) to drive continuous improvement

Also, if the system is intuitive enough, there is an opportunity to engage your staff more fully in the continuous improvement process, right to line level. Ideally, when your team leave at the end of their shift, they leave with a fair understanding of how things went and why supported by robust and actionable information.

Converting the unconstrained supply time from your ERP system to one that the cost to supply while accounting for your supply constraints requires the application of a mathematical model of your plant. It is likely that you will need to consider:

• Work centre capacity and the need to stock build around overhaul plans.
• Step changes in the crewing plan to model how inventory builds and dissipates around periods of labour surplus and labour shortage.
• Material constraints for long lead time and seasonally available items.

There are a number of options available from load levelling heuristics to liner and mixed integer programming.

We have experience in deploying both in-house solutions and commercially available advanced planning systems that come with their own inbuilt optimisation engine.