In short: Seed and planting planning runs on a multiplication cycle of three or more field seasons, so the quantity available this autumn was fixed by a decision taken before anyone knew the variety would sell. What the business commits to is contracted area; what it receives is area times yield times the share that passes certification. Germination is a threshold rather than a gradient, and a lot that decays below the certification minimum loses most of its value at once. The carryover choice between holding surplus seed and selling it into feed is dominated by the probability that the lot fails retest next season, which is a number most operations never compute.
The sales team has a good problem. A variety that came onto the recommended list two autumns ago is being asked for by every distributor in the eastern region, and the orders are running at double the plan. Somebody asks production whether they can make more.
Production can make more in three years. The certified seed being sold this autumn was grown last summer from basic seed, which was grown the summer before from pre-basic, which came from a breeder lot bulked up the season before that. Every one of those steps is a growing season and none of them can be compressed. The quantity available this autumn was fixed by a decision taken when nobody knew whether the variety would perform.
That is the defining feature of this business. The commitment horizon is measured in growing seasons, the demand signal arrives late, and the only real flexibility is at the bottom of a pyramid where the quantities are tiny.
The multiplication cycle is a lead time measured in seasons
Seed moves through generations, and the nomenclature has been standardised internationally through the OECD Seed Schemes since 1958: breeder seed, pre-basic, basic, and then one or two generations of certified. Each generation is a field season, and each multiplies the quantity by a factor set by the crop's seed rate and yield.
Work the pyramid for a cereal with an effective multiplication factor of 25, after allowing for grading losses and rejected lots. To sell 12,000 tonnes of certified seed you need to sow 480 tonnes of basic. To have 480 tonnes of basic you need to sow 19.2 tonnes of pre-basic the season before. To have that you need around 770 kilograms of breeder seed the season before that.
Three quantities and three years. And notice how the decision changes character as you move down the pyramid. At 12,000 tonnes you are managing a commodity operation. At 770 kilograms you are managing a bet on a variety that has not yet been tested commercially, in a quantity that fits in a small van, where the cost of over-producing is negligible and the cost of under-producing is three years of a market you cannot serve.
The asymmetry at the bottom of the pyramid is the single most actionable thing in this article. Producing twice as much pre-basic as your central forecast requires costs a rounding error on the production budget. It buys you the option to double the following generation, and then the one after. Businesses that ration pre-basic to the forecast are optimising the cheapest stage of the chain and paying for it at the most expensive one.
Hybrid crops work differently and the logic holds. Parent lines are maintained separately, the crossing block is the constraint, and multiplication factors are lower, which makes the horizon shorter in generations and tighter in capacity. The bet still gets placed years before the answer arrives.
Contracted area is the production order, and yield is the variance
Seed is grown under contract by growers on defined fields, inspected in the field for varietal purity and isolation, and bought back. What you commit to is area. What you receive is area times yield times the share that passes certification.
Two sources of variance sit on top of each other. Yield moves with the season like any crop. The pass rate moves with the season too, because wet conditions raise the risk of admixture and disease, and a field that fails inspection produces no certified seed at all, only grain.
Plan on a distribution rather than an expectation. Contracting area equal to the central yield estimate means falling short in roughly half of years, and the cost of falling short is asymmetric because the shortage lands in a specific variety that customers have already been sold on. A useful discipline is to contract to a quantile rather than to a mean, chosen per variety according to how substitutable it is. A variety with three close alternatives in the portfolio can be contracted near the middle. One with a distinctive disease resistance package that a region depends on should be contracted well above it.
The genuinely awkward feature is that supply and demand are correlated through the weather. A hard winter that kills established crops creates a spring replanting spike, and the spring seed available to serve it was grown the previous summer and is fixed. That correlation makes the buffer worth more than an uncorrelated analysis would suggest, and it is measurable from a decade of your own order intake against winter severity.
Germination is a threshold, and the asset decays toward it
Certification requires a lot to meet minimum standards for germination, varietal purity and analytical purity, tested to the International Rules for Seed Testing published by the International Seed Testing Association, which has set the reference methods since 1924. A cereal lot below the germination floor cannot be sold as certified seed. There is no discount at which that transaction becomes available.
This belongs in the plan as a hard filter applied before any allocation runs, generated from the test certificate on each lot. Never model it as a cost with a penalty, because a penalty is a price and an optimiser will find the case where paying it looks attractive, producing a recommendation that a certification officer will reject. Build the sellable set from the test results, then optimise inside it.
What makes it interesting rather than merely restrictive is that germination declines in storage, at a rate that depends on moisture, temperature and the crop. So every lot is walking toward a cliff at a knowable pace, and the distance to that cliff is a planning variable that almost nobody carries in their inventory record.
The carryover decision collapses near the floor
Surplus certified seed can be held for next season or sold now, often into feed at a fraction of the value. The arithmetic looks simple and it is dominated by a probability most people never compute.
Take 800 tonnes of surplus. As certified seed it is worth 620 a tonne. As feed grain it is worth 210. Germination declines by around two points a year in ambient storage, with a standard deviation of roughly 1.2 points across lots, and the certification floor is 85 percent.
A lot testing 90 percent today will test around 88 next autumn. The probability it clears 85 is the probability a normal variable with mean 88 and standard deviation 1.2 exceeds 85, which is about 99 percent. Expected carryover value is roughly 616 a tonne before storage cost, so holding it is clearly right.
A lot testing 87 percent today will test around 85 next autumn, and the probability it clears the floor is about a half. Expected value is 0.5 times 620 plus 0.5 times 210, which is 415, before storage. Selling it now at any price above 415 beats carrying it, and selling it now at a 15 percent discount to full value, which is 527, beats carrying it comfortably.
The shape of that function is what matters. Carryover value is nearly flat well above the floor and falls off a cliff within three or four points of it, so the seed most likely to be carried by habit, the marginal stuff nobody wants to discount, is exactly the seed where carrying destroys the most value. Sorting the carryover list by current germination rather than by volume changes which decisions get made first.
Variety life cycles make the long-horizon forecast a portfolio bet
A variety typically climbs for two or three years after listing, plateaus, and then declines as newer material with better yield or disease resistance displaces it. Very few varieties hold a market for a decade.
That means the three-year forecast underpinning your pre-basic decision is a forecast of where a variety will be in its life cycle, which no time series method will produce because most varieties have no history at the point the decision is made. The workable approach is analogue-based, using the observed adoption curves of past varieties with similar trial performance and similar market positioning as the shape, which is the general launch forecasting problem (D4) applied to a portfolio where a third of the range is in its first three years at any time.
The demand signal that matters arrives on a known date. National recommended list trials, such as the AHDB Recommended Lists in the UK and their equivalents elsewhere, publish each autumn, and a variety's position in those results moves distributor ordering more than anything your commercial team does. Building the pre-basic decision calendar around the trial publication date, rather than around the financial year, gets you one extra year of information at no cost.
Your largest competitor is the grower's own barn
For most open-pollinated field crops, a grower can keep part of the harvest and drill it next season. That right is bounded by plant variety rights, under the UPOV Convention of 1961 as revised in 1991 and, in the EU, through the farm-saved seed provisions of Council Regulation (EC) No 2100/94, which permit the practice for listed species subject to a royalty. What it means commercially is that a substantial share of the sown area is served by nobody.
The certified seed replacement rate, meaning the share of area sown with newly purchased certified seed, is therefore a demand driver in its own right, and it moves with grain prices, with the age of the varieties in circulation and with how wet the previous harvest was. A five point movement in that rate across 2.1 million hectares at a 180 kilogram seed rate is nearly 19,000 tonnes of demand appearing or disappearing, which is larger than almost any variety-level movement in the portfolio.
Very few seed businesses forecast it explicitly. Building even a crude model of replacement rate against grain price and harvest quality, from whatever regional statistics exist, adds a term that is currently sitting inside the residual and being attributed to sales performance.
Where this stops
The pyramid arithmetic assumes multiplication factors that hold in a normal season. In a poor season the factor drops and the pass rate drops together, so a shortfall at the basic stage propagates into a shortfall at the certified stage two years later, and the propagation is invisible for a year while everything looks fine. A simple annual report of realised multiplication factor by generation and by crop is the early warning, and remarkably few businesses produce one.
Treating each generation as an echelon and placing buffers across them is the natural framing, and the standard multi-echelon methods (I1) assume replenishment lead times measured in weeks with the option to reorder. Here the lead time is a year and there is exactly one opportunity to order per year. That changes the problem enough that the buffers have to be sized against a season-length review period rather than against a lead time, and the results from a system configured with weeks in the lead time field will be badly wrong in a direction that looks reassuring.
The last limit is that the pre-basic hedging argument works only where the field capacity and the isolation distances exist to grow the extra generations. In crops requiring wide isolation or specialised production regions, pre-basic capacity is genuinely scarce and doubling it is a real cost rather than a rounding error. Establish which situation you are in before treating the bottom of the pyramid as free.
Pull the germination certificate for every carryover lot you are currently holding, sort by how many points sit between it and the certification floor, and work the top of that list this week.