Portolan Operations decision desk
Open Portolan
All playbooks
Playbook · Inventory

Inventory and replenishment

Inventory is a consequence of policy, lead time, variability and data quality. This path fixes records first, sets policy by item class, then aligns the information and incentives that create bullwhips.

Start a decision

6 steps · 5 cited sources · free during beta

The question this path answers

How much stock should we hold, where, and what should trigger a replenishment?

The path, step by step

  1. Segment items by value, variability and criticality

    Classify items by annual value, demand variability and operational criticality. Assign a policy tier to each segment instead of managing every SKU with the same rule.

    Why it worksInventory effort should follow consequence. Segmentation lets a tight policy sit on the items that carry most value or risk, while slow movers get a simpler, cheaper rule.

    EvidenceSilver et al. (2016) · Syntetos et al. (2005)

    CheckpointSegments covering most of the value and all critical items, with a named policy per segment.

    Common failureApplying one safety-stock rule to every item because the ERP defaults to it.

  2. Fix record accuracy before tuning policy

    Audit system quantities against physical counts by category, trace errors to receiving, returns and cycle counts, and fix those processes before changing reorder points.

    Why it worksA large retailer study found 65% of inventory records inaccurate, with error concentrated by SKU, store and process. Replenishment maths applied to a wrong quantity produces confident nonsense.

    EvidenceDeHoratius & Raman (2008) · Silver et al. (2016)

    CheckpointRecord accuracy measured by segment, with the largest error sources named and owned.

    Common failureRaising safety stock to compensate for records that are wrong.

  3. Set the policy item by item

    For each segment, set review period, reorder point or order-up-to level, lead time and target service level. Document why the parameters differ.

    Why it worksReorder-point and order-up-to policies translate a service-level target and lead-time variability into a decision rule. The parameters only work when lead time and demand variation are measured, not assumed.

    EvidenceSilver et al. (2016) · Lee et al. (1997)

    CheckpointA policy table with parameters, rationale and the service level each item is meant to achieve.

    Common failureCopying a reorder point from a spreadsheet whose assumptions nobody remembers.

  4. Align information and incentives across the chain

    Share demand and stock signals with suppliers and downstream sites, shorten lead times where possible, and stop local order batching that amplifies upstream demand.

    Why it worksThe bullwhip effect comes from forecast updating, order batching, price fluctuation and rationing. Information sharing helps, but its value depends on whether it changes decisions and on lead-time and batch-size improvements.

    EvidenceLee et al. (1997) · Cachon et al. (2007)

    CheckpointOrder variability relative to demand variability, before and after the information change.

    Common failureInstalling a data-sharing portal while every site still orders in monthly batches.

  5. Pilot and measure both sides of the trade-off

    Run the new policy on a controlled group and track fill rate, inventory days, expedites and write-offs together. Do not judge it on stockouts alone.

    Why it worksInventory improvements that only look at availability can be bought with excess stock; improvements that only look at inventory can be bought with service failure. The evidence supports a joint measure.

    EvidenceCachon et al. (2007) · Silver et al. (2016)

    CheckpointFill rate and inventory days move in the intended direction together, with expedites counted.

    Common failureDeclaring success because stockouts fell while inventory and write-offs quietly rose.

  6. Review slow movers and obsolescence

    Review the tail on a fixed cadence: delist, substitute, return or write down items whose demand is intermittent or fading. Do not let a generic service target keep dead stock alive.

    Why it worksIntermittent-demand methods and lifecycle reviews treat slow movers differently from fast movers. Holding the same service level for a dying item locks up cash and space.

    EvidenceSyntetos et al. (2005) · Silver et al. (2016)

    CheckpointA documented tail review with delist/write-down decisions and the cash released.

    Common failureLetting the long tail inherit the service target of the best sellers.

The evidence base

Each step above cites hand-selected research. These are the studies behind the path.

More from the library

Further reading from Portolan's research library, matched by subject. These papers are not proof that any step works; check each checkpoint against your own operation.

  • Çetinkaya & Lee (2000) — Stock Replenishment and Shipment Scheduling for Vendor-Managed Inventory Systems
  • Graves & Willems (2000) — Optimizing Strategic Safety Stock Placement in Supply Chains
  • Kök & Shang (2007) — Inspection and Replenishment Policies for Systems with Inventory Record Inaccuracy
  • CACHON & FISHER (1997) — CAMPBELL SOUP'S CONTINUOUS REPLENISHMENT PROGRAM: EVALUATION AND ENHANCED INVENTORY DECISION RULES
  • Ha (1997) — Inventory Rationing in a Make-to-Stock Production System with Several Demand Classes and Lost Sales
  • Oroojlooyjadid et al. (2021) — A Deep Q-Network for the Beer Game: Deep Reinforcement Learning for Inventory Optimization
  • Graves (1999) — A Single-Item Inventory Model for a Nonstationary Demand Process
  • Heese (2007) — Inventory Record Inaccuracy, Double Marginalization, and RFID Adoption
  • Anily & Federgruen (1990) — One Warehouse Multiple Retailer Systems with Vehicle Routing Costs
  • Deshpande et al. (2003) — A Threshold Inventory Rationing Policy for Service-Differentiated Demand Classes

Guides for this path

Other playbooks