In most industries a stockout costs a sale. In pharma it can cost a patient their therapy.
That single fact reshapes the whole forecasting problem. Pharmaceutical demand planning runs under constraints that consumer and retail planners never face: product that expires on a fixed shelf life, supply that moves through a regulated and serialized chain, demand that is shaped by formulary decisions and prescriber behavior rather than price, and a service-level floor that is not a commercial target but a clinical and sometimes legal obligation. Get the forecast wrong on the high side and you write off expensive product against an expiry date. Get it wrong on the low side and a pharmacy or hospital cannot fill a prescription.
This guide covers where pharma demand planning is genuinely different and the framework that holds up against those differences.
What makes pharma forecasting different
Four structural features separate pharma from general demand planning, and each one changes the math.
Shelf-life expiry caps your buffer. In a non-perishable category, the answer to demand uncertainty is more safety stock. In pharma, every unit carries an expiry date, and product that ages out is a write-off — often a large one, because pharmaceutical unit costs are high. You cannot simply buffer your way to a high service level; over-buffer and you trade stockout risk for expiry waste. The forecast has to be accurate enough that the safety stock you hold actually turns before it expires.
Regulated, serialized supply. Lot tracking, serialization, and cold-chain requirements mean supply is not fungible. You cannot freely substitute one lot for another, and a temperature excursion can quarantine inventory you were counting on. The forecast feeds a replenishment process with hard constraints on what can ship, when, and from where.
Demand is shaped, not priced. Pharmaceutical demand responds to formulary placement, prescriber adoption, payer coverage, patent cliffs, and competitive launches — not to the promotional price mechanics that drive CPG and retail. A forecast built on demand history alone misses the structural breaks: a formulary win that doubles volume, a generic entry that collapses it, a competitor recall that spikes it overnight.
Service level is an obligation. For many products, especially in hospital and specialty channels, the target fill rate is not a number a commercial team chose. It is a clinical floor and, for some drug classes, a regulatory one. That floor sets the minimum the forecasting and inventory process has to deliver, which means the cost of getting it wrong is asymmetric in a way most demand planning ignores.
A framework that holds up
The pharma teams that plan well do not treat the forecast as a single number. They structure it around the constraints above.
Forecast the structural breaks separately from the baseline. The statistical baseline handles the stable, in-market demand pattern. Layer on event-based adjustments for the things history cannot predict on its own: formulary wins and losses, patent cliffs, competitive launches and recalls, and seasonal disease patterns for relevant therapeutic areas. Forecasting these as part of one smoothed series guarantees the model lags every structural shift by a quarter or more — and in pharma, a quarter of lag is a write-off or a shortage.
Set service levels by clinical criticality, not by volume. General inventory practice sets safety stock by ABC volume class. Pharma should overlay a criticality tier: a low-volume drug with no therapeutic substitute may warrant a higher service level than a high-volume product with three alternatives on the shelf. Segment the portfolio by both volume and clinical substitutability, then set the service-level target — and therefore the safety stock — accordingly.
Plan against expiry explicitly. The replenishment policy has to know the shelf life. A forecast-driven order that lands product which cannot turn before expiry is a guaranteed write-off, no matter how good the demand forecast was. Tie the order policy to remaining shelf life and projected turn, and flag any position where the projected sell-through runs past the expiry date.
Watch forecast bias closely — it is more expensive here. In pharma, directional bias is punished twice. A persistent over-forecast does not just tie up working capital the way it would in CPG; it actively destroys product at expiry. A persistent under-forecast does not just lose a sale; it risks a clinical shortage. Track signed error by product and by channel every cycle, because the asymmetric cost makes uncorrected bias far more damaging than the accuracy number alone suggests.
What good looks like
A pharma demand planning process is working when three conditions hold together.
Forecast accuracy is measured where the decision is made — by product, by lot-relevant location, and by channel (retail, hospital, specialty pharmacy each behave differently). A national accuracy number that hides channel-level misses is not good enough when one channel carries the clinical service-level obligation.
Bias is near zero and tracked against the asymmetric cost. The planning review should show not just how accurate the forecast is but which direction it leans and what that lean costs — expiry write-off on the high side, shortage risk on the low side.
The process is fast enough to absorb structural breaks. When a formulary decision lands or a competitor announces a recall, the forecast has to re-base in days, not at the next quarterly cycle. A planning process that can only refresh on a slow cadence will always be reacting to the last regime, not the current one.
The implementation reality
Pharma planners know all of this. The obstacle, as in every regulated industry, is that the validated enterprise planning suites built to handle serialization, cold chain, and expiry are heavy to implement and slow to change — and the regulatory validation burden makes every modification a project. The forecasting improvement that would actually reduce write-offs and shortages sits behind a multi-quarter implementation queue.
There is a faster, lower-risk path for the forecasting layer specifically. The Geneva Forecast API produces the demand forecast — algorithm selection per product, structural-break handling, bias diagnostics — through a single API call against the history you already hold, and feeds it into your existing validated replenishment and serialization systems. You improve the number that drives the order without re-validating the system of record. Start with one therapeutic area, prove the write-off and service-level numbers, then expand.
Carrying significant expiry write-offs or service-level risk on a critical product line? Contact RoadMap — we will run a therapeutic area against your history and show you where the forecast is driving both the waste and the shortage risk.
Leave a Reply