Fresh produce has no inventory buffer. What's planted is what's available, in roughly the week it matures, for a few days before it loses its value. If demand isn't there in that window, the crop is disced under. If demand is there and the crop isn't, the buyer sources elsewhere and might not return.

Most growers plan for this from experience. Sales knows roughly what the major accounts took last year; the farm manager knows roughly how many acres that was. They produce a planting plan in the fall. The plan works well enough to stay in business, but it loses money at both ends: acres planted that never sell, and orders declined that could have been filled.

This page describes a quantitative approach. It covers the sources of produce demand, how to build a weekly demand curve, how to convert that curve into planting dates and acreage given yield and maturity uncertainty, and how to revise the plan during the season. It's written for growers who ship to both programs and the open market.

Demand channels

Produce demand arrives through three channels with different behavior. Model each on its own terms; a single combined demand figure loses the distinction between firm commitments and estimates.

Program business. A retailer or foodservice distributor commits to a weekly volume for a defined period, usually at a fixed or formula price. The commitment is rarely a hard contract, and the buyer adjusts it with their own sales, but it's the closest thing produce has to a firm order. Programs are agreed months in advance, which makes them plannable, and they determine most of what growers plant.

Open-market (spot) business. A buyer who is short calls on Tuesday for loads that ship Thursday. Prices move with supply, sometimes sharply. There's no advance commitment, so the only way to plan for spot is statistically: in a given week of the year, this much spot demand has historically appeared at roughly this price. Some growers plant for spot deliberately, betting on a supply gap. Most treat it as an outlet for what programs didn't absorb.

Processing and value-added. A salad processor needs raw product every week and contracts for it, but specifications differ from fresh-market cartons and prices are usually lower. For growers who serve both, the processor is often where off-grade or surplus fresh-market product goes, so processor demand depends partly on how the fresh-market plan performs.

Build the weekly demand curve

The unit of demand planning for produce is the ship week, not the month or the season. A program that averages 40 loads a week for 12 weeks is different from one that takes 20 loads for 6 weeks and 60 for the next 6, even though the totals match. A planting plan that satisfies one misses the other.

For program business, derive the curve from the commitments and adjust by history. If an account has historically taken 90% of its stated volume in shoulder weeks and 110% around a holiday, plan for those figures rather than the stated ones. This requires shipment history by account and week, which most farms have in their sales ledger and few have loaded into a planning tool.

For spot business, the curve is a distribution, not a value. For each ship week, history gives a range of volumes the farm has sold on the spot market and a range of prices. Choose a point in that range to plant for. The choice is a risk decision: planting for the median spot volume means being short half the time; planting for the 80th percentile means having surplus most weeks and being paid for it only when the market is tight. Make the choice explicitly, once, rather than implicitly across 40 block-planting decisions.

Layering the channels produces a total demand curve by week with the firm and speculative portions shown separately. That curve is the target for the planting plan.

Convert demand to planting dates and acreage

Converting a weekly demand curve into a planting plan means working backward from each ship week through harvest timing, maturity, yield, and pack-out to a planting date and an acreage. Each step has uncertainty, and the plan must carry it rather than hide it.

Harvest timing. A crop that ships in week 20 must be harvested in week 20, or a few days earlier if it holds in the cooler. Working back through the variety's maturity, adjusted for expected heat units, gives a planting window. For spring lettuce the window might be four or five days wide. Planting outside it makes the crop early or late for the week it was meant to supply.

Yield. Demand is in cartons; the plan is in acres. Expected yield per acre is the conversion, and it isn't constant. It varies by variety, ranch, time of year, weather that hasn't happened yet, and market conditions at harvest: a crew cuts deeper when the price is high and leaves more behind when it isn't. A farm with block-level yield history can build a distribution for each combination that matters. A farm without it uses a single planning yield and gets surprised.

Pack-out. Not every carton cut ships as a fresh-market carton. Some goes to processing, some is culled, some is downgraded. Fresh-market demand must be met from fresh-market pack-out, so planted acres must cover demand divided by the pack-out fraction.

Combining these gives, for each ship week, a planting window and an acreage expected to meet demand at a stated probability. Those weekly targets are then assigned to specific blocks based on ground availability, rotation constraints, and crew location. That assignment is the planting schedule, which is covered separately. The demand plan is what the schedule is trying to achieve.

Set coverage per account

Growers overplant programs because the two possible errors aren't symmetric. Being short means a buyer who counted on product doesn't get it, fills the gap from a competitor, and reduces or cancels next year's program. Being long means discing a few acres or selling them at a loss on the spot market. The second is expensive; the first can end the relationship.

The right amount of overplanting is a calculation, not a habit. It depends on yield variance, the variance of the buyer's actual take against their commitment, the value of the account, the cost of the acres, and the expected spot price for surplus that week. A grower who overplants 15% across the board is overplanting too much for stable accounts on predictable ground and too little for volatile accounts on new ground.

The demand plan should show, for each program and week, the planned acreage, the expected surplus or shortfall at several yield outcomes, and the expected cost of each. Choose a coverage level per account rather than applying one rule everywhere, and see what each choice costs in expected disced acres.

Revise the plan during the season

A demand plan made in November is a forecast. Revise it as three things change:

  • Demand. A program account increases its take because a promotion succeeded, or cuts it because a competitor undercut the price. A new account signs mid-season. The spot market for week 22 looks tight because a competing region had weather. Each change adjusts the demand curve, and the plan should show which ship weeks are now over- or under-covered.
  • Supply. A block was planted late, so its harvest moved from week 19 to week 20. Another block has a poor stand and its yield estimate was cut by a third. A cold spell slowed everything by four days. Each change moves projected supply between weeks, and the plan should show the resulting gaps and surpluses against demand.
  • Options. Planting windows remain open for late weeks. Ground could be added. A neighboring grower has product to co-pack. The plan should show which corrective actions are still available for each week and what they cost.

The output is a rolling view, updated at least weekly, that shows for every future ship week the projected supply from every block against the projected demand from every account, with a confidence level for each. Where they align, nothing needs to happen. Where they don't, someone makes a decision: eight weeks out while a planting window is still open, not in the week the trucks are loaded.

Required data

Most farms have the data this requires, but few have it organized. The plan needs:

  • Shipment history by account, item, and ship week, for at least three seasons. This is the foundation for the demand side. It usually exists in the sales system but is rarely extractable in this shape.
  • Program commitments as weekly volumes with agreed dates, not season totals. These often live in email and spreadsheets on the sales desk.
  • Block-level yield and pack-out history with planting date and variety. This is the foundation for the supply side. It depends on harvest records carrying block identity and on pack-out records being linked back to them, the same lot tracking that traceability programs require.
  • Maturity models by variety, ideally heat-unit based, plus weather history and forecast for each ranch. Variety data often comes from the seed company and needs local calibration.
  • The current planting schedule with planned and actual dates, so the supply projection reflects what happened in the field.

A farm with all five can build a quantitative demand plan. A farm with three can build a useful one. The most common gap is yield history, and the most common cause is block identity lost between the field and the cooler.

Results

The fall planning conversation changes first. Instead of two people arguing from memory, sales brings a demand curve by account and week with firm and speculative portions separated, and the farm brings supply capability by ranch and week with yield distributions attached. The discussion is about coverage levels and risk, not about whose recollection of last year is correct.

In season, gaps and surpluses are visible weeks in advance instead of days. Sales can find a home for surplus while there's time. The farm can pull a planting forward or push one back while the window is open. Fewer acres are disced, fewer orders are declined, and the orders that are declined are declined deliberately.

Over multiple seasons, each year adds account-level take history, block-level yield history, and plan-versus-actual variance. The overplanting percentage that was a guess becomes a calibrated figure per account. The single planning yield becomes a distribution per ranch and season. The farm relies on a record instead of on the two people who remember.