Use demand that matches the replenishment decision
Demand should be measured in the same sellable unit and time interval used by the replenishment model. Exclude cancelled orders, separate bundles when their component consumption differs and account for stockout days that suppressed observed sales. A simple recent average may be useful for stable products, while seasonal or promoted products need a comparable-period forecast.
Separate baseline demand from planned events. A promotion, new channel or price change is a scenario, not historical proof. Keep a baseline forecast and add an explicit event adjustment so the team can later compare which part of the forecast was wrong.
Measure lead time from order release to availability
Lead time should include supplier confirmation, production, origin handling, transport, customs, destination delivery, receiving and quality release. Using transit time alone creates a reorder point that is systematically late. Record actual order dates and available-for-sale dates to build an observed lead-time distribution.
Use separate lead times for suppliers, lanes and service levels that behave differently. Maximum lead time should represent a plausible operating case, not a one-off shutdown that would create excessive stock indefinitely. Exceptional risks are better handled as named scenarios with an expiry date.
Reorder point and safety stock answer different questions
Average lead-time demand estimates inventory consumed while a normal replenishment is in progress. Safety stock is the additional buffer for demand and lead-time uncertainty. Reorder point adds the two. Inventory position, not only on-hand stock, should normally trigger the decision: on hand plus confirmed inbound minus committed demand.
The maximum-average safety-stock method is transparent and conservative when detailed distributions are unavailable. Statistical service-level methods can be more precise, but only when demand variability, lead-time variability and independence assumptions are understood. A more complex formula does not repair poor input data.
Order quantity must include operating constraints
Classic EOQ balances order-processing cost with annual holding cost under stable demand, immediate replenishment and no shortages or quantity discounts. It is a useful reference, not an automatic purchase order. Round the result to supplier minimums, case packs and pallet quantities, then test warehouse capacity, shelf life, cash and forecast risk.
A large discount can still destroy value if it creates markdowns, obsolescence or cash pressure. Compare the discount with additional holding cost and a downside demand scenario. For constrained cash, rank purchases by expected contribution protected per unit of cash rather than by revenue alone.
Worked replenishment example
Suppose average demand is 20 units per day and normal lead time is 18 days, so average lead-time demand is 360 units. A reviewed safety-stock estimate is 140 units, producing a reorder point of 500 units. If on-hand stock is 430, confirmed inbound is 180 and committed orders are 150, inventory position is 460, below the reorder point, so replenishment review is triggered.
The purchase quantity still needs a target. If the operating policy covers another 45 days after receipt, target cycle stock is 900 units before pack-size and cash constraints. Subtract projected stock at arrival, then round to a valid supplier quantity. This separates the timing decision from the quantity decision.
Monitor whether the policy works
Review forecast error, stockout days, service level, excess stock, lead-time variance, receiving variance and cash tied in inventory. Inventory turnover and GMROI add financial context, while sell-through helps diagnose a specific receipt or season. No single metric should reward availability at any cost.
- Update high-volume and long-lead-time products more often.
- Record overrides and the reason for each exceptional purchase.
- Expire temporary buffers when the named risk has passed.