The Panacea of VMI
Much of what is called Vendor Managed Inventory (VMI) in the industry is not really managed by the vendor. The specifics of VMI have become diluted as the term continues to be thrown about generically for every supplier collaboration program inside the four walls.
Let’s look at some of the more common inventory management programs found in manufacturing facilities; and what differentiates them:
- Min/Max: (Line-Side or Stockroom) this program can be managed by either data thresholds or physical/visible replenishment models. In both cases, the customer typically purchases the inventory at the time of receipt.
- Kanban: Similar to min/max in that a two kanban system is effectively a min/max system. However, min and max have been separated into two physical containers creating visual triggers for replenishment. The return of the empty kanban container or an equivalent data signal is what triggers the supplier to deliver additional parts. Kanban quantities are typically computed to meet the lean inventory requirements of flow scheduling; or alternatively the supplier’s min/mult package quantity. Kanbans can also be used to physically deliver materials from the supplier to the stockroom or the manufacturing floor; or internally from the stockroom to the manufacturing floor. Direct delivery to the manufacturing floor is typically lowest total cost and fewest transactions.
- Bonded Inventory: (aka Sequestered) Inventory is held at the supplier in two forms: physical or computer segregated on-hand balances reserved for one customer; and/or “pipelined” inventory timed to arrive recurrently from the manufacturer. Both physical balances and pipeline allow the manufacturer to order materials to a perceived lead time that is significantly shorter than the factory’s “build from scratch” quoted lead time.
- In-Plant Stores: (aka Hub, Out-Plant) A stockroom or warehouse in which the supplier staffs and owns the materials, and is located in or near the customer facility. The supplier stages materials in this location but cannot deliver or invoice until a demand signal is received from the customer. Typically the stocking levels are expressed as so many days/weeks of forward demand from either an MRP share or production scheduling feed. The supplier will treat this inventory as “bonded” and “pipelined” per above.
- Consignment Stores: An In-Plant stores, but staffed by the customer. Typically the customer will provide the supplier with daily/weekly transaction summaries on the materials taken. This is lower-cost to the supplier but requires a tremendous degree of transaction control and accuracy to mitigate downstream liability and “trust” issues with the supplier.
- VMI: The In-Plant Stores or Hub model, but with the added requirement that the supplier is responsible for monitoring factory usage, projecting forward demand and providing inventory to satisfactory negotiated service levels.
It is this responsibility for forecasting and service-level replenishment that differentiate VMI from prior models. Currently the term “VMI” is bantered about loosely for any and all of the above supplier/manufacturer relationships; but it is an inaccurate use of the term without the required shift of demand management from the manufacturer to the supplier. This is a key point with tremendous ramifications.
It is this responsibility for forecasting and service-level replenishment that differentiate VMI from prior models.
VMI was originally developed as a replenishment method for independent demand items; such as blue jeans sold at Wal-Mart. The underlying assumption was that “Levi’s” would be able to replenish and aggregate current demand and forecast future demand better than either the local store or corporate purchasing organization. This is due to their familiarity with the overall market; including seasonality, fashion trends and marketing promotions. They can overlay their expertise on top of point of sale (POS) data supplied by the retailer. The same can be said of sunscreen sales during a heat wave or mosquito repellant after summer rains. The product marketing team at the manufacturer holds the advantage over the retailer on predicting demand.
These end products, referred technically as “independent demand” items, must be supplied by their factories at lead time; or supplied immediately from buffer inventories that are then replenished through scheduled pipeline orders or at lead time. In any situation where the demand inside of lead time is greater than the buffer quantity (i.e., finished goods); forecast data must be used. The VMI model thus shifts the responsibility for forecasting and inventory replenishment to the party most familiar with the item and its market: the supplier. This often makes sense in retail.
However, in the Electronics Manufacturing Services (EMS) industry, VMI is being used to manage “dependant demand”. Dependant demand items are those that have a direct relationship to a higher-level assembly. For example, if I am building one car then it is certain that I will need one engine, two doors and four tires. The engine, doors and tires are “dependant” on the assembly of one car.
A statistical model calling for 0.98 engines, 2.23 doors and 3.86 tires does not increase accuracy. The suppliers of engines, doors and tires are not going to be able to fulfill dependant demand any more accurately than the manufacturer was able to do. In fact, they will most likely perform worse.
VMI fixes nothing on dependant demand items. It is a convenient tool for masking the problems and inefficiencies of a supply chain organization by outsourcing apparent accountability. It is actually counter to the JIT philosophy of lowering the stream, uncovering the rocks and fixing the problem. It simply blames the supplier for the rocks and makes them pay for the water.
Since inventory has been shifted from manufacturer to supplier; “VMI” amounts to exchanging internal inventory value and carrying cost for higher supplier unit costs or cost of capital charges.
VMI for “dependant demand” items is both a strategic blunder and a tactical nightmare. Sharing current transactional usage data instead of forward bill of material consumption is a recipe for failure. Having suppliers responsible for building statistical forecasting and replenishment models on top of current consumption data, instead of providing forward-looking bill-of-material demand, replaces good data with less accurate statistical averages in all but the most steady-state production. A “VMI” model in the electronic component supply chain must be founded on the sharing of accurate MRP data for planning, scheduling and capacity; combined with point of use demand for delivery and payment.
But this begs the questions; “If your data is accurate, then why go to a VMI process? What value can the supplier add to demand management with a VMI program?”
VMI, if implemented on dependant demand items, is nothing more than a shift of ownership of inventory to point-of-use. In this sense it is nothing more than the latest synonym and buzzword for the in-plant stores and bonded inventory programs of the past. VMI in manufacturing is a cash-management tool, not the material planning and control solution that it is for retail end-item sales.
VMI in manufacturing is a cash-management tool, not the material planning and control solution that it is for retail end-item sales.
In the incredibly complex environment of electronics manufacturing the distinction between dependant and independent demand must be understood. Then, the right tools can be chosen for the job. For managing dependant demand items consumed in the manufacturing process, “VMI” is not the right tool. This is because the burden falls on the manufacturer to provide current consumption and forward scheduling bill of material demand; not the supplier, Can collaboration take place and service levels be agreed to? Of course, but the manufacturer must take the lead in this dance.
If manufacturers take ownership of the data quality they share with component suppliers, then “VMI-like” programs can be adopted as the cash management tool they are; without needing to pretend that they are a demand fulfillment method strategically greater than in-plant stores, kanban or min/max… or that they have any bearing on improved customer service beyond the capabilities of these earlier models.
© David Ginsberg
[Filed in the Tactical Lean section]
“…an interesting thesis which is certainly to a large extent right.”
-Vice President Manufacturing and Logistics, Nokia Siemens Networks
