Planning Around a Six-Week Lead Time: Buffer Stock and Order Cadence

Order on a fixed calendar interval, and size what you hold against the spread of your arrival dates rather than their average. If a consignment typically lands about six weeks after you place it but has twice taken ten, the number that governs your planning is the four-week gap between typical and worst, multiplied by how fast you consume the item — not the six. The average tells you when to expect a parcel. The spread tells you how much has to be on the shelf already for the week the parcel is late, and across most African destinations the spread is the larger of the two problems.

Two honest caveats before the arithmetic. First, there is no published dataset of transit and clearance times for research-reagent consignments into most African countries. The logistics indices that do exist measure general cargo and aviation freight, so they bound the problem rather than answer it, and anyone quoting a confident continental figure is extrapolating. Second, everything below assumes you will keep records, because six of your own receipts will outperform any published average for your own gateway. The long-standing description of laboratory constraint in the region as infrastructural rather than intellectual ^[8] applies exactly here: the limiting factor is usually not the work, it is the interval between deciding you need something and being able to use it.

Why the average lead time is the wrong number to plan against

Arrival times are not symmetrically distributed. Nothing arrives early — a carrier does not deliver in three weeks a consignment it quoted at six — while the delays are open-ended: a missed linehaul, a documentary query, a referral to a second agency, a public holiday at the gateway. The distribution has a hard floor and no ceiling, so the mean sits well below the value you actually have to survive. Planning to the mean guarantees you run out on roughly half of all interruptions, and the half you lose are the long ones.

Published measurements of adjacent flows show how wide that tail gets. The World Bank's 2023 Logistics Performance Index used carrier tracking data to publish aviation import dwell — the interval between cargo being notified as ready and actually being delivered — and found the fastest economies averaging well under a day while the slowest averaged close to ten ^[2]. For sea freight the spread is larger again: a study of six sub-Saharan ports found average container dwell of around twenty days against a few days in most large ports elsewhere, and concluded the difference was not simply administrative failure but partly a rational response to cheap storage and unreliable onward transport ^[4]. Neither figure describes your parcel. Both establish where the variance lives, which is at the border rather than in the air — and a consignment waiting there is also accumulating a temperature excursion someone has to assess on receipt.

The practical move is to stop planning to a date and start planning to a percentile. Decide what proportion of resupplies you are willing to have arrive after you have run out — five per cent is a normal working answer — and size the holding so the other ninety-five per cent are covered. That one decision converts an argument about optimism into an arithmetic problem with a defensible answer.

Where the six weeks actually goes

Before trying to shorten a lead time, split it. The four intervals below are controlled by different parties and contribute very different amounts of variance, and compressing the one you control does nothing at all if the variance sits in one you do not.

That last row is worth dwelling on because it is the one people leave out of their own figures. A consignment sitting unopened in a receiving room because nobody has checked it in is exactly as unavailable as one still on a gateway floor. Lead time ends when the material is in its storage location and in the stock record, with the lot, expiry and condition-on-arrival detail that keeps a batch traceable captured at the same moment — not when the tracking page says delivered.

Sizing the buffer: the arithmetic, with real numbers

Standard laboratory quality management guidance frames stock control as two numbers per item: a reorder level at which a new order is raised, and a reserve quantity that the reorder level is built on top of ^[1]. The reorder level covers expected consumption across the lead time. The reserve — the buffer — covers the lead time being longer than expected. Both are computed from your own records, in whatever unit you actually issue the item in.

Take a worked example. You consume about 20 units of an item per month, which is roughly 4.6 per week, and your last six consignments arrived 5, 6, 6, 7, 10 and 6 weeks after ordering. The mean is 6.7 weeks, the longest is 10, and the sample standard deviation is about 1.8 weeks. Two methods follow, neither of which needs software ^[1].

• Range method. Buffer = (longest observed lead time − mean lead time) × consumption rate = (10 − 6.7) × 4.6 ≈ 16 units. It makes no distributional assumption and is the approach most field inventory systems actually use.
• Statistical method. Buffer = z × standard deviation of lead time × consumption rate, where z is 1.28 for about 90 per cent coverage, 1.65 for 95 and 2.33 for 99. At 95 per cent: 1.65 × 1.8 × 4.6 ≈ 14 units.
• Reorder level. Mean lead time × consumption rate, plus the buffer: 6.7 × 4.6 + 16 ≈ 47 units. Once stock on hand reaches about 47, an order is not upcoming — it is already late.

The two buffer figures land within two units of each other, and that agreement should be read carefully rather than triumphantly. With six observations the statistics are not doing much work, and the range method is defensible precisely because it claims nothing about the shape of the distribution. Be equally explicit about what the arithmetic ignores. It assumes consumption is steady: if your usage is lumpy — three quiet months, then a burst — the buffer must also absorb demand variability, and these figures will understate it. It assumes the six observations describe the same route, the same carrier and the same clearance treatment; change any of those and the history stops applying. And six is a small sample. Treat the output as a starting position revised at every receipt, not as a result.

Fixed-interval ordering beats waiting to run low

There are two ways to trigger an order. Continuous review raises one the moment stock crosses the reorder level. Periodic review raises one on a fixed calendar date — the first Monday of every second month, say — bringing stock back up to a target level. Where transit is fast and freight is cheap, continuous review wins on efficiency. Where a consignment takes six weeks, carries a fixed clearance and documentation overhead per shipment regardless of size, and needs paperwork that is easier to get right when it is routine, periodic review usually wins.

The target level for periodic review has to cover consumption across the review interval plus the lead time, because once you have ordered today you cannot order again until the next review date, and that later order will not land for another lead time on top. With an eight-week review interval on the example above: (8 + 6.7) × 4.6 + 16 ≈ 84 units as the target. On each review date you count what is on hand plus what is already on order, subtract from 84, and order the difference. The discipline is that you order on the date even when the shelf looks comfortable. A small order placed on schedule is the mechanism that keeps the buffer intact; skipping it because stock looks healthy is how the buffer quietly becomes the working stock.

The obvious cost is concentration. Ordering four times a year rather than twelve makes each consignment larger, and a single loss, seizure or clearance dispute then removes a quarter of your annual supply instead of a twelfth. Where a route is unproven or an item is critical, split the order: two smaller shipments a fortnight apart on separate waybills convert a total loss into a partial one. That is an insurance premium paid in freight, and it is usually worth paying for the first few orders on a new route, when you have no history to price the risk with.

The costs of holding more

A buffer is not free, and in a hot climate on unreliable mains power it is not merely a capital cost. Three limits bound how large it should be allowed to get.

• Shelf life. A retest period or shelf life is only meaningful in relation to the storage condition it was established under ^[6], and a reserve held for months has to actually sit in the conditions and containment lyophilised material is specified for. Holding more than you will consume inside that period converts working capital into expired material, and an expiry date does not wait for your next review.
• Exposure. Every additional unit held is additional storage-days under the labelled condition, which the model guidance on storage and transport treats as a requirement across the whole chain rather than at its endpoints ^[5]. In economies where a large share of firms report regular electrical outages ^[7], more inventory held cold is simply more material exposed to a single failure.
• Concentration. Money sitting in a store room is money unavailable for the item you did not anticipate needing, and one storage failure costs more when the reserve is large than when it is lean.

Rotation is the cheap mitigation. Issue oldest-first, record lot and expiry at receipt, and check the reserve's own dates at every review — buffer stock that is never issued is exactly the stock that expires, and a store where the reserve sits untouched at the back of a shelf will discover the problem on the day it is finally needed ^[1].

Measure your own border before trusting anyone's average

Every figure above depends on a lead-time history you probably do not have yet. Customs administrations measure the same interval formally: the World Customs Organization's Time Release Study method measures the time between arrival and release and attributes each portion of it to a specific actor, and some administrations publish their results ^[3]. Asking your national administration whether it has conducted one is a reasonable question with a factual answer, and a published study tells you the realistic floor at your own border rather than a regional average that describes somebody else's.

Your own record is simpler and more useful. Five fields per consignment, kept in one file:

• Date the order was placed and the date dispatch was confirmed — this separates supplier performance from everything downstream.
• Date the consignment reached the destination gateway — this separates the carrier network from the border.
• Date customs released it, plus the verbatim reason for any hold, which turns a recurring documentary defect into something fixable once.
• Date it was physically received and the date it entered the stock record, which are usually not the same day.
• Quantity received against quantity ordered, because short shipments distort a consumption history quietly and permanently.

Six entries give you a mean, a maximum and a standard deviation, which is everything the arithmetic above requires. Twelve give you something you can argue with: when a consignment sits for a month, a file showing that the same route with the same paperwork has historically cleared in four days is the difference between a complaint and a case ^[3].

What to do before you have any history

For the first few orders there is no distribution to plan against, so plan against pessimism and shrink it as evidence arrives. A defensible provisional position: assume a lead time of roughly double whatever is quoted, hold a reserve equal to one full quoted lead time of consumption, order smaller quantities more often than you eventually intend to, and treat the first four consignments as measurements as much as supplies. Frequent early orders generate observations faster and spread the risk of a first-time route failing outright.

Then let the numbers pull the reserve back down. If four consecutive consignments land within a week of one another, the tail you were insuring against is narrower than you assumed and the buffer can shrink. If one lands three weeks late for a documentary reason you can correct, correct the document rather than inflating the buffer: an avoidable delay is a defect, not variance, and paying for it in permanent inventory is the expensive way to solve it. Where a national requirement genuinely cannot be confirmed from a primary source, plan the interval around the documented process and the responsible agency rather than around an assumption about the rule.