It is Tuesday, the strawberries came in light, and three retail accounts are all expecting a full order on Thursday. Sales has already said yes to two of them. The biggest account has a weekend promotion built around this week's volume. The lot that would have covered the gap is sitting on a quality hold while the lab confirms a result. And the product will not wait. Whatever you cannot move in the next few days becomes a markdown or a bin.
None of this is a mistake anyone made. It is a normal short week in a business that grows, packs, or makes food. The supply arrived the way weather and biology decided, not the way the forecast hoped. So the question in the room is no longer what demand will be. It is who gets served, who gets less, who gets a substitute, and what you tell each of them today, before shelf life, service levels, and margin all move against you at once.
That is the real problem behind demand and allocation planning for perishable products, and it is why a spreadsheet rebuilt every Monday never quite fixes it. Sales wants to protect the order it promised, operations wants a stable plan, finance is watching the margin, and quality is holding a lot. Everyone is working from a slightly different version of the number, while the one clock nobody can pause keeps ticking on the fruit.
This guide is for the people who own that call. Commercial and planning leads, demand and sales-and-operations planners, sales operations, and the COOs and founders of fresh-produce suppliers, grower-packers, and food manufacturers who sell perishable product into retail, foodservice, or distribution. It is about the commercial decision once supply is set: given what you actually have this week, what do you promise, to whom, and when. Deciding what to grow and pack, or when to buy more stock, are their own workflows in food and agriculture. Here you are working with the supply you already have.
The job is the allocation call, not the forecast
Demand planning in food and agriculture usually gets described as forecasting, so it helps to keep two things apart. A forecast is a prediction about what demand or yield will be. Allocation is a decision about who gets the finite product now that reality has changed. In a short week the forecast is already history. You can have an excellent forecast and still make poor allocation calls, because the forecast does not tell you which account to protect when the crop comes in light or a lot fails grade.
The output of this workflow is a decision, with an owner, communicated to the customer and to the internal teams it affects. Analysis that never turns into a committed promise, or a promise nobody hears in time, has not done the job it exists for. If the work stays in spreadsheets and side conversations, the team either overpromises, wastes product, gives up margin, or disappoints a customer with no clear reason why.
Pick the one allocation decision that hurts most
The instinct when allocation feels chaotic is to reach for a bigger system: a new planning tool, an ERP module nobody switched on, a forecasting platform someone demoed last quarter. That almost always makes things worse first, because you end up modeling every SKU and deciding nothing.
Start smaller. Pick the single product family and the week-shape where getting it wrong costs the most. A short premium-grade week where two accounts both want the same pallets. An over-supplied flush you have to move before it ages. A contract you cannot afford to miss. Each of those is a real decision someone already makes, usually under time pressure and with incomplete information. Choosing one gives the work a finish line, so you can tell whether allocation actually got better, and it forces the honest question underneath any planning software: when this call has to be made on Wednesday at two in the afternoon, what does the person making it need in front of them?
Separate demand types before you argue about the number
Teams often argue about one demand number when the real problem is that the number is a mix of very different promises. A standing contract order is not the same promise as a promotional lift, a forecast placeholder, a spot inquiry, or a soft customer hold. Treat them all as one figure and you will either overcommit against demand that was never firm, or waste product holding stock for a promotion nobody confirmed.
The first useful move is to label each piece of demand by type and confidence. That is most of the work. It turns "we have 400 of demand against 300 of supply" into something you can actually decide from: this much is firm commitment, this much is confirmed and dated promotion, and the rest is forecast and spot you are allowed to release.
| Demand type | What it actually commits you to | How it should behave when supply is short |
|---|---|---|
| Standing or contract order | A volume you agreed to serve, often with a service level or a penalty behind it | Protected first, up to the contracted quantity; anything above that is negotiable |
| Promotional demand | A lift tied to a retailer promotion with a fixed on-shelf date | Protect only once the promotion is confirmed and dated; do not hold stock for a promo that might not run |
| Forecast or plan | A placeholder for likely demand, not a promise to anyone | Fills after committed demand; the first thing you release when supply is tight |
| Spot or opportunistic | An inbound inquiry, usually a better price, no relationship debt | Served from genuine surplus only, priced for the risk of the week |
| Soft hold or booking | Reserved but not yet confirmed by the customer | Time-boxed; released automatically if it is not confirmed by an agreed cutoff |
Once the demand is split this way, most short-week panic drains out of the room. A shortfall against firm commitments is a real problem that needs a decision. A shortfall that only touches forecast and spot demand is not a crisis, it is a release. The team stops debating the size of the number and starts deciding which promises to keep.
Follow one short week from order to commitment
Before you change any tool, trace how a real short week actually moves, not the way a process document says it moves. Pick one product and follow it forward through the week.
In a lot of perishable businesses the path looks something like this. An order lands from a retailer portal or a sales sheet. Someone checks it against what the pack floor expects to have, using yield estimates that may already be a day old. Sales, wanting to keep the account happy, has already verbally promised two customers a full order. Quality places a hold on a harvested lot pending a result, which quietly removes part of the usable supply. Finance notices the spot order carries a better margin than the contract one and asks why the contract gets priority. And the customer conversation finally happens from whatever version of the number the person on the phone happened to print.
The point of tracing this is not to assign blame. It is to see where the commitment loses its basis. A promise gets made against supply that was only an estimate, before a hold posts, and the customer hears a figure the floor cannot actually deliver. Most allocation pain is not bad math. It is commitments made in parallel, off different versions of the truth, with no single place they ever reconcile before the truck is loaded.
The minimum better version is an allocation board
The first improvement is not a full inventory system or a batch-tracking screen. Those are their own projects. It is a single allocation board that, for the constrained items only, holds the few things the decision needs side by side: the product and the usable available quantity after holds and shelf-life screens, demand broken out by type and confidence, each account's priority and any contract commitment, the margin context, the shelf-life window, the recommended call, the owner, and whether the customer has been told yet.
Keep the board focused on the handful of at-risk items, not every line you sell, and let it surface the conflict rather than just the numbers. When usable supply, committed demand, account priority, and shelf life sit together, the planning meeting can start at the decision instead of spending its first half-hour rebuilding a shared picture from four inboxes. The value is the one sentence those facts add up to: you are short against your own promises, and you have to choose who gets served in full, who gets less, and what you tell them today.
Keep the first version deliberately plain. A board a commercial manager can open on a Wednesday and act on beats a dashboard that needs three people to explain it, and the value shows up when the same view works every week without a scramble to assemble it.
Write the allocation rules before the short week arrives
The worst time to decide how you ration product is in the middle of a short week with three accounts on hold and a promotion counting down. Agree the rules while it is calm. Who ranks ahead of whom, what a contract actually guarantees, whether you allocate strictly by priority or fair-share within a tier, when a substitute is on the table, and where the markdown line sits.
| Situation | The rule to agree in advance | Who owns the call |
|---|---|---|
| A contract order and a higher-margin spot order both want the last pallets | Serve the contract commitment first; spot is filled only from genuine surplus | Commercial lead |
| Two priority accounts, not enough for both in full | Fair-share within the tier, or protect the strategic account by an agreed policy | Commercial lead with the account owner |
| Aging stock with three days of shelf life left | Move it to a faster channel or mark it down before it becomes waste, on a trigger set by days to expiry | Commercial with finance |
| A grade shortfall on a premium order | Substitute a cleared lot with the customer's sign-off, or short the order with early notice | Sales with quality |
| A promotion at risk of running empty | Confirm the promotion is live and dated before reserving stock against it | Account owner |
Substitution and markdown belong in the same set of rules, because they are the same allocation call made from the other direction. Substitution moves demand onto product you actually have. Markdown moves product you must clear onto whoever will take it fastest, trading margin against waste. A supplier that has agreed its markdown triggers in advance, by days to expiry and by channel, turns a panic discount into a planned one and usually keeps more margin than a business deciding it live at four on a Thursday.
A worked example: a short strawberry week across three accounts
Say a fresh-produce supplier sells strawberries into three retail accounts out of a single packhouse. This is an illustrative scenario rather than a real client, and the numbers are there to show the shape, not to report a result. On Monday the plan looks fine: firm orders in, fields expected ready midweek, both grades covered.
By Tuesday, three things have moved. The midweek field came in lighter than planned, so usable premium-grade supply is down. Quality has put a hold on a harvested lot pending a lab result. And Account A, the largest, has a live weekend promotion that needs full premium volume on the shelf by Thursday. Account B is a steady weekly contract. Account C is a spot buyer paying a premium this week because the market is tight everywhere.
Individually, each of these is a phone call. Together they are an allocation decision: there is not enough premium-grade product, ready at the right time, to serve all three the way each would like. Someone has to decide what to protect. The useful version of the workflow does not hide this behind a green dashboard. It puts the conflict on the board, with the call, the owner, and the reason explicit.
| Account | What they expect | The call | Why |
|---|---|---|---|
| Account A (promotion) | Full premium volume for a live, dated weekend promotion | Serve the promotion volume in full; short the discretionary top-up | Confirmed and dated, with real relationship and penalty risk if the shelf runs empty |
| Account B (contract) | Its standing weekly order | Serve the contracted volume; hold the extra it sometimes asks for | Contract commitment is protected first; the extra was forecast, not a firm promise |
| Account C (spot) | A hundred cases at this week's premium price | Offer forty from surplus, or a substitute grade at an agreed price | Served from genuine surplus only; no contract or relationship debt behind it |
Framed this way, the Wednesday call takes ten minutes instead of an hour, because everyone is looking at the same three decisions rather than reconstructing the week from their own inbox. Each customer gets a clear, early update instead of a surprise short-ship at the dock. The plan changed, but the change is visible, owned, and communicated, which is about all a supplier this exposed to weather and biology can honestly promise.
Fit the data to the decision, and mind its shelf life
Perishable businesses are rarely short of systems. There is usually an ERP, sales orders, a CRM, inventory or a WMS, a quality system, a forecast, maybe retailer portals and a BI layer on top. The mistake is starting from the system with the most fields and asking what it can show. Start from the allocation call and ask what it needs.
For a weekly allocation decision, the short list is usually the usable available quantity after holds and shelf-life screens, demand by type and confidence, each account's priority and contract terms, a margin estimate, the promised date, and whether the customer has been updated. Most of that already exists somewhere. The work is pulling the few fields that actually change the call into one current view, not integrating every system end to end.
Timing is the part teams underestimate. Data has a shelf life just like the product does. A quality release or a revised yield that lands after you have already committed the week is a postmortem, not planning information. When you map the data, map when each input arrives as carefully as where it lives, because a perfectly accurate number that shows up an hour after the promise is made changes nothing.
Close the loop with the customer
The allocation call is not finished when the room agrees. It is finished when the customer knows. This is the step teams quietly drop: the decision gets made, and the category manager at the retailer hears about it when the short-ship arrives. The real output of the whole workflow is the updated commitment, sent early, in a form the account can act on.
A good board keeps the communication status next to the decision, so "decided" and "customer told" are two different states and nobody assumes the second just because the first is done. When you have to short an order, an early call to retail procurement with the revised quantity and a substitute offer protects the relationship far better than a full truck that turns up wrong. Give the person making the call the language to use: what changed, what you can serve, what you are offering instead, and by when. A confident, early update reads as a supplier in control. A late surprise reads as a supplier who lost the plot, even when the underlying shortfall was exactly the same.
Where AI helps, and where it must not decide the allocation
Once the board has some structure, AI can take real work off the desk, mostly by turning messy inputs into signals. Grower texts, quality updates, and sales emails often arrive as free text, and a model can read those and pull them into structured signals: this lot released, this account moved its order, this promotion is confirmed and dated. It can compare the week's committed demand against usable supply and flag where the two do not reconcile, draft the customer update from an approved decision, and search prior short weeks so a planner can see how a similar constraint was handled before.
The rule that keeps this safe is simple. AI prepares the allocation; people make it. A model should never decide which account gets shorted, release a held lot, or send a commitment to a customer on its own. The judgment about margin, service, and a relationship you plan to keep for years stays with the person who owns the outcome. If a suggested allocation cannot be traced back to the demand and supply it came from, do not act on it.
What usually goes wrong
Most demand and allocation planning breaks in a handful of predictable ways. Naming them makes them easier to catch before they cost you a shipment or a relationship.
| What goes wrong | Why it happens | What to do instead |
|---|---|---|
| The team argues about one demand number | Standing, promotional, forecast, and spot demand get added into a single figure | Split demand by type and confidence before you allocate anything |
| Sales promises move ahead of supply and quality | The customer call happens off a printed number, not the current board | Make every commitment from the same board that holds usable supply and holds |
| Shelf life is ignored until it forces a markdown | Aging stock is not on the allocation view until it is nearly waste | Put days to expiry on the board and agree markdown triggers in advance |
| Nobody owns the call when supply is short | The "who gets served" decision has no owner, so it is made late or by default | Agree priority tiers and decision owners before the short week hits |
| Substitutes are offered inconsistently | Each rep improvises a substitute in the moment, at a price they invent | Agree substitution rules and capture the customer's sign-off in the workflow |
| The customer hears last | "Decided" is treated as the same thing as "communicated" | Track communication status separately and send the update while it still helps |
| AI is added before the rules exist | A model is layered on top of an undefined allocation process | Define demand types, priority, and owners first; use AI to prepare, not decide |
Set a weekly cadence that protects the allocation call
An allocation function needs a predictable weekly cadence, or every short week turns into a fresh argument about deadlines and who owes what. The point of the cadence is not more meetings. It is to move the team out of collecting numbers and into deciding commitments before the week gets away from them.
At the end of the prior week, confirm firm orders, promotional commitments, and the supply the pack floor expects, and draft the allocation. At the start of the week, lock the base commitments and publish what each account is expected to receive. Midweek, pull the changes into the exception view, the yield revisions, quality holds, and order or promotion changes, so shortfalls become visible while there is still time to act. Then hold a short midweek call to decide the constrained items, reallocate, substitute, or renegotiate, and assign who tells which customer. At the end of the week, note what changed and why, so the same shortfall is not rediscovered as fresh bad luck next season.
The exact days depend on your crop and your customers' order windows. The discipline is what matters: a base set of commitments the team trusts, a designed moment midweek to catch what moved, and a quick decision step so promises change on purpose rather than at the loading dock.
A first 30, 60, and 90 days
If allocation is painful today, do not try to rebuild it all at once. A narrow first version that ships beats a complete one that never does.
In the first thirty days, pick one product family and one short-week shape, and build the allocation board with demand by type, account priority, shelf life, margin, and communication status. Run one real week through it, even if half of it is still a shared sheet. The goal is one week where the commitments came off the board instead of from parallel promises made on the phone.
In the next thirty days, connect the few sources that keep the board current: sales orders, usable supply after holds, quality release, margin, and the retailer portals that carry order and promotion changes. Agree the priority and substitution rules, and tighten the decision and communication steps so the midweek call gets faster and less contested.
In the following thirty days, add AI where the workflow is stable enough to trust it, for variance summaries and draft customer updates, and only then consider widening to more products or channels once the first one holds.
What to measure
A few measures tell you whether the workflow is earning its place. On the process side, watch the constrained items reviewed each week, customer updates sent on time, demand carried too long without confirmation, and repeat allocation conflicts. On the business side, watch the measures the calls are supposed to move: fill rate against committed demand, waste and markdown as a share of supply, margin held through constrained weeks, service misses and short-ships, and aging inventory.
To put a number on the manual cost and judge where a workflow is worth automating, the cost of manual work guide and the Workflow Readiness & ROI Calculator are a good place to start.
How Ubisar would implement this workflow
In week one, Ubisar would pick one planning cycle and one product family where allocation is painful, then map how firm orders, promotional commitments, usable supply after holds, shelf life, margin, and account priority turn into the final call. The first output would be an allocation board carrying demand by type, usable available quantity, a margin signal, the promised date, the owner, and the customer communication status.
In weeks two and three, we would connect the smallest useful set of sales, inventory, quality, and retailer-portal data, then add the priority and substitution rules and the exception view. AI would come in only after the weekly cadence is agreed, to summarize what changed since the last cycle and draft the customer updates, while commercial and quality owners still decide the trade-offs. By week four, the team should be able to run the next short week from the board, decide allocations in the open, and send consistent updates to sales and operations.
At the end of month one, keep going if the board changed commitments before orders went late or wasteful, and stop or narrow the scope if the demand and supply inputs are still too unreliable for one product family. This is a strong fit for AI, Data & Tech Implementation. If allocation is eating your short weeks, tell us the workflow that is slowing you down through contact us.
Where to go next
For more on this sector, the food and agriculture page covers the workflows that come up most often in growing, packing, and food manufacturing. If you are weighing this against production and harvest planning, inventory, or traceability work, the workflow guide library lays them out side by side. If you are still deciding where to start, how to choose the first workflow to improve with AI is a good next read.
If you need to make the business case first, the implementation cost guide helps with budget, and if you are comparing a consultant, an agency, or software, the comparison guide is written for exactly that choice.
Sources and useful references
For background while you design your own allocation cycle, the USDA Agricultural Marketing Service market news at ams.usda.gov/market-news tracks fresh-produce prices and movement, the FAO platform on food loss and waste at fao.org/platform-food-loss-waste covers the waste side of getting allocation wrong, and the Microsoft Dynamics 365 shelf-life planning reference at learn.microsoft.com shows how one ERP models shelf life in planning. Use them as background, and design the allocation cycle around your own products, customers, and constraints.