The dominant methodology for managing materials replenishment in a global distribution network is Distribution Resource Planning, or DRP.
How does DRP work? It applies a variation of MRP logic to distribution networks:
- Sales forecasts are made for all items and distribution centers.
- Safety stocks are set up for all articles and locations, possibly in the form of days’ supply for articles with sufficient turnover.
- Requirements — expressed as dates and quantities at the source site — are propagated from downstream to upstream, based on sales forecasts and applying minimum order and delivery frequency constraints.
The global deployment of a DRP is often a very significant investment. It involves multiple national/regional teams — and potentially incorporates multiple levels of distribution centers (global, regional, national) that are replenished in a cascading fashion. Implementation is time consuming, costly, and often must also consider organizational obstacles, as it interferes with the structuring of sales and marketing organizations beyond the supply chain team.
It is not uncommon to find that results are not satisfactory, even after this colossal effort of implementation and an investment of a few million Euros or Dollars:
- The factories and suppliers that supply the DRP-controlled network are subject to strong variations in demand — adrenaline rushes followed by slowdowns. The real consumption of the markets — the “sell out” — is difficult to perceive when you are upstream of the network.
- The stock doesn’t always end up in the right place — we’re out of stock here and in excess there…
- Managing scarcity situations is complex. The planner at the upstream site receives orders in quantities to date — without having a clear view of what the minimum requirement is. Are there backlogged customers two notches downstream, or is it just a requirement from a forecast and a safety stock? Should I serve Japan or Korea instead?
No problem, you may say, technology can help. If we put business intelligence and a control tower on our DRP, we will be able to help with the decision… at the risk of adding another layer of complexity, right? Some will even propose predictive algorithms based on artificial intelligence.
The Trouble With DRP
If we take a few steps back, the shortcomings of the DRP model become clear:
- A DRP is driven by detailed downstream forecasts. Forecasts are made at the SKU level — material + distribution site — and on the horizon of the replenishment time of the destination site. If, for example, you are replenishing supplies to Japan from Europe by sea, you must take several weeks into account. We know that the more detailed and distant the forecasts are, the more wrong they are — so you feed the DRP with an inherently bad demand signal. Detailed, and over a distant horizon.
- Safety stocks, lead times, and batching policies mask real demand, and generate the well-known bullwhip effect.
- It is difficult to assess what is a priority, due to the lack of systemic visibility. It is also possible that commercial organizations bring a bit of power play into the equation.
Evolving Beyond DRP
Could we take a simpler approach? Yes, at several levels:
- Simplify the structure of the distribution network itself. Incorporating fewer sites and fewer levels enable us to increase the frequency of replenishment and simplify our flows.
- Simplify the forecast. We need a consumption rate per item and distribution center. In many cases, recent actual consumption is sufficient at the distribution center level. If a forecast is needed, preferably a global or regional forecast, exploded by site according to relative weights. In all cases, smooth / average the forecast by rolling period.
- Pull-flow control, by re-filling consumption, with visual management.
- Give control at source to the sending site — of replenishments — in a VMI / shared supply management type mode.
- Use relative priority allocation logic to manage shortages or allocate surpluses.
- Drive continuous improvement of inventory and flow health in the network with a visual dashboard.
The Demand Driven Approach
Below is the replenishment situation in Intuiflow for my five destination sites. Each destination has a red/yellow/green sized buffer — and I should replenish to the top of the green.
I have visibility into downstream inventory, replenishment requirements, acceptable operating ranges, and relative priorities.
I don’t have enough to replenish all the sites to top of the green, so the quantity is evenly distributed to the sites that need it. It’s simple, visual… and automatable.
In distribution as in production, being simple requires a well-designed operating model, but brings considerable benefits for quick and relevant decision making.
