12,000 planned purchase orders a night
Fabrikam runs one master plan nightly against 9,000 SKUs. About 60% are purchased parts with a 21-day lead time; the rest are made across two BOM levels. Coverage on almost every item is Min/Max with a minimum of 50 and a maximum of 200, and item coverage carries no order multiple and no minimum order quantity. The plan creates around 12,000 planned purchase orders a night, most for one or two units, and a large share carry need dates that are already in the past. Buyers firm orders manually within two days and the planner spends six hours a day on action messages. Finance flags growing excess on the seasonal line while three stock-out escalations hit the same product family last month. Forecasts exist for the seasonal family only and are entered at site level.
1. Redesign coverage for the purchased parts so the plan stops producing 12,000 one-unit orders. Name the code and parameters, and justify with arithmetic.
Read the worked answer
Min/Max on a 21-day lead time is the engine of this problem: every time projected on-hand dips below 50, the plan orders up to 200, so it reacts to noise and hoards in lumps, while the absence of an order multiple means every requirement becomes its own order.
Recommended per purchased part
- Coverage code: Period with a coverage period of 14–28 days (start at 21, matching the lead time). All net requirements inside the window consolidate into one planned purchase order per item per period instead of one per requirement. Rough arithmetic: if a part has demand on 20 days of the month at 15 units a day, Requirement coverage produces ~20 orders; a 21-day period produces 1–2 orders of the same total quantity.
- Order multiple (for example 100) and minimum order quantity to match the vendor's pack size, so the order count falls further and the quantities are orderable as placed.
- Safety margins: a receipt margin and reorder margin in days so the 21-day lead time is protected against late receipts, instead of using max quantity as an accidental buffer.
- Negative days set so the plan can net requirements slightly backward rather than firing urgent orders for every tiny shortfall.
Keep Min/Max only where it genuinely fits - cheap, high-volume consumables where space or budget caps matter more than order count. Applying it to 60% of the catalogue is why the nightly output is thousands of lines. Also confirm coverage is maintained at the dimension the plan nets on (site and warehouse): coverage set per warehouse while demand and supply live at different warehouses produces both the excess and the stock-outs Finance is reporting.
2. Why do so many planned orders carry need dates in the past, and what should you do about them?
Read the worked answer
Past-due dates are usually a signal, not a bug. Planned order dates are computed backwards from the requirement date using lead time plus receipt and issue margins. If the requirement date is already inside the 21-day lead time, the order date must fall in the past - the plan is telling you the requirement cannot be met as dated.
Check these in order
- The requirement itself is late - sales order lines or forecast lines whose requested dates have already passed and were never rescheduled. Fix the demand, not the supply.
- Lead time is wrong. Released product or vendor lead time shorter than reality makes every calculated order date optimistic and past-due. A 21-day nominal lead time that is really 28 days will generate permanent urgency.
- Margins are missing so there is no buffer to absorb normal variation.
- Items with Manual coverage or no coverage group are excluded from planning entirely and their demand surfaces unexpectedly when someone notices - check the exceptions, not just the output.
- Action messages are being ignored. Advance and postpone messages tell you which existing orders to move; if the planner is spending six hours a day on them and the count is not falling, the underlying coverage settings are regenerating the same problem each night.
Policy decision needed, per order: firm and expedite (paid freight, alternate vendor, partial quantity), or accept the date and communicate it to the customer. The plan cannot make that call. What it can do is stop generating the same past-due order every night by consolidating demand into a period and getting lead times and margins right.
3. Forecast exists only for the seasonal family and is entered at site level. What must you verify so the season is neither double counted nor ignored when coverage is reworked?
Read the worked answer
Four checks, in this order.
- Dimension alignment. Compare the dimensions on the forecast lines with the dimensions the plan nets on. Forecast at site with no warehouse is fine if coverage is maintained at site; it is effectively invisible if coverage was set per warehouse, because the requirements do not meet. Given the mixed excess-and-stock-out picture in the same family, dimension misalignment is my first suspicion, and it explains both symptoms at once.
- Model assignment. The master plan must be pointed at the forecast model that actually holds these lines. A forecast sitting in an unassigned model is simply not consumed, and nobody notices until the season is over.
- Forecast reduction. Confirm the reduction principle on the item model groups covering the seasonal items, so firm sales orders reduce the forecast rather than being added to it. If the seasonal family straddles two model groups with different reduction settings, half the family double counts and half does not - which is precisely how you get excess and stock-out in the same family.
- Buffer versus coverage. For a season with a real demand spike and unreliable supply, decide deliberately whether the protection is a minimum quantity (safety stock via Min/Max on the seasonal items only) or safety margins on a period-based coverage. Do not leave the whole catalogue on Min/Max and hope. Then validate with a forecast plan run and compare planned supply against the forecast by month before switching the master plan over.