The current focus of lean implementations in the warehouse - whether it be wholesale/retail warehouses, distribution centers (DCs) or the finished goods inventories of manufacturers - is the elimination of waste and the changing of business processes. However, applying a lean flow distribution strategy is about more than just eliminating waste. It is also about creating a replenishment strategy that focuses on pull, uses a statistical approach to sizing the right inventory levels and ensures the business reacts quickly to changes in demand. Implementing these types of lean flow principles to inventory models can greatly improve operations, inventory days of supply and much more.
Before proceeding, let's answer what seems like a basic question: What is the purpose of a DC or warehouse?
To be able to ship product when it is sold. You can't sell it if you don't have it or quick access to it.
To supply it to customers when they want it. This means high service levels, fill rates and on-time delivery performance.
Converting to Pull
To keep costs low. Any cost of distribution or shipping is non-value added and takes profit directly from the bottom line. What is the biggest expense in the warehouse/DC? It's all of the finished product ready to sell, i.e., the inventory.
How does a warehouse/DC improve performance to achieve these purposes? One way is to employ a lean flow distribution strategy. However, with many traditional lean implementations in warehouses, there is so much focus on the elimination of labor waste and improving business processes that there is little attention paid to the improvement of service levels, the assurance that product is available and keeping the largest cost (inventory) as low as possible.
The focus needs to be on the lean flow distribution strategy. This strategy addresses all of the above items and also focuses on where the largest financial opportunity exists\m>reducing inventory while improving service-level performance.
Implementing a lean flow distribution strategy first requires the input of goods into a DC to progress from the traditional push to a pull process. Traditional push of product into a warehouse/DC is most often driven by a forecast or a sizing methodology that is marginally tied to what occurs in the DC. Therefore, when product stops being consumed, the input from the supplier may or may not change based upon how planners/buyers react. As a result, inventory is going to pile up somewhere, whether it's in the DC, in the pipeline, or at the supplier (see Figure 1, “Push Analogy”).
Planners/buyers may assume that the failure to sell something is an anomaly or that the rapid consumption of a product is not real demand and therefore should be ignored. What information are they using when they make these assumptions? Reams of data, statistical charting/graphing of demand patterns, mathematical comparisons between the forecast and actual sales? No, more often than not, they rely heavily upon their experience.
Although the experience of planners/buyers is invaluable, the process of spending thousands or millions of dollars to replenish a DC may require a little more analysis and analytical tools than planners/buyers have available. It has been my experience that often planners/buyers are trying to do the right things to keep inventory low, but they are so overwhelmed that they make the safe decision, based on the long-held belief that “if we don't have it in stock, we can't sell it.”
The conversion to a pull process works on the assumption that a company needs to maintain an acceptable quantity of inventory in the DC and the supply chain to meet customer demand but not create large excesses. The need to replenish additional product is driven by the consumption of product at the DC, not a planned (forecasted) future requirement. Once the inventory on hand and on order is reduced below the statistically sized re-order point, additional product is ordered from the supplier. If there is no product consumed, then there is no need to order product and put it into the pipeline.
What if we went away from a forecast or fancy formulas trying to predict the future for thousands of items and gave planners/buyers tools that ensured the time necessary to analyze the bulk of the dollars purchased? For inexpensive, low-value inventory items, the design of the replenishment system would make sure those items are, statistically, always available. These items (traditionally, C and D items) would be on a pure pull system, and the signal to replenish product would be an electronic kanban trigger or re-order point (ROP).
This statistically sized ROP would use high service levels to cover almost any variation seen in historical usage. The signal to replenish product would not be forecast driven, it would be just a signal to purchase when product has been consumed. Then, only periodically and by exception — when the recalculation told buyers to review the data — would there be any interface between the planner/buyer and these items.
The B items, those with some dollar value and historical usage, would be sized and replenished the same way as the C and D items, but service levels would be more conservative and the exception management effort slightly higher. There may be some analysis of the forecast projections, consumption history and statistical patterns (review of spike usage), and this analysis would ensure sufficient information to make informed decisions when necessary. However, these informed decisions are again on an exception basis, driven by differences between the forecast, consumption and statistically sized inventory. More analytical effort than a C and D class item may be required, but most certainly less than is being required today.
Finally, the A items are where the experience, focus and 75% to 80% of planners'/buyers' time will be spent. This is where 80% or more of the inventory buy and sales will occur. These items the planners/buyers will manage almost daily. The methodology will still be an electronic kanban trigger (ROP), statistically sized and signaled for replenishment by consumption of product, or a pull. These items will be reviewed often, compared to the forecast and historical statistical projections and influenced by the experience of the planner/buyer, any available market data and feedback or information coming from the customer base. Large consumption or sales will be analyzed, and future inventory builds will be planned carefully to ensure availability, but the statistical ROP will keep inventory as low as possible while meeting desired service levels.
There will be a lot of interaction with suppliers: meetings, e-mails, conference calls and a constant focus on taking waste out of the supply chain. Why so much focus on these items? The reason is, this is where large inventory savings can occur and where lack of attention to details will break the bank. Here, engineering changes will be choreographed like a Broadway musical, and every aspect of these parts will be planned. Planners/buyers can spend all or most of their time analyzing A items because they have relinquished control of the C and D items, which represent 50% or more of the part numbers in the kanban system.
Some companies may already be using a pull system to determine when to purchase additional products. They may call it a reorder point, which, if properly applied, can be an electronic kanban signal triggering a replenishment. The question concerns the methodology used to size the reorder point and signal the replenishment. What is the requested minimum order quantity (MOQ)? The lean flow method of sizing the ROP is based on some of the same data that any ROP-sizing would be based on: historical usage, lead time from the supplier, service levels, etc. The devil is in the details of how the information is used to create the ROP.
The lean flow distribution strategy approach uses historical usage data to statistically size the ROP to ensure:
Sufficient inventory to meet variations in demand to prescribed service levels. This will allow a DC to keep the minimum inventory necessary to meet customer demand but have enough to meet the desired service level. This will also allow the DC to rapidly change ROPs to increasing or decreasing demand patterns and anticipate future demand based upon statistical projections.
Very high service levels are met or maintained. Anyone can easily reduce inventory to any level management requests by arbitrarily controlling the total dollars purchased and received, but reducing inventory while improving customer fill rates or service levels is what the lean flow distribution strategy achieves and where the return on investment will be created.
MOQs are set to provide adequate on-hand inventory and determine the frequency and quantity of product delivery. The MOQ provides a consistent quantity to be purchased but requires a variable delivery schedule with a consistent supplier lead time. If demand is down, on-hand inventory will not drop below the ROP, and no material will be replenished. If material is consumed at a higher rate than normal, then the ROP will signal for replenishment more often than normal, asking for more material. Product requested uses a pipeline of replenished materials (pull analogy) in transit to the DC.
When there is a long lead time between the supplier and the DC, the inventory received needs to be shipped periodically in smaller quantities rather than all shipped at once. This is especially true of materials with longer than four-week lead times. To receive nine weeks or more of something in today's fast-paced global market reflects poor use of cash, space, labor and betting the next nine weeks' sales on a crystal-ball projection of what has been placed in inventory.
There are many examples of companies that have used lean flow pull principles to achieve success in improving or maintaining high service levels while dramatically reducing inventory in their warehouses and DCs.
To properly implement a lean flow distribution strategy, the statistical ROP will need to be resized frequently to ensure that trends in demand are properly applied to the ROP. This means that A items should be resized at least once per month, B items quarterly and C items every six months. To achieve this ability to resize the kanban trigger efficiently, a robust interface between the MRP/inventory/ROP repository software and the lean flow tool used to create the statistical ROP will be required. This needs to be considered when planning the implementation of the statistical kanban trigger.
Receiving benefits from a lean flow implementation of a pull system in a warehouse or DC will probably not require a huge rearrangement of facilities, no new foundations for machines, no large changes to the physical processes of receiving, material handling, picking orders and shipping product. In fact, some additional savings beyond the potential inventory reductions include:
With less inventory, less distribution space is required. This could allow the use of the space made available to be sold, leased or used for other business purposes.
With less excess inventory in the warehouse/DC, some of the new space could be made available for more efficient storage of items, so there is no stacking item A on the floor in front of B, then digging item B out when needed. This reduces labor waste and therefore costs.
With less material coming in large batches, there is less labor needed to cover peak demands and therefore less overall labor required.
Can any warehouse or DC take advantage of a lean flow distribution strategy? Absolutely. So, what are you waiting for?
Preston J. McCreary is a founding partner at FlowVision, based in Dillon, Colo. He has been educating companies and implementing lean flow manufacturing concepts for more than 17 years, in such industries as power generation, aerospace, railroads, machine tools, compressors, construction equipment, foundries, high-tech, air-conditioning equipment and vitreous china.