Systems and records
CPQ configuration exception data: can the quoted build ship as written?
By SourceX Editorial · Reviewed by Noah Loul ·
10 sections
Short answer
A configured quote must pass more than its price rule. A task built from CPQ configuration exception data should check that the option combination is allowed, required dependencies are present, the approver signed this exact version, the promised date is buildable, the item may ship to that destination and the terms fit any signed agreement. Link each test to its record, then label the quote with what was built, shipped and accepted.
Provenance trail
CPQ configuration exception
Key takeaways
- Price rules are one of at least seven tests; compatibility, dependencies, approval, feasibility, compliance, agreement terms and validity also decide whether a quote is valid.
- CPQ statuses and approvals record workflow steps at a point in time, not whether the quote could be delivered as written.
- Each constraint family lives in a different system and leaves different evidence, which the packet links by record and date.
- The label is the verified outcome from orders, shipments and returns, and open packets stay unlabeled.
- Exceptions with an approver carry the most signal, and passing counterexamples keep a model from flagging everything.
Which constraints beyond price should a valid sales quote task include?
A valid quote has to pass at least seven tests besides its price: the configuration is technically allowed, every required option and dependency is present, anyone whose approval the deal needs has approved this version, the requested delivery date is feasible, the item may legally ship to that customer and destination, the terms fit any agreement already in force, and the customer accepts within the quote's validity window. A price rule can be satisfied while any of the others fails.
A task built from CPQ configuration exception data should therefore present the starting state, the requested configuration, the rule versions in force and the linked evidence, then ask for a decision on each test, not only a number. The label is the verified outcome: what was actually ordered, built, shipped and accepted, and whether an exception, change order or return followed.
Most quote histories contain the raw material for this, spread across the CPQ, the approval log, the ERP and the shipping records. The work is linking them so that a model learns why a quote was valid, not just that it was sent.
Why can a saved and approved quote still be unbuildable?
Because a status records which workflow step finished, not whether the quote could be delivered as written. The CPQ's own checks are narrower than the business question, and they run at a point in time.
Salesforce CPQ, for example, uses product rules of several types. A validation rule whose error conditions are met blocks the quote from saving and shows an admin-written message, and the Configuration Validator API runs selection, validation and alert product rules along with configurator-scoped price rules against a configuration. Advanced Approvals evaluates approval rules when a rep submits a record, and the record is not approved until every approver in every chain has approved. All of these produce useful evidence, and none of them confirms that the factory could build the configuration by the promised date.
Approval and order states can also drift apart. A quote approved for one quantity can be edited, re-evaluated and approved again; an order can be accepted and later changed. The task outcome has to come from records created after the decision, not from the last status on the quote.
Where each constraint family is enforced, and what it leaves behind
Each family has a different system of record and a different footprint. Knowing which one a check lives in tells you what to export and which timestamp to compare against the quote date.
| Constraint family | Example system mechanism | Evidence it leaves | Typical exception |
|---|---|---|---|
| Compatibility | Table or expression constraints in a configuration model; CPQ validation rules | Rule ID and version, error or pass result | Engineering approves a combination the model forbids |
| Dependencies and selections | Selection rules that add, remove or hide options | Options added automatically versus chosen by the rep | Rep removes an auto-added option and the order fails later |
| Approval authority | Approval rules, conditions and approver chains | Approver, step, timestamp and the quote version approved | Approval obtained on an earlier quote version |
| Delivery feasibility | Order promising methods such as sales lead time, ATP or CTP | Promised date and the method used to compute it | Requested date accepted without an availability check |
| Trade compliance | Export control checks on quotations and orders by jurisdiction and classification | Check result, classification, license reference | Quote released, order held for a license |
| Existing agreement terms | Master agreement, entitlements and contract pricing | Agreement ID and the clause applied | Quote terms conflict with a signed agreement |
| Validity window | Quote expiration and price book effective dates | Expiry date and acceptance date | Customer accepts after expiry |
The configuration approval and feasibility task packet
The packet below is the original asset. One packet is one quote decision, assembled from linked records so a reviewer or a model can reach the same verdict a careful deal desk would. Each element names the record that proves it and the check that keeps the packet honest.
At a glance
The configuration approval and feasibility task packet
| Packet element | Contents | Linked record | Check |
|---|---|---|---|
| Starting state | Account, existing agreements, installed base, open orders on the quote date | CRM account snapshot, agreement register | Snapshot dated on or before the quote |
| Requested configuration | Product, options, attributes and quantities as requested | Quote lines and configuration attributes | Matches the version sent to the customer |
| Rules in force | Configuration model, product rule and approval rule versions | Rule change log | Versions effective on the quote date |
| Constraint results | Pass, fail or overridden for each compatibility and dependency rule | Validation output and override records | Every override has an approver |
| Approvals | Required approvals and those obtained, per quote version | Approval history | Approved version equals the accepted version |
| Feasibility | Requested date, promised date and the method used | Order promising result or planner note | Promised date not earlier than the computed date |
| Compliance | Classification, destination, check result and any license | Export control check log | A failed check has a recorded resolution |
| Decision | Valid as quoted, valid with exception, or invalid, with reasons | Deal desk or reviewer note | One reason per failed test |
| Verified outcome | What was ordered, shipped and accepted, and any change order or return | Order, shipment, invoice and service records | Outcome dated after the decision |
| Evidence label | Real and approved for use, or illustrative | Data owner sign-off | Illustrative packets never mixed with real ones |
Which engineering note, approval version or shipment record overturns the configurator's verdict?
The contradictions worth capturing are the ones a CRM view cannot show on its own. Each of the following produces a quote that looked valid in the CRM and was not, and each needs a record dated after the quote to detect.
- A signed master agreement fixes terms, such as a warranty or delivery commitment, that the quote overrode without an amendment.
- The approval on file covers an earlier quote version with a smaller quantity or different options.
- The promised date came from a fixed sales lead time that ignores stock, while an availability-based method would have produced a later date.
- An export control check passed on the quote but the order was held because the ship-to location changed.
- The opportunity is marked won, but the order was canceled or reconfigured before shipment.
- A configuration was overridden by engineering, built, and later returned for the defect the rule was meant to prevent.
Worked case (illustrative): a pump skid quote that passed its price rule
Every detail in this case is illustrative. An industrial distributor quotes a configured pump skid: a pump, a motor, a seal type, a base frame and a control panel, with a requested delivery in six weeks. The configuration model includes a table constraint that lists allowed combinations of seal material and pumped fluid; the rep selects a seal that the table does not allow for the customer's fluid.
The quote still saves, because the CPQ validation rule in place checks only voltage and motor frame size, so the compatibility failure is invisible at the price and validation layer. A discount above the rep's limit routes to a sales manager, who approves it. The order promising method on this product is a fixed sales lead time, so the system shows the six-week date even though a seal component is on backorder.
The task packet runs every test and finds two failures. Compatibility: the seal and fluid pair is not in the allowed table, and the only override is an engineering note dated two days after the quote. Feasibility: an availability-based check, run against the inventory snapshot of the quote date, would have produced a date about nine weeks out. Compliance: the destination and classification pass, so this test is satisfied. The decision is valid with exception, conditional on engineering approval and a revised date.
The verified outcome closes the case: the order shipped about three weeks late with a change order replacing the seal. The counterexample is the same quote with a fluid the table allows; it passes every test, ships on time, and belongs in the dataset too, because a model that only sees failures learns to flag everything.
From signed quote to shipment: the records that settle a disputed packet
The order of proof runs from signed documents to physical events. A signed quote or order acknowledgment proves what was agreed; the approval history with timestamps and quote versions proves who authorized it; the constraint results and any override record prove the configuration decision; the promising result proves the date logic; and shipment, invoice, installation or return records prove what happened.
When records disagree, the later physical record wins over the earlier status, and the disagreement itself is recorded. If an export check log is missing for a regulated product, the packet's compliance test is marked unverified rather than passed. If no outcome record exists yet, the packet stays open; an open packet is not a valid quote by default.
Quotes to keep as exceptions and quotes to leave out
Keep the exceptions; they carry the signal. Overrides with an approver, quotes revised after a failed check, and accepted quotes followed by change orders are the cases where a model learns what the rules miss.
Leave out, or flag separately, packets that cannot be judged fairly. Quotes for configurations that no longer exist in the current model can stay only with the model version that applied. Quotes with no linked outcome after a reasonable period stay unlabeled. Quotes for regulated products where compliance records are incomplete should be excluded rather than labeled as passing. Test quotes, internal demos and canceled drafts do not belong at all.
What revenue operations should remove before packets leave
Customer identities, negotiated pricing and margin are the sensitive parts, and most validity tests do not need them. Pseudonymize accounts consistently, replace absolute prices with the approval band they triggered, and remove contact names from approval notes.
Agreements with customers often treat their terms as confidential, and price data shared between competitors can raise competition issues, so counsel reviews scope before quote packets leave the company. Access to the CPQ, privacy permission over people named in the records, ownership of the records and the contractual right to license them are separate questions; removing names settles none of the last three by itself.
Quote exception histories as licensable records
Linked quote, approval, feasibility and outcome records are the kind of enterprise workflow history that is hard to reproduce synthetically, because the value is in the exceptions a real deal desk resolved.
Manufacturers and distributors weighing that option can have SourceX run the rights review, contracting, data preparation and delivery for selected quote packets. Packets are delivered only after scope and terms are approved and signed, and eligibility is first checked through a short assessment that takes no upload.
Frequently asked questions
What should a configuration approval and feasibility task packet include?
It should include the starting state, the requested configuration, the rule versions in force, constraint results and overrides, required and obtained approvals per quote version, the feasibility check and its method, the compliance check, the decision with reasons, and the verified outcome. Each element links to the record that proves it.
What is CPQ configuration exception data?
It is the record of quotes where a configuration, approval, delivery or compliance check failed, was overridden or was later contradicted, together with how the exception was resolved. Linked to the eventual order and shipment, it shows where the rules and the real outcome diverged.
Is an approved quote a valid quote?
Not necessarily. Approval shows that the people required by the approval rules signed off on a version of the quote. It does not show that the configuration was buildable, the date was feasible or the shipment was permitted, and the approved version may not be the one the customer accepted.
How do you label a quote whose outcome is not known yet?
Leave it open. Record every test you can verify and mark the outcome as pending, then label it when an order, shipment, acceptance or cancellation record arrives. Defaulting pending quotes to valid teaches a model that unverified quotes succeed.
Do quote packets need real prices to be useful?
Usually not. Most validity tests concern configuration, approvals, dates and compliance. Replacing absolute prices with the approval band a discount triggered keeps the decision logic and removes the most sensitive commercial detail.
Sources
- Dynamics 365 configuration constraints control which attribute values can be selected on sales orders and quotations; table constraints list allowed combinations and expression constraints use arithmetic and Boolean expressions. Source
- Dynamics 365 product configuration models use declarative constraints and imperative calculations, and items can be configured on sales orders, sales quotations, purchase orders and production orders. Source
- Dynamics 365 order promising methods include sales lead time, which does not consider stock availability, ATP, which includes uncommitted inventory, lead times, planned receipts and issues, and CTP. Source
- Dynamics 365 advanced export management verifies export control restrictions before sales orders are confirmed, picked, packed, shipped and invoiced, can check sales quotations, and uses jurisdictions, codes, restrictions, exceptions and licenses. Source
- A Salesforce CPQ validation product rule uses error conditions and a message, and blocks saving when the conditions are met. Source
- The Salesforce CPQ Configuration Validator API runs selection, validation and alert product rules and configurator-scoped price rules against a configuration. Source
- Salesforce Advanced Approvals evaluates approval rules when a record is submitted to decide approvers, and a record is not approved until all approvers in all chains approve. Source
See if your company qualifies
A short company assessment. No data uploads are needed.