AI uses for records
Engineering change inventory dataset: effectivity is not a disposition
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
An engineering change inventory dataset has to record two decisions that systems often blur. Effectivity says when, or from which serial or lot, the new design applies. Disposition says what happens to each bucket of old stock: on hand, in process, finished, in transit and on order. Label every bucket use as is, rework, return, scrap or convert, with the signer, the authority, the evidence and the realized outcome.
Checklist
Engineering change inventory
Key takeaways
- Effectivity says when the new design applies; disposition says what happens to old stock. Record both, per bucket.
- A use-up effectivity is a disposition in disguise, and a planning setting can decide whether stock in transit or on order is swept in.
- Who may say use as is depends on who owns the configuration: the company, the customer or a regulator.
- IAS 2 write-downs make the economics visible, so keep the cost basis beside every rework and scrap label.
- In the illustrative bracket episode, 1,580 affected units split five ways, and the fast use-up answer fails the fit the change was made to fix.
- Grade answers on bucket coverage, constraints, authority, evidence dated at approval, economics and outcome.
What should an engineering change inventory dataset capture?
An engineering change inventory dataset should capture, for every part a change order touches, the stock that existed when the change was approved, the disposition chosen for each bucket of that stock, who chose it under what authority and what happened afterward. A change record that holds only the new revision and its effective date shows what engineering wanted. It does not show what the company did with the parts it had already bought, built or put on a truck.
The decision being modeled is the one a materials or quality lead faces the day a change is released: of the old-revision parts on the shelf, in process, finished and on order, which may still be used, which must be reworked, returned or scrapped, and who has to sign. For a team building evaluation or training episodes, that per-bucket decision is the label.
A usable label needs five kinds of evidence linked into one episode: the change order with its effectivity, a stock snapshot dated at approval, the disposition for each bucket with its signer, the cost basis at the time and the realized outcome. Requests, constraints, approvals and results that live in different systems only become a decision episode once they are joined on the change number and the part.
Why effectivity and disposition are two different decisions
Effectivity answers when the new design starts; disposition answers what happens to the old parts, and one does not imply the other. A date or serial cut-in can leave hundreds of old-revision parts with no instruction at all, and a disposition can be made with no change to effectivity.
Planning software shows how easily the two blur. In Oracle's E-Business Suite, planning can suggest a change's effective date from the use-up date of an item, the day on-hand inventory is projected to run out against gross requirements, and whether scheduled receipts on open orders count toward that date is a profile option (Oracle's MRP and Advanced Supply Chain Planning guides). A use-up effectivity is a disposition in disguise: it decides that existing stock will be used as is, and a configuration setting decides whether stock in transit or on order is included.
PTC's Windchill, by contrast, keeps disposition as a field of its own. On a change task's affected objects, users choose a disposition in separate On Order, Work-in-Process and Finished columns, from values such as Use Existing, Return, Rework, Scrap, Convert, Service, Repair and Retrofit, and can add disposition comments. A dataset should keep that per-bucket structure instead of collapsing it into a single status.
Which stock buckets does one change touch?
Every change touches at least five buckets of stock, and each lives in a different record with a different person able to act on it. A dataset that snapshots only on-hand quantity misses much of the exposure, which is often already committed on purchase orders.
The snapshot has to be dated at approval, not pulled later. Quantities move every day a change sits unresolved: work orders consume old parts, receipts land, and a supplier keeps cutting metal against an unchanged order line. A label judged against a later snapshot rewards hindsight, and it hides the stock that was used up while the decision waited.
The buckets, where their records live, the usual options and who usually decides:
| Bucket | Where the record lives | Usual options | Who usually decides |
|---|---|---|---|
| On hand at the plant | Inventory balances by lot and location | Use as is, rework, scrap, hold for service | Engineering with quality, or a review board |
| Work in process | Open work orders and routings | Finish as is, rework at a step, scrap | Manufacturing engineering with quality |
| Finished goods | Finished inventory and serial records | Ship to unaffected serials, retrofit, hold | Engineering, with the customer where the contract requires |
| In transit from a supplier | Advance ship notices and receipts | Receive and rework, return, refuse | Procurement with quality and the supplier |
| On order, not started | Open purchase order lines | Convert to the new revision, cancel, keep | Procurement, with the supplier's written acknowledgment |
| Installed base and service spares | Service stock and field records | Keep for older serials, retrofit, recall | Engineering and service, or a regulator |
Who is allowed to say use as is?
Whoever owns the configuration the part has to fit may say use as is, and that can be the company, its customer or a regulator. The record should name which, because the cheapest disposition, using old parts as they are, is also the one most often outside the decider's authority.
Defense configuration management sorts changes by their effect. An Air Force Space and Missile Systems Center instruction, citing MIL-HDBK-61A, describes Class I changes as those that affect a configuration item's form, fit, function or interface, and adds that classifying a change takes more than those four characteristics (SMC Instruction 62-109). The class is itself a judgment about authority, so a dataset should record who assigned it and when, not only the class.
A regulator can take the decision away entirely. An FAA airworthiness directive for Boeing 787 airplanes, published in February 2025, contains a parts installation prohibition: from its effective date, no one may install on any airplane a No. 1 flight compartment window with the part numbers and serial numbers listed in its appendix. Spares with those serials became uninstallable by rule, whatever anyone's disposition said.
Another directive shows how rework restores eligibility. An August 2025 directive covering various airplanes bars installing an affected buckle on any airplane unless it is marked MOD. A or INS. A. For a dataset, that marking is the evidence that rework happened; without it, a reworked part cannot be told apart from a prohibited one.
What does the cheapest or fastest disposition get wrong?
The cheapest or fastest disposition optimizes the wrong objective. Use as is avoids rework cost and keeps the line running, and scrap clears the shelf quickest, but each can break the constraint the change exists to enforce or destroy value a rework would have kept.
At a glance
What does the cheapest or fastest disposition get wrong
The accounting makes the trade-off visible. IAS 2 says the cost of inventories may not be recoverable when they become wholly or partially obsolete or when the estimated costs to complete them rise, and requires writing them down to net realizable value, usually item by item, with a reversal if circumstances later change. Rework raises the cost to complete; scrap writes the stock off. A dataset that records the cost basis and any write-down beside each disposition lets a grader check whether the chosen path was sensible, not merely permitted.
Three shortcuts recur in change records and make useful negative examples:
- Use-up by default: a planning setting consumes old stock before the cut-in, even though the old part does not fit the new design.
- Scrap by default: a whole bucket is written off when a cheap rework would have restored it, often because nobody wrote the rework instruction.
- Silence on open orders: the change is approved without converting or cancelling purchase order lines, so the supplier keeps building the old revision.
The disposition constraint matrix
The matrix states, for each disposition, when it is allowed, who signs, what the record must hold and the shortcut that makes it wrong. It doubles as the grading key for the episodes the dataset collects.
Each episode carries the six field groups listed, and each bucket in it is labeled with the matrix that follows them:
- Change header: change number, old and new revision, reason, class and approval date.
- Effectivity: type (date, serial, lot or use-up), the cut-in value and any planning setting that produced it.
- Stock snapshot at approval: quantity per bucket, lot and serial ranges, and open purchase order lines with supplier status.
- Disposition per bucket: value, quantity, signer, the authority relied on and the comment.
- Cost basis: unit cost, estimated rework cost and any write-down.
- Outcome: what was actually reworked, returned, scrapped or converted, and any later field issue.
| Disposition | Allowed only when | Who signs | Evidence the record must hold | Shortcut that fails |
|---|---|---|---|---|
| Use as is | The old part still meets form, fit, function and interface for the units it will go into | Engineering, plus the customer or regulator where they own the configuration | Serial or lot range it may serve, the change class and any directive in force | Old stock used on units past the cut-in |
| Rework | A written instruction restores the new revision and the result can be verified | Engineering and quality | Rework instruction, inspection result, marking or new part number | Reworked and untouched parts left indistinguishable |
| Return to supplier | The supply agreement allows it and the supplier accepts | Procurement, with the supplier's acknowledgment | Return authorization, quantities and who bears rework or replacement | Shipping back stock the contract obliges the buyer to keep |
| Scrap | Rework is impossible or costs more than replacement | Quality, with finance for the write-down | Reason, quantity, cost basis and the write-down entry | Scrapping a bucket a cheap rework would have saved |
| Convert on order | The supplier has not started or can switch to the new revision | Procurement, with the supplier's written change acknowledgment | Purchase order change naming the new revision and any price or date change | Approving the change while open lines still name the old revision |
Illustrative episode: one bracket change, five buckets
In this invented episode, a bracket moves from revision C to revision D because its mounting hole must shift to fit a redesigned housing that starts at product serial 1001. Revision C still fits every unit built before that serial, which is what makes some old stock usable and the rest not.
Of 1,580 affected units, the correct labels are 210 use as is (150 brackets kept as service spares for older serials and 60 finished assemblies shipped against older serials), 330 rework (210 on hand and 120 in process, redrilled to revision D), 40 scrap (an older lot without enough edge distance to redrill), 300 returned to the supplier for rework while still in transit, and 700 converted to revision D on the open purchase order before the supplier starts.
The fast answer fails. A use-up effectivity would have sent 400 revision C brackets onto new housings before the cut-in, or 1,400 if the planning setting counted every open purchase order receipt (700 counting only the 300 units in transit), and each would have failed the fit the change was made to fix. Grading rewards the per-bucket labels, the named signers and the serial restriction on the use-as-is stock, not the lowest cost. The packet for the episode reads:
| Bucket | Units | Disposition | Signer | Evidence kept |
|---|---|---|---|---|
| On hand at the plant | 400 | 150 use as is for service; 210 rework; 40 scrap | Engineering and quality; finance for the write-down | Serial restriction on spares, rework instruction, scrap reason |
| Work in process | 120 | Rework at the drilling step | Manufacturing engineering and quality | Revised routing and inspection record |
| Finished assemblies | 60 | Use as is for serials before 1001 | Engineering | Serial allocation showing each unit's destination |
| In transit from supplier | 300 | Return to supplier for rework | Procurement, with supplier acknowledgment | Return authorization and agreed rework terms |
| On order, not started | 700 | Convert to revision D | Procurement, with supplier acknowledgment | Purchase order change naming revision D |
How should a model be graded on these episodes?
Grade the per-bucket decision against what was knowable at approval, then check the signer and the outcome. A model that proposes the right dispositions with the wrong approver has not solved the task, because the authority is part of the decision. A grader should ask six things of every answer:
At a glance
How should a model be graded on these episodes
- Coverage: is every unit in every bucket assigned, including open purchase order lines?
- Constraint: does any use-as-is quantity go to units past the cut-in or to parts under a prohibition?
- Authority: is each disposition signed by someone the change class, the contract or the regulation allows?
- Timing of evidence: does each label rest on records dated at or before approval rather than on the later outcome?
- Economics: where rework and scrap are both allowed, does the choice follow the recorded cost basis?
- Outcome agreement: did the realized result match the plan, and if it did not, is the deviation itself recorded?
Requesting engineering change episodes, and where this design stops
Teams that need these episodes get further by describing them precisely than by asking for change orders in bulk. SourceX finds data for each buyer request among U.S. businesses rather than keeping an inventory of datasets, and its request form asks for scope, volume, format, timing and licensing needs; a request that names the task, the buckets and dispositions required, the time span and an acceptance test is the one worth sending. Nothing here implies that such records are available today.
The design has limits. Public documentation describes how systems store effectivity and disposition, not how often companies fill those fields in; the airworthiness directives are regulator examples, not typical commercial changes; and the episode's counts are invented. Changes to customer-owned or export-controlled designs raise rights questions that come before any dataset work.
Method: sources were checked on October 27, 2026, all through domain-limited search excerpts that a reviewer should open: PTC's Windchill help, two Oracle planning guides, an Air Force instruction citing MIL-HDBK-61A, IAS 2 and two FAA directives in the Federal Register. The episode counts come from a short script that checks each bucket sums to its total and prints the use-up exposure without scheduled receipts, with only the in-transit receipts and with every open purchase order receipt.
Frequently asked questions
What is the difference between effectivity and disposition in an engineering change?
Effectivity says when the new design applies, by date, serial, lot or the day old stock runs out. Disposition says what happens to the old-revision parts in each stock bucket: use as is, rework, return, scrap or convert. A change can set one without settling the other, so a dataset should record both.
What does use as is mean in a change order disposition?
It means old-revision parts may be used without modification, usually only for a stated range of units, such as serials built before the cut-in. It is acceptable only when the part still meets form, fit, function and interface for those units and the person signing has authority over that configuration.
Why do open purchase orders matter when a design changes?
Parts on order are often the largest exposure. If the order lines are not converted or cancelled when the change is approved, the supplier can keep building the old revision, and the buyer then faces a return, rework or write-off later.
Who approves the disposition of stock affected by an engineering change?
It depends on who owns the configuration. Engineering and quality usually decide for the company's own designs, procurement handles supplier stock with the supplier's acknowledgment, customers may need to approve changes that affect fit or interfaces, and a regulator can prohibit installation outright.
Can engineering change records be used as AI training or evaluation data?
They can when the change, the stock snapshot, the dispositions, the signers and the outcomes are linked into episodes and the company has the right to license them. Changes to customer-owned or export-controlled designs usually need to be excluded or cleared first.
Sources
- PTC's Windchill help says dispositions for a change task's affected objects are set in On Order, Work-in-Process and Finished columns, with examples including Use Existing, Return, Rework, Scrap, Convert, Service, Repair and Retrofit, plus disposition comments. Source
- Oracle's MRP guide describes letting planning suggest an engineering change order's effective date from the use-up date of a chosen use-up item. Source
- Oracle's Advanced Supply Chain Planning guide computes the use-up date by applying on-hand inventory against gross requirements, with a profile option deciding whether scheduled receipts are included. Source
- An Air Force Space and Missile Systems Center instruction, citing MIL-HDBK-61A, describes Class I changes as affecting form, fit, function or interface and says classification needs more than those characteristics. Source
- IAS 2 says inventory cost may not be recoverable when items become wholly or partially obsolete or completion costs rise, requires write-down to net realizable value usually item by item, and allows reversal when circumstances change. Source
- An FAA airworthiness directive for Boeing 787 airplanes (February 2025) prohibits installing No. 1 flight compartment windows with listed part and serial numbers from its effective date. Source
- An FAA airworthiness directive for various airplanes (August 2025) bars installing an affected buckle unless it is marked MOD. A or INS. A. Source
Tell us what data you need
Share scope, volume, format, timing and licensing needs. Datasets are sourced on request.