Systems and records
Where demand forecast and supplier capacity data loses the decision
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
Demand forecast supplier capacity data teaches a model only when each record links a dated forecast version, the commitment level of every quantity, the supplier's capacity answer, the gap per period, who accepted each part of that gap and what changed next. UN/EDIFACT can carry forecasts and capacity side by side, yet none of the 25 commitment codes in its D.24A directory records who accepted a shortfall or why.
Record timeline
Demand forecast supplier
Key takeaways
- Keep the commitment level of every forecast quantity; it decides whether a gap is the supplier's failure or the buyer's unprotected estimate.
- UN/EDIFACT D.24A offers 25 commitment level codes, five of them for supplier capacity, and none for accepting a shortfall.
- Planning software records the supplier's reason for a commit mismatch; the buying-side acceptance usually lives elsewhere and has to be joined in.
- Split every gap into parts, each with its own response, accepting role and review date.
- Grade reconciliation against what was known on the decision date, apart from forecast accuracy and outcome.
What should demand forecast and supplier capacity data prove?
It should prove that someone compared a dated demand forecast with what the supplier said it could make, saw the gap, and decided who would carry it. A usable record links six things: the forecast version, the commitment level of each quantity, the supplier's capacity answer, the gap computed per period, the person who accepted each part of that gap, and the change that followed.
Many systems keep the first and the last of these and lose the middle. A planning tool stores the current forecast and the current commit, while the version that triggered the decision, the call where a shortfall was accepted and the reason the supplier gave sit in email or a portal note. A model trained on what remains learns that forecasts and commits converge, not how people made them converge.
This page sets out what demand forecast supplier capacity data has to contain for that lesson to survive. It uses the EDI vocabulary that carries schedules between companies, the contract rules that say what a forecast obliges, and one illustrative episode worked through from forecast to outcome.
Firm, authorized or planning: whose problem is the gap?
The commitment level attached to each forecast quantity decides who bears a gap, so the record must keep it per line, not per message. A firm quantity is a promise to take delivery; a planning quantity is an estimate the supplier plans against at its own risk; in between, a buyer can authorize the supplier to buy material or start production without making the finished quantity firm.
Federal procurement rules state the planning end plainly. Under FAR 16.503, the estimated total quantity in a requirements contract "is not a representation to an offeror or contractor that the estimated quantity will be required or ordered," and the contract should state, if feasible, the maximum limit of the contractor's obligation to deliver. The standard clause at FAR 52.216-21 adds that the scheduled quantities "are estimates only and are not purchased by this contract."
Commercial sales law adds a limit that binds both sides. Section 2-306 of the Uniform Commercial Code limits a requirements term to quantities that occur in good faith and bars demands unreasonably out of proportion to a stated estimate. Read together, these rules make the estimate, the maximum obligation and the commitment level three separate fields. A dataset that collapses them into one forecast quantity cannot say whether a shortfall was the supplier's failure or the buyer's unprotected bet.
Twenty-five commitment codes, none for accepting a shortfall
The EDI vocabulary already lets a forecast and a capacity statement travel side by side; what it has no code for is the acceptance. In the UN/EDIFACT D.24A directory, as mirrored in the public php-edifact/edifact-mapping repository, the delivery plan commitment level list (data element 4017) holds 25 codes, and the DELFOR delivery schedule message uses its scheduling conditions segment to attach one of them to the scheduled quantities that follow it.
Sorted by what each code asserts, 6 describe a buyer commitment, 5 an estimate, 5 a supplier capacity statement and 9 other scheduling context. A case-insensitive search of all 25 titles and descriptions for accept, shortfall, gap, reject, approv and reason matched nothing. The capacity available code entered the list in D97B; the three tooling capacity codes are already present in D94B, the earliest version these files hold.
The wider directory supplies the links a reconciliation chain needs. A document name for a delivery schedule response, defined as a message responding to a previously transmitted delivery schedule, first appears in these files in D99A. A quantity qualifier for the maximum supply quantity a supplier endorses appears in D00A, and a reference qualifier for the previous delivery schedule number in D96A. A supplier's answer can therefore point back to the forecast version it answers, while who on the buying side accepted the difference, and why, has to travel as free text or in another record.
Method: a script read only the XML data files at commit c875cda (20 November 2025) for every folder named D plus two digits plus a release letter, 60 versions from D94B to D24A, and counted a code as present when that version's code list holds it. First appears means the earliest version in these files, not the first UN release. The four groups are an editorial map that covers every code with none left over. The groups, with what a reconciliation record must add to each:
| Group (editorial) | Codes in D.24A | Examples from the list | What the record must add |
|---|---|---|---|
| Buyer commitment | 6 | Firm; commitment for manufacturing and material; commitment for material | Who authorized the liability, for which periods, and the date it was given |
| Estimate or forecast | 5 | Planning/forecast; potential order increase; proposed | The forecast version, its issue date and the version it replaced |
| Supplier capacity statement | 5 | Capacity available; tooling capacity; required tooling capacity | Whether the number is a dedicated allocation, a shared line or a tooling limit |
| Other scheduling context | 9 | Promotion; special demand; strike protection | Why the period is unusual, so the gap is not read as normal demand |
| Shortfall accepted, with reason | 0 | No code in the list | Accepting role, authority, response chosen and review date |
How planning software flags a commit mismatch, and what it keeps about the cause
Collaboration software such as Oracle's detects the gap and records its cause from the supplier's side. Oracle's Supply Chain Collaboration guide lists three exceptions in order forecasting: past-due commits, where the supplier has not committed by the commit due date in the service level agreement; forecast changes, where the current cycle's quantity differs from the previous cycle's; and commit mismatches, where the committed quantity is less than the forecast quantity for a given date.
Oracle's REST documentation for those exceptions gives each one a type, commit mismatch or forecast deviation, and a commit mismatch reason code that identifies why the supplier could not commit to the forecast. That is the supplier's explanation. Among the fields checked for this page, none names who on the buying side reviewed the mismatch and chose to live with it, move demand or pay for more output.
The capacity answer needs its own context to mean anything. The FAR responsibility standard at 9.104-1 asks whether a prospective contractor can meet the delivery schedule "taking into consideration all existing commercial and governmental business commitments." A supplier's capacity figure is what remains after its other customers, so the record should say how the figure was built and on what date it was given.
Fourteen reconciliation fields and a response-by-constraint grid
A reconciliation episode is complete when a reviewer can rebuild the decision without asking anyone who was in the room. The fourteen fields below are the minimum. The grid after them tests five common responses against the constraints that most often make the obvious answer wrong.
Every response in the grid is legitimate in some episode. What the record must show is which constraint each was checked against and who signed for the part of the gap it covered. The fields, then the grid:
- Forecast version identifier and issue date, plus the identifier of the version it replaced.
- Item, ship-to site and time bucket for every quantity.
- Commitment level of each quantity: firm, production authorized, material authorized or planning.
- Contract limits: estimated quantity, maximum delivery obligation and any per-period order limits.
- Supplier capacity answer per bucket, with its date and its basis.
- Signed gap per bucket, with shortfalls and spare capacity kept apart rather than netted.
- The supplier's stated reason for any mismatch, as a code and in its own words.
- Options open on the decision date, including doing nothing.
- The response chosen for each part of the gap, in units.
- Who accepted each part, their role and the authority they acted under.
- Cost or liability created: a premium, stranded material or a moved customer date.
- The next forecast version and what moved between versions.
- Realized demand and deliveries for the affected buckets.
- Outcome label: served on time, deferred within customer acceptance, or unserved.
| Response | Contract limit | Material liability | Customer date | Who must sign |
|---|---|---|---|---|
| Supplier extra shift or overtime | Must stay inside the maximum obligation or be amended | Usually unchanged | Protected | Supplier operations, plus the buyer's budget owner for any premium |
| Defer demand into spare buckets | Unaffected | Unaffected | Moves, so customer acceptance is needed | Sales or customer operations owner |
| Second source for part of the gap | Check exclusivity and requirements clauses | Can strand material the first supplier bought under authorization | Protected only if the source qualifies in time | Category manager, with quality for qualification |
| Build ahead in earlier buckets | Unaffected | Raises inventory one party must carry | Protected | Planner, with finance for the inventory |
| Accept the shortfall as open risk | Unaffected | Unaffected | At risk | Named planning owner, with a review date |
Illustrative: closing a 180-unit gap in weeks 3 and 4
In this illustrative episode, a pump maker's forecast for one casting asks for 400, 400, 520, 560, 480 and 440 units over six weeks. Weeks 1 and 2 are firm, weeks 3 and 4 carry a commitment for material, and weeks 5 and 6 are planning quantities. The supplier answers with capacity of 400, 400, 450, 450, 500 and 500.
That leaves a 70-unit shortfall in week 3 and 110 in week 4, 180 in all, against 80 units of spare capacity in weeks 5 and 6. The record splits the decision three ways: the supplier's operations manager accepts a 60-unit extra shift in week 4 and the buyer's category manager approves its premium; sales operations accepts deferring up to 80 units into weeks 5 and 6 for orders with flexible dates; and the planning owner accepts the last 40 units as open risk until the next cycle, because those weeks were not yet firm.
Then the outcome. Realized demand in weeks 3 and 4 was 490 and 520 units, 1,010 against a forecast of 1,080, so the forecast ran 70 units high. Capacity plus the extra shift covered 960, the other 50 moved into the spare weeks, nothing went unserved and 30 spare units were left unused. The deferral assumes the week 5 and 6 forecasts held.
A grader learns three things here that a final-state extract hides: the 40-unit risk was a decision with an owner, not an oversight; the premium had an approver separate from the supplier's agreement; and forecast error and reconciliation quality are different scores. Week by week:
| Week | Commitment level | Forecast | Capacity answer | Gap |
|---|---|---|---|---|
| 1 | Firm | 400 | 400 | 0 |
| 2 | Firm | 400 | 400 | 0 |
| 3 | Commitment for material | 520 | 450 | 70 short |
| 4 | Commitment for material | 560 | 450 | 110 short |
| 5 | Planning/forecast | 480 | 500 | 20 spare |
| 6 | Planning/forecast | 440 | 500 | 60 spare |
Is cutting the forecast to fit capacity a valid answer?
Only when demand really fell; otherwise it closes the gap on paper and hands the shortfall to the customer. A revised forecast that matches the supplier's answer looks like a reconciled plan, and a model rewarded for reaching zero gap will learn to produce one.
The record tells the two cases apart with evidence. A forecast cut should carry a demand reason dated before the cut, such as a cancelled order or a revised customer schedule, rather than the supplier's capacity reply alone. Oracle lists forecast changes between cycles as an exception type of their own, which keeps a moved quantity visible instead of silent.
The fastest answer can fail as well. FAR 52.216-21(e) lets the government buy urgently required quantities from another source when the contractor will not accept an order for accelerated delivery. In a commercial contract, the same move can collide with an exclusivity term or strand material the first supplier bought under a material commitment, which is why the grid tests each response against liability as well as cost.
Scoring a model on reconciliation episodes
Score the reasoning against the constraints known on the decision date, and score the outcome separately. A proposed deferral is right only if the deferred units fit spare capacity, the customer dates allow it and someone with authority is named; it does not become right because demand later came in low.
At a glance
Scoring a model on reconciliation episodes
Five checks cover most of the grading, each answerable from the fourteen fields:
- Commitment use: firm and authorized quantities are treated as obligations, planning quantities as estimates.
- Arithmetic: gaps are computed per bucket, with shortfall and spare kept apart rather than netted across the horizon.
- Constraint check: each proposed response passes the contract limit, liability and customer date tests in the grid.
- Ownership: every part of the gap has an accepting role and a review date.
- Change discipline: any forecast cut carries a dated demand reason separate from the capacity reply.
Sourcing reconciliation episodes, and where this design stops
Ask for episodes, not tables: each forecast cycle in which a commit fell short, with the versions before and after and the decision trail between them. SourceX sources this kind of data from U.S. businesses against a buyer's written specification rather than selling from stock, so a request should name the item types, the planning horizon, the number of cycles and the acceptance test the episodes must pass.
The design has limits. The vocabulary count covers one public mirror of the UN directories, not the implementation guides trading partners actually exchange, which often narrow or extend the code lists. The Oracle fields are documented software behavior, not evidence that any company fills them in. The six-week episode is illustrative, and no company's records were analyzed for this page.
Frequently asked questions
What is the difference between a demand forecast and a supplier commit?
A demand forecast is the buyer's estimate of what it will need, usually with only the nearest periods made firm. A supplier commit is the supplier's statement of what it will make or ship in each period. The difference between them is a decision someone has to make, and a useful record shows who made it.
Which EDI message carries forecasts and capacity answers?
In UN/EDIFACT, the DELFOR delivery schedule message carries scheduled quantities with a commitment level code for each group, including codes for planning quantities and for capacity available. The directories also define a delivery schedule response document. Trading partners' implementation guides decide which of these codes they actually exchange.
Should superseded forecast versions be kept?
Yes. The decision was made against one specific version, and later cycles overwrite the quantities that explain it. Keep every version with its issue date and a pointer to the version it replaced, so a reviewer can see what the planner saw.
Who should accept a capacity shortfall?
Whoever owns the consequence: a budget owner for a premium, a customer-facing owner for a moved date, a planning owner for open risk. One person rarely holds all three, so a single approval field is not enough. Record each part of the gap with its own accepting role and review date.
Can a model learn reconciliation from ERP tables alone?
Rarely. ERP tables keep current forecasts, purchase orders and receipts, while supplier answers, mismatch reasons and acceptance decisions sit in portals, EDI logs and meeting notes. An episode needs those joined by forecast version and time bucket before a model can see the reasoning.
Sources
- In the UN/EDIFACT D.24A code lists mirrored by php-edifact/edifact-mapping, data element 4017 holds 25 delivery plan commitment level codes, and the directory defines a delivery schedule response document, a supplier-endorsed maximum supply quantity and a previous delivery schedule number reference. Source
- The D.24A DELFOR message structure places a scheduling conditions segment ahead of each group of scheduled quantities and allows free-text segments. Source
- FAR 16.503 says a requirements contract's estimated quantity is not a representation that it will be ordered and asks the contract to state the maximum limit of the contractor's delivery obligation. Source
- FAR 52.216-21 says scheduled quantities are estimates only and lets the government buy urgent quantities elsewhere when the contractor will not accept accelerated delivery. Source
- FAR 9.104-1 requires a responsible contractor to meet the delivery schedule taking into account all its existing business commitments. Source
- UCC 2-306 limits output and requirements terms to good-faith quantities that are not unreasonably disproportionate to a stated estimate. Source
- Oracle's Supply Chain Collaboration guide lists past-due commits, forecast changes and commit mismatches as order forecast exceptions. Source
- Oracle's REST documentation types forecast exceptions as commit mismatch or forecast deviation and carries a reason code for why the supplier could not commit. Source
Tell us what data you need
Share scope, volume, format, timing and licensing needs. Datasets are sourced on request.