Forecasting the vaccine season spike: stock cover and lead times

Between September and December, gel pack demand from vaccination programmes runs at roughly three and a half times the summer baseline, and the orders arrive with about two weeks of notice. SAP powder has a lead time measured in weeks, not days, and a freezer full of charged packs cannot be conjured. Every year, somebody runs short in the second week of October and pays a premium to somebody who planned in June.

The season is predictable to within a fortnight and it still catches people out, because the forecast is usually built from last year’s orders rather than from the programme calendar that generated them.

Start with the campaign, not the history

Public vaccination programmes publish their start dates, target cohorts and delivery models months ahead. Those three numbers give you a much better forecast than a moving average of your own order book. A campaign that adds a school-based delivery channel changes the pack mix as well as the volume, because school clinics use smaller carriers and more, smaller packs.

Ask your three largest customers for their campaign start date and their carrier format in August. Most will tell you, because a supplier who runs short is a problem for them too. The ones who will not are usually the ones whose own planning is late, and that is worth knowing.

Lead times stack, they do not overlap

A common planning error is to treat the lead time as a single number. It is a chain: purchase order to SAP dispatch, dispatch to your warehouse, warehouse to fill line, fill to freeze, freeze to dispatch. Each link has its own queue, and they do not compress just because you are in a hurry.

Map the chain once with real numbers from last season, then add the queue time you actually experienced rather than the one in the contract. In most plants the gap between contracted and actual is two to three weeks in September, which is precisely the period when everyone else is also ordering.

Freezer capacity is the hidden constraint

Stock cover is usually calculated in packs, but the binding constraint in a spike is frozen volume. A plant that can hold 40,000 charged packs holds them for a few days at most before the freezer becomes the bottleneck for production. Building stock cover without building frozen volume produces a warehouse full of warm packs and a line that cannot charge them.

Work the number backwards: peak day demand in packs, multiplied by the days of cover you want, divided by packs per cubic metre, compared with your usable frozen volume. If the answer is more volume than you have, the cover has to come from production scheduling rather than from stock, which means you need the fill line available earlier, not more inventory.

The two-number safety stock rule

Set safety stock on two numbers, not one: the forecast error you actually experienced last season, and the lead time variability you actually experienced. Most plants hold stock for the first and ignore the second, which is why they run short in exactly the year a supplier has a maintenance shutdown.

A practical starting point is cover equal to the peak week’s demand multiplied by one plus your historical forecast error, plus lead time variability expressed in days of peak demand. It sounds fussy and it is arithmetic you can do on one sheet.

Then decide, in advance, what happens if you are 20% short: which customers get allocation, and at what point you announce it. A plan made in June is a commercial decision. The same plan made in October is a crisis.

If you are building stock cover for a season, the grade-to-application mapping in our SAP powder product range lets you hold one grade across several pack formats, which reduces the number of SKUs you have to forecast separately.

Ring your three biggest customers this week and ask for their campaign dates. The forecast you build from that call will beat anything your order history can tell you.