What Happens When The Clinical Forecast Changes?

By Tom Walls, principal and founder, Axon Bridge Consulting

A clinical supply forecast is wrong the day it is published. That is not a criticism of the forecaster. It is the nature of the inputs. Enrollment location, timing, and arm are uncertain by design. Yields and release lead times vary batch to batch. Expiry is dynamic while material sits on the shelf and stability data accumulates. Regulatory approvals arrive country by country with conditions attached. Any demand or supply assumption can change at almost any time, and most of them will.

The commercial planner's instinct is to fix this with a better forecast. That is the wrong instinct here, because the uncertainty is not in the model, it is in the world. The right response is a planning process that expects the change, absorbs it cheaply, and turns it into a decision. This article is about what that process looks like.

Where The Change Comes From

The sources are familiar to anyone who has run a program. Enrollment runs faster or slower than modeled, and slower is the more common of the two, because sponsors and CROs overestimate site activation and enrollment far more often than not. Sites open unevenly across countries, pulling demand from one depot to another. A protocol amendment adds an arm or expands the sample size. A program is accelerated after a data readout, or paused. A batch fails, comes in short, or a lot is placed on hold pending an investigation. A health authority asks for a tighter specification and only one of four batches meets it.

What distinguishes these events is not their frequency but their asymmetry. Faster enrollment creates stockout risk, and a stockout halts randomization for a whole study. Slower enrollment creates waste, and waste in a clinical program is an expiring batch that cost a million dollars. The plan has to be able to show both risks at once, and the process has to decide which one the company would rather carry.

Refresh, Do Not Rebuild

The first requirement is that a change does not mean starting over. The plan should be refreshed on a monthly minimum cadence, more often during heavy activation or enrollment, and immediately after a failed batch or a major upstream slip. The refresh has three moves, and the order matters.

Actualize first. Replace forecast with fact wherever fact exists: subject demand from the dispensing report, supply from batch release, expiry from any extension granted. Actual demand below forecast pushes the remaining subjects out; above forecast pulls them in. Then run the requirements review from the site end of the chain toward the drug substance end: how does the new demand, the new distribution picture, and the new expiry data change resupply timing and quantity at the regional depot, the central depot, finished product, drug product, and drug substance, in that sequence? Then reconcile in the opposite direction: align every dependent order to the actual quantity and release date of the upstream batch that will feed it, because a short drug substance batch means a smaller drug product order, which means a smaller packaging run, which means a different depot allocation. The two directions loop until the orders and the yields agree.

None of this is hard if the plan was built the way the previous article described. Safety stock and order targets expressed in weeks of supply recalculate themselves. More demand pulls orders in and makes them larger; less demand pushes them out and makes them smaller. The planner's job is to read what the plan is now saying, not to reconstruct it.

Scenarios As The Core Discipline

The second requirement is that "what if" is a routine operation rather than a special project. When a program lead asks what a third arm and 50% more subjects would do to supply, the answer should be available in days, and it should come from a separate version of the plan, never from overwriting the live one. The live plan reflects what is real. A scenario reflects a question.

The most useful scenario technique I know is the low-case-high-case test. Build the plan against the high-case demand, with conservative safety stock and order targets. Build the low case. Copy the high-case production plan into the low-case demand and see what expires. Trim it until nothing does. Copy the trimmed plan back against the high case and see what stocks out. Iterate until you have a production strategy that survives both, and present that strategy with the two triangles, waste and stockout, drawn on the same chart. Leadership can look at that and make a decision. They cannot make a decision from a single forecast with a 30% overage.

That last point deserves its own paragraph. The traditional response to uncertainty is to inflate demand: a large overage, subjects forecast on multiple arms, sites stocked as if every one will enroll. It feels safe. It is the opposite. Inflated finished product demand becomes inflated drug product orders becomes inflated drug substance campaigns, and the bullwhip at the top of the chain is a campaign the company did not need and cannot cancel. Forecast accurately, shrink the overage as the study enrolls, and handle the uncertainty with inventory strategy and scenarios instead.

Before any scenario, though, comes triage. When new information lands, the discipline is to pause and ask six questions: what exactly is being asked, does the plan actually need to be updated to answer it, what are the steps, where does this land in the planning cycle, does the cycle need to be accelerated, and who needs to hear the answer? The second question is the biggest time saver in the list. Much of the time the answer is already in the plan.

When Supply Cannot Flex

Sometimes the scenario comes back and the answer is no. The drug substance campaign is committed inside the CDMO's cancellation window. The comparator is purchased and re-blinded on an 11-week lead time. The stability-restricted batch supports one subject in the two countries that asked for the tighter limit. At that point the planning question becomes a demand shaping question, and it is a different conversation.

Demand shaping in a commercial company means pricing and promotion. In a clinical program it means a conversation with clinical operations about which sites activate first, whether a country can wait for the next batch, whether a cohort can be sequenced differently, or whether an expansion is gated on a campaign that is still four months out. The planner does not decide any of that. The planner's job is to call the constraint out the moment demand outruns supply, quantify the options in the same format the team has seen every month, and get the decision made by the people who own the demand. Waiting until the shortage is visible at a site is the one unforgivable outcome, because by then every option is gone.

The escalation path should be defined in advance. Operationally, a monthly cycle handles most changes. Strategically, a drug substance or drug product commitment point should trigger a separate review that weighs stockout risk against the cost of hedging, at executive level, several weeks before the commitment window closes. The calendar drives the first; the CDMO contract drives the second.

Report The Delta, Not The Plan

The final requirement is about communication, and it is where most planning functions underinvest. Leadership does not need the whole plan every month. They need what moved since last month, why, what it costs, and what decision is now in front of them. A one-page change-over-change view, in the same format each cycle, does more for a planning function's credibility than any dashboard.

It also needs triggers. The R3M framework I have written about elsewhere applies here directly: measure the risk in every item and every assumption, monitor the ones that matter on a cadence, and define in advance what level of movement turns monitoring into mitigation. A yield that comes in short twice in a row is a trigger. Weeks of supply at a depot falling below the replenishment lead time is a trigger. A stability timepoint added without a corresponding change to the demand plan is a trigger. When the triggers are written down, the monthly meeting stops being a status update and becomes a decision forum.

This brings the series to its last question. If a clinical-stage company needs a monthly forum where four demand owners, a mostly outsourced supply side, quality, and finance look at one set of numbers and make decisions about constrained material, then it needs something that looks a great deal like S&OP. Whether it should be called that, and who should own it, is the subject of the final article.

Editor’s Note: You can read Part 1 here and Part 2 here.