Pharma carries more inventory, for longer, than almost any sector, and expiry turns every forecast miss into a write-off rather than a markdown. Here is the economics, with the data.
Sources: The Hackett Group, 2026 Working Capital Survey (2025 data); Moderna FY2023 Form 10-K (SEC).
Pharmaceutical distribution carries more inventory, for longer, than almost any other sector. The Hackett Group's 2026 working-capital survey put days of inventory in pharma at around 251 days in 2025, the single largest drag on the sector's cash. That is not sloppiness; it is structural, driven by cold chain, regulated batches, long import lead times and a duty never to stock out on a critical line. But the same structure means every forecast miss is expensive in a way it is not elsewhere.
In most industries an over-order is a markdown: you discount it and recover part of the cost. In pharma and diagnostics, an over-ordered lot that passes its expiry is destroyed, at full cost. There is no clearance rack for an expired vaccine or reagent. The scale is real: one vaccine maker took a $2.2 billion inventory write-down in 2023, attributed in its own annual filing to shelf-life expiry and inventory held in excess of demand. The lesson for a distributor is that forecast accuracy is not an efficiency metric here; it is a loss-prevention metric. Every unit you over-order against a short-dated product is a candidate for total loss.
Temperature-controlled stock, vaccines, biologics, many reagents, carries both a short shelf life and a cold-chain cost. That combination removes the usual escape hatch of “just order extra to be safe.” Extra units of a short-dated, temperature-sensitive product are not a safety margin; they are a scheduled write-off and an added storage and energy cost. The buffer has to be sized precisely, not generously, which is only possible if the demand forecast and the lead-time variability behind it are reproducible.
Pharma demand rarely trends gently. It arrives in lumps: public tenders, regional buying-group awards, insurance-coverage changes, seasonal campaigns. A moving average smooths those waves away and is therefore wrong exactly when it matters, in the month a tender lands or a campaign starts. Intermittent, tender-driven demand needs a forecasting method built for lumpiness, not one that quietly averages the peaks into the troughs.
Most reagents and many finished drugs are imported, so the reorder decision is made months before the demand is known. The longer and more variable the lead time, the more the forecast has to hold over a long horizon, and the more a small error compounds into either a stockout or a pile of short-dated stock arriving too late to sell. Lead-time variability, not just the average, is what should size the buffer.
A distributor that places analyzers in labs can never fall out of stock on the reagents those instruments consume; a stockout stops a diagnostic line. That service obligation pushes teams to over-stock “to be safe,” which collides directly with expiry. The only way out of the trade-off is a per-SKU buffer sized on that SKU's real forecast error and lead-time variability, high enough to protect uptime, low enough not to feed the expiry pile.
Three disciplines separate distributors that bleed cash on expiry from those that do not. First, a pre-obsolescence view: flag the stock heading for destruction while it still has months of value, not at the annual write-off review. Second, reproducible forecasting of intermittent demand: model the tender and campaign waves instead of averaging them away. Third, per-SKU safety stock on real error and lead-time variability, so analyzer uptime is protected without over-buying short-dated product. Because the stakes are regulatory, the planning also has to be auditable and run on-premise, numbers a quality or finance team can reproduce and defend months later, with data that never leaves the building. This is exactly the discipline DemandIntel brings to pharma distributors, deterministic and explainable, sitting on top of the ERP.
Cold chain, regulated batch and expiry tracking, long import lead times and strict no-stockout obligations on critical lines all push buffers up. Around 251 days of inventory in 2025 (Hackett) reflects that structure, which is also why forecast accuracy matters more here than almost anywhere.
Catch slow-moving and short-dated stock early with a pre-obsolescence view while it can still be sold or redeployed, and size safety stock per SKU on real forecast error rather than a round number. That protects service on the lines that matter while removing the padded buffers that become write-offs.
Yes, with a method built for intermittent, lumpy demand rather than a moving average. The waves (tenders, buying groups, seasonal campaigns) are modelled explicitly so the plan anticipates the peak instead of being surprised by it.
For most health-sector actors it is a requirement, not a preference: patient and commercial data should not leave the organization's walls, and every number must be reproducible and audit-ready long after it was produced.
In pharma, the forecast is not an efficiency metric. It is a loss-prevention metric: every unit over-ordered against a short-dated product is a candidate for total loss.
Picture your next S&OP meeting: one number on screen that nobody argues with. See it free, on your own data, in 45 minutes; and if it does not hold up, we part as friends.