In short: Coverage horizon in commodity procurement timing should be set by how long your own selling price is fixed, so an annual price list needs cover closer to a year than to the eleven weeks a supplier happens to quote. The price decision and the physical delivery decision are separable, which lets coverage extend without extending storage, working capital or a bet on your own demand. Laddering into monthly tranches removes the chance of committing everything at the worst moment and leaves you fully exposed to a sustained move in the market. Score the buyer against the average price over the period rather than against the annual low, because a coverage policy exists to bound risk rather than to reward timing skill.
The buyer covers eleven weeks forward. He has covered roughly eleven weeks forward for four years, because that is how far ahead his supplier will quote a firm price without asking for a commitment he does not want to make. Nobody chose eleven weeks. It emerged.
Meanwhile the commercial team agreed the annual price list with three of the four largest customers in October, fixed until the following October. The raw material behind those lines is bought eleven weeks at a time, and the plan for the remaining forty-one weeks is that the market will probably be alright.
When the market moves 22 percent, which agricultural commodities do regularly, the conversation that follows is about the buyer's timing. The problem is upstream of the buyer, in a mismatch between how long the business has committed its selling price and how long it has secured its buying price, and nobody has ever written that mismatch down as a number.
Coverage horizon should be set by how long your selling price is fixed
Here is the question that reorganises the whole discussion. For each stream of volume, how long is your selling price fixed?
Volume that reprices monthly needs very little cover, because a movement in raw material passes through before it does much damage. Volume sold on a twelve month price list needs cover approaching the length of that list, because for twelve months you have converted a variable cost into something you have to absorb.
Put the arithmetic on it. A processor where raw material is 47 percent of cost of goods, running a 19 percent gross margin, with 60 percent of volume on annual price lists. The input moves 22 percent over the list year and none of it is covered. Cost of goods rises by 0.47 times 0.22, which is 10.3 percentage points, and gross margin on that volume falls from 19 to under 9. On the 40 percent repricing quarterly, most of it passes through with a lag of a quarter and the damage is a fraction of that.
So the coverage policy has a shape, and the shape comes from the sales book rather than from the commodity. Cover the annual-list volume for the length of the list. Cover the quarterly volume rolling three to five months. Leave genuinely spot-priced volume largely uncovered, because covering it converts a matched position into an open one.
The reason this is worth stating explicitly is that a single company-wide coverage percentage, which is what most businesses have, is wrong in both directions at once. It over-covers the volume that reprices and under-covers the volume that does not, and the aggregate number looks reasonable while both halves are mismanaged.
Froot, Scharfstein and Stein made the general argument in the Journal of Finance in 1993 that a firm hedges to avoid being forced into expensive external finance at the moment its cash flow is weakest, and Stulz reframed it in the Journal of Applied Corporate Finance in 1996 around where a firm has a comparative advantage in bearing risk. Neither says a processor should have a view on the market. Both say the coverage should be sized against the commitments the business has already made.
The physical decision and the price decision are separable
The reason coverage gets tangled with logistics is that most people buy both at once. A purchase order fixes a quantity, a delivery window and a price, so extending coverage means extending physical commitment, which means storage, working capital and a bet on your own demand.
Agricultural markets have separated these for a century. Contract structures exist that fix the delivery without fixing the price, or fix the price without fixing the delivery. Deferred pricing and price-to-be-fixed arrangements let you take material on a formula and set the flat price later against a reference. Basis contracts fix the local differential and leave the reference open, or the reverse. Working laid out the underlying logic in the American Economic Review in 1953, in the argument that hedging in commodity markets is about the relationship between two prices rather than about eliminating price risk outright.
The practical value is that it lets you match each decision to its own constraint. Physical timing gets decided by plant requirement, storage and shelf life. Price timing gets decided by the coverage policy above. A buyer who has to move both together will always compromise one of them, usually the price one, because the plant runs out of raw material more visibly than the margin runs out.
What the split leaves behind is the difference between your reference price and your actual delivered price, and how much of that difference to accept is a hedge design question with its own treatment (N17).
Laddering reduces less risk than everyone assumes
The standard defence against timing risk is to buy in tranches. Take a twelfth of the annual requirement each month and your average cost becomes the year's average, which removes the possibility of committing everything at the worst moment.
It does that, and the size of the benefit is usually overstated by a wide margin, because it assumes the twelve monthly prices are independent draws. They are strongly autocorrelated, since a commodity that is expensive in March is usually still expensive in April.
The arithmetic is the same as for any average of correlated variables. With a monthly price standard deviation of 11 percent and an average pairwise correlation of 0.8 across the twelve months, the standard deviation of the twelve-month average is 11 percent times the square root of one plus eleven times 0.8, all over twelve, which comes to 9.9 percent. Laddering across a whole year took the variability from 11 percent to 9.9, a reduction of about a tenth.
Now run the same calculation with a correlation of 0.2, which is closer to what you see across a crop-year boundary or in a market with two independent supply regions. The factor becomes the square root of 3.2 over 12, which is 0.52, and the variability falls from 11 percent to 5.7. Halved.
So laddering works when the tranches sample genuinely different information, and it mostly does not when they sample the same slow-moving market twelve times. The correlation is measurable from your own price history in ten minutes, and it tells you whether your tranche policy is buying risk reduction or just buying activity. Where it is the latter, the honest alternative is a smaller number of larger decisions taken at moments that matter, which is a harder discipline and at least an intentional one.
The crop year has a structure worth planning around
Agricultural commodities carry a seasonal shape that industrial inputs do not. Supply arrives in a window, so the market typically prices a carry through the crop year to pay for storage, and the shape breaks down in years when the crop is short and the market inverts to pull material out of store now.
Two things follow for a buyer. The forward curve is telling you something about expected availability rather than about direction, and reading a carry as a forecast of higher prices is a common and expensive misreading. And the choice between buying forward and buying now to store is a comparison you can do arithmetically, because the market is quoting you the price of storage. If the six month carry is 8.40 a tonne and storing it yourself costs 9.60 in space, finance and shrink, the market will store it more cheaply than you will, and the working capital stays in the business.
Origin is the other lever the calendar hands you, and it is the one most often unavailable in practice. A southern hemisphere crop harvests six months out of phase with a northern one, so a buyer with two qualified origins has two harvest windows a year and a genuinely different coverage problem from a buyer with one. Whether the second origin is usable depends on whether the plant runs it without a process change and whether the customer specification allows it, which is a technical question answered long before the price becomes attractive. A price advantage on an origin your process cannot take is information rather than an option, and treating it as an option is how a buyer ends up explaining a rejected consignment.
The other structural feature is the crop-year boundary. Old crop and new crop are different commodities in an important sense, and a coverage policy that runs smoothly across the boundary is quietly assuming a relationship between them that only holds when both crops are adequate. Tomek and Robinson's Agricultural Product Prices, a standard reference since 1972, sets out the seasonal and storage relationships that govern this, and the part worth internalising is that the boundary is where the correlation assumption behind laddering is weakest, which makes it the one place a tranche genuinely diversifies.
Score the buyer against the period average
Every procurement review has the same failure. Someone points at the annual low, observes that the buyer did not achieve it, and the buyer explains why. This teaches the organisation that timing skill is the objective, which is the opposite of what a coverage policy is for.
The defensible benchmark is the volume-weighted average market price over the period in which the requirement had to be bought. A buyer who achieves that has done the job. A buyer persistently below it has either genuine skill or has been carrying an open position that happened to work, and the way to tell is to look at whether the deviations from policy were authorised in advance.
The measurement that matters more is compliance with the coverage policy itself, reported as covered percentage against target by stream and by month. That number tells you whether the business is exposed. The price achieved tells you how a past decision turned out, and the two get confused constantly in review meetings.
Where this stops
Matching coverage to the price commitment horizon assumes the price commitments are knowable, and in businesses where large customers renegotiate mid-year or where a tender can reset a price list in six weeks, the horizon is itself uncertain. The workable response is to treat the commitment horizon as a distribution too, cover the portion of volume where the commitment is contractual, and keep the rest shorter.
The separability of physical and price decisions also depends on your market. Major grains, oilseeds, sugar, coffee, cocoa and dairy powders have instruments and contract structures that make the split easy. Many speciality crops, most fruit and vegetable inputs and a good deal of the ingredient market have neither a futures contract nor a liquid forward, and there the only coverage available is a physical forward purchase with a counterparty, which brings back the storage and credit questions the split was meant to avoid.
The last limit is organisational rather than analytical. A coverage policy written properly will occasionally produce a year in which the business paid materially above the market and did so deliberately. Surviving that review is the actual test, and policies that have not been signed off by the people who will be in that meeting tend to be abandoned in exactly the quarter they were designed for.
Split your volume by how long its selling price is fixed, compare each stream against how far forward it is currently covered, and bring that one table to the next procurement review.