When Clinical Demand Becomes A Supply Plan

By Tom Walls, principal and founder, Axon Bridge Consulting

The first article in this series argued that demand in a clinical-stage company is a set of owned assumptions rather than a number. This one is about what happens when those assumptions meet supply. On the surface the mechanics look like any other supply plan: demand in, inventory and capacity out, orders in between. Underneath, almost every step is different, and a planner who imports the commercial playbook unchanged will produce a plan that is confident, tidy, and wrong.

From Subjects To Kits To Batches

The chain starts with a subject and ends with a bioreactor, and the plan is built in the reverse order to the physical flow. Demand planning comes first, then distribution, then finished product supply, then expiry, and only then the upstream drug product and drug substance plans. Each step introduces something a commercial planner does not usually have to deal with.

The unit of forecasting is not the product. It is the product at a depot, because the depot is where demand meets supply and it is the deepest level a sponsor can plan without stepping into territory the randomization system owns. Below the depot, the RTSM allocates kits to subjects and triggers site resupply. Above it, the planning function aggregates depot demand and plans backward to drug substance. Agree on that boundary first. Every other design decision follows from it, and most integration arguments between planning and RTSM vendors are really an argument about where the line sits.

The enrollment curve is the demand signal, and it is a curve, not a line. Sites activate in waves, screening lags activation, and enrollment accelerates as more sites come online. A linear assumption front-loads demand, which means overproducing for a short-dated product and holding inventory that expires before it is needed. Build the curve from screen rate, screen-fail rate, site counts, and activation dates, and let it be slow at the start.

Then the SKUs multiply. A blinded study needs a kit that is its own SKU, with the active, the comparator or placebo, and the packaging all inside one bill of materials. Label text and label countries create additional SKUs of the same drug product. And the blinded kit does not stop at the kit: its demand blows back through the bill of materials to the drug product inside it, and that drug product is often shared by several studies. Plan the kit at the depot, but aggregate every study's demand onto the shared drug product, or manufacturing will be planned against a fraction of what is actually needed.

Country approvals add a timing dimension that commercial planners never see. A single protocol is approved country by country, with each health authority on its own calendar and often with its own conditions: a label variation, a shelf life limit shorter than the sponsor's, a locally sourced comparator, and an import license. Demand in a country does not exist until that approval lands, and supply is not usable there until it meets the conditions attached to it. Both sides therefore have to be tagged. Demand carries the country and the approval date it is gated on; each supply lot carries the label, release status, and expiry position that determine where it can legally be dispensed. Net the two without the tags and the plan will show a depot fully stocked with kits that cannot be shipped to the sites that are enrolling.

Finally, shelf life caps everything. Expiry runs from the drug product manufacture date, not the packaging date, and it is consumed by every step downstream: fill, release, packaging, labeling, distribution, and the days trimmed off by do-not-dispense and do-not-ship buffers. An 18-month shelf life can leave less than a year of usable life at the site. That ceiling, more than any cost calculation, sets how large a finished product batch is allowed to be.

Planning On The Future, Not The Past

Commercial replenishment models lean on history: time series, regression, seasonality. None of that applies. There is no history for a study that has not enrolled, and the history of the last study tells you very little about this one. Clinical supply planning runs on future demand only, from the first unplanned week of the current plan forward.

The practical consequence is that safety stock and order quantities have to be expressed in weeks of supply rather than units. Safety stock is future demand summed over the lead time to replenish. The order target is future demand summed over however many weeks the planner wants one batch to cover. State both in weeks and they flex automatically as the forecast moves; state them in units and every enrollment change means a manual recalculation. Weeks of supply becomes the single most useful metric in the plan, because it shows inventory against the demand it is meant to serve, and it updates itself.

The same logic runs upstream, with one addition. Upstream demand is dependent demand, driven by downstream orders rather than subjects, and it is converted through the bill of materials and the yield. Requirement at the upstream level equals the downstream quantity, times the bill of materials quantity, divided by yield. That last term is where clinical plans are won and lost.

Yield As The Pivot Point

In a mature commercial process, yield is a number with two decimal places that finance has already baked into standard cost. In an early-phase or ATMP process it is a range, and a wide one. A biologic drug substance batch can come in well under plan or well over it, fill yields in early programs vary with modality and operator experience, and cell therapy processes add patient-to-patient variability on top.

A yield assumption is not a planning parameter. It is a joint commitment between planning and manufacturing, and it moves with every process change and every batch. The plan has to be built on an agreed number, that number has to be revisited when each batch completes, and the plan has to be updated when it changes. If planning is carrying 80% and the process engineers privately expect 60, the plan is fiction, and the campaign it commits the CDMO to will be short.

Netting Everything Against Everything

The output of all this is a balance: every supply source netted against every demand stream, at every level of the bill of materials, with expiries treated as a first-class data point rather than a footnote. Supply sources are on-hand inventory by lot and expiry, material in process, material in quality release, purchased comparator, and committed capacity. Demand is the four streams from the first article, in the same unit, on the same horizon.

Two things fall out of the netting that commercial planners rarely have to think about. The first is that inventory is best held as far upstream as possible. A vial of drug product committed to a specific study's blinded kit is spoken for; the same drug product held in bulk can serve any study. If safety stock exists upstream, finished product safety stock can drop to little more than packaging lead time. The second is that the plan should show where material will expire before it is consumed, by lot, months in advance, because an expiring lot in one program is often the answer to a shortage in another, and the window to redirect it closes early.

The balance also has to be expressed in standardized outputs: agreed units, agreed currencies, and the same charts every cycle. Program teams cannot compare a plan they have never seen the shape of before. If the format changes every month, the planning conversation becomes a conversation about the format.

The CDMO Dimension

The supply side of a clinical-stage company is mostly outsourced. That does not outsource the accountability. A sponsor that hands a CDMO a 12-month forecast and waits for batches has not planned; it has hoped. The forecast to a CDMO should run at least 36 months, be refreshed on the same cadence as the internal plan, and be governed at two levels: a tactical monthly review of schedule, yields, and open issues and a strategic review, at least twice a year, of capacity, campaign commitments, and the material cost levers both sides can pull.

The CDMO has no sales history to sanity-check your forecast against and no way to know that your enrollment assumption is optimistic. They will build what you tell them to. That makes the sponsor a steward of the supply chain rather than a customer of it: the yield conversation, the raw material risk, the commitment window before which a campaign can still be cancelled – all of that stays with the sponsor. The stewardship mindset is the difference between a supply plan the CDMO executes and a purchase order they fulfill.

Everything in this article assumes the demand number holds still long enough to plan against. It will not. The third piece in this series is about what happens when the forecast changes, which in a clinical program is a description of every month.