Lean philosophy has become popular as the basis for avoiding waste and enhancing operational efficiencies primarily in manufacturing, but more recently within services organizations. While Lean is more of a philosophy than a collection of tools, firms intelligently embracing some of those tools have gained considerable advantages.
One such tool is Demand Pull, which has tremendous advantages over push-based approaches. Firms still relying on push to set inventory and guide production planning should rethink their strategy and consider Demand Pull, which places the primary emphasis on the consumption or shipment of goods rather than forecasting. The difference is akin to setting a fixed schedule to refuel your car every Monday (push) instead of monitoring the gas gauge and filling the tank when the level approaches empty (pull).
Does this somewhat minimalist approach to supply chain design stifle growth? Can a firm adopting this approach ramp up quickly enough to support the rapidly increasing demand that's bound to occur as the U.S. economy recovers? While consumption-based planning might appear to be incapable of handling this eventuality, the beauty of pull-based planning is that it's designed to handle uncertainty. The challenge, of course, is identifying how to incorporate and implement demand pull in your supply chain design.
Setting the Inventory for Pulling
To illustrate how this is done, we'll use the example of a Manufacturing OEM which was severely affected by the rapid fall in demand during the recession. Their forecast-driven planning system could not reduce production quickly enough, leading to obsolete inventory. The firm decided to adopt a Lean-based supply chain using the Demand Pull technique. It also enhanced the collaboration with its manufacturing partners by investing in IT systems that allowed it to share the demand forecast and actual customer orders visible to Tier-1 and Tier-2 suppliers.
The inventory of each component at different nodes was set based on actual historical consumption at that node, and replenishment was triggered to maintain this set level. This was just enough to support actual demand during lead time and equal to the average historical consumption of the item. Additionally, to account for uncertainties, a safety stock, also based on historical variability in consumption was added. This inventory is identical to the traditional ROP (Re-Order Point) with the difference being that historical consumption data was used instead of the forecast. In the language of Lean, we call this the Kanban, and use this formula:
ROP = Average Historical Consumption during Lead time + Safety Stock
Whenever total inventory dropped below Kanban a build signal was triggered and transmitted downstream to initiate the production or the replenishment of the item.
To estimate average demand, the lead time was first defined as the amount of time from the point at which a complete customer order is received to the point when the inventory is available for the customer's use. Afterwards all of the discrete historical shipments during a particular lead time were bucketed into a single number called consumption. Finally the average (statistical mean) of all those consumptions for the previous two to four quarters was computed.
Safety stock was calculated using the traditional formula of Demand Variability * Customer Service Level factor. Once the demand data was bucketed into lead time, the variability of those buckets was measured using standard deviation. A customer service level (instances when the requirement for the item is filled completely within the commit date) was selected for different product groups and the corresponding service level factor was determined using z-score. Once the safety stock was calculated, it was added to historical consumption during lead time to determine the Kanban size.
Whenever total inventory dropped below Kanban, a build signal was triggered and transmitted downstream to initiate replenishment of the item. The total inventory was computed by adding the on-hand and on-order inventory. The order quantity depended on the set-up and capacity of the manufacturing process and the logistics operation, which influenced the package size. Order Quantity was larger of minimum production lot or an integral multiple of the minimum package quantity.
Limitations to the Approach
As useful as this pull-based method is in smoothing out supply chain inventory, it's not a one-size-fits-all solution. It is critical to be aware of circumstances where such pull-based supply chain is not the optimal design.
While safety stock takes care of variability in demand quantity, it is assumed that there is no supply side variability in quantity or lead time. If there is any supply side uncertainty, than this particular design is incapable of protecting the firm from such variations. This can be particularly severe for items with a very long lead time. A pull-based design is not necessarily the best options in those scenarios. Several approaches, including co-located JIT-hubs or switching to more regional suppliers instead of relying on the global low-cost vendors, should be considered before adopting this new design. Another issue could be known variation in demand. The safety stock is meant to support only an uncertain change in demand. Hence any variation known in advance, such as a large one-time order, should be planned for separately.
For items with large but a very steady average demand, the calculated safety stock was very low and hence a minimum safety stock was proposed for these items. Initially this was set at a fixed percentage (say, 30% to 40%) of average demand. Alternatively for items with extremely volatile demand, this approach was considered unsuitable.
Now, many experts have suggested that the best supply chain planning system should mature to the extent that forecasting is not at all necessary. We disagree. We have found that a forecast is still the best guess predictor of long-term future demand and should continue to be generated to assist with capacity planning (both equipment and manpower), collaborate with trading partners, performance management of trading partners, and launch of new product.
How to Manage the New Process
Most of the data manipulation to achieve pulled-based demand is done outside the ERP (Enterprise Resources Planning) system. Does it add unnecessary complexity to the operation? We've found that the benefits of Demand Pull are so overwhelming that a manual process used quarterly to adjust the parameter is not burdensome.
Unfortunately, popular enterprise management applications do not yet support the data manipulation capabilities required for pull-based design. The best current solution is to download necessary data into a spreadsheet for calculation and afterwards upload it into the ERP. Tailor made solution to automate the process of retrieval, data manipulation, calculation and upload is also an option -- recently a few commercial applications have come up to automate this process.
Using a Pull-Based Design to Improve the Bottom Line
In spite of being very popular in many industries such as automotive and, recently, high-tech, most firms have yet to adopt pull-based supply chain management-even if they embrace some other Lean philosophy. Using the simple analytical approach we've described, supply chain managers who adopt demand pull can become experts at managing uncertainty-preventing costly excess inventory crises during downturns and helping to ramp up as business improves. With a robust pull-based supply chain, organizations can improve on-time shipments, increase customer service levels, free working capital, reduce components obsolescence and improve the reliability of the supply chain. It's the essence of Lean.
Manoj Nanda is a Consulting Manager for Wipro Consulting's Supply Chain Practice and is based in Dallas.