Reading Demand Before It Arrives

Part of Clinical Trial Assistant (CTA) Training: Supply Forecasting and Stockouts.

About 4 minutes. Free, in your language, no account needed.

Why this matters

Supply problems are almost never sudden. They are visible weeks ahead in enrollment rates, visit schedules and shipment lead times, and they become emergencies only because nobody was looking.

Forecasting is not a specialist discipline here. It is arithmetic done regularly.

The inputs

Enrollment rate. How many participants are actually enrolling per site per month, not the planned figure. Use the last two or three months, and treat a site that has enrolled nobody for six weeks differently from one enrolling steadily.

Visit schedule and dosing. The protocol determines how much product each participant consumes and when. A study dosing every four weeks has a predictable draw; a titration design does not, and needs a range.

Current stock on site, and critically its expiry profile. Stock that expires in six weeks is not stock for a participant starting a twelve-week course.

Resupply lead time, measured rather than assumed, and measured per country. The same depot shipping to two countries can have very different real-world lead times once customs is included.

The arithmetic

Weeks of cover equals usable stock divided by expected consumption per week. Usable excludes anything expiring before it could be consumed, and anything quarantined.

Compare weeks of cover against lead time plus a buffer. If cover is less than lead time, a stockout is already inevitable and the only question is how bad.

That comparison is the whole model. It can live in a spreadsheet, and the discipline of running it on a schedule matters far more than its sophistication.

What distorts the picture

Screening surges. A site that suddenly screens many participants converts them to dosing several weeks later. The demand signal arrives before the demand does, and it is the earliest warning you get.

Expiry clustering. Product delivered in one shipment often expires together. A site can look well-supplied and have a cliff.

Returns and destruction in transit. Units already counted at the site but physically gone.

Assumed lead times. The most common distortion, and the easiest to fix: measure what actually happened on the last three shipments to that country.

Reporting a forecast

Report weeks of cover per site, the date cover runs out, and the date a resupply must ship to prevent it. Three numbers. A forecast that reports stock levels without the ship-by date makes the reader do the arithmetic you were supposed to do.

A worked forecast

Site 210, a cardiology site in Pune. Protocol doses every four weeks, one kit per participant per visit.

Current state: 14 kits on site, of which 3 expire in five weeks. 9 participants active. Two more screened last week and likely to enroll.

Usable stock is 11, not 14, because the 3 short-dated kits cannot cover a participant starting a twelve-week course. Expected consumption once the two screened participants convert is 11 participants every four weeks, so roughly 2.75 kits a week.

Weeks of cover: 11 divided by 2.75, so four weeks.

Measured lead time to this site on the last three shipments: 19, 24 and 31 days. Use the worst, 31 days, because customs variability is the thing you are buffering against. That is 4.4 weeks.

Cover is four weeks and lead time is 4.4. The stockout is already inevitable unless a shipment is requested today, and it was inevitable a week ago without anyone noticing.

The three numbers to report

Not stock levels. Cover, run-out date, ship-by date.

"Site 210: four weeks cover, runs out 21 October, must ship by 20 September to avoid a gap. Requesting today."

A study manager can act on that in ten seconds. "Site 210 has 14 kits" requires them to do the work you were supposed to do, and they will not, so nothing will happen until somebody misses a dose.

Key takeaways