Compare Supplier Quotations Without Hiding the Assumptions

02Similar descriptions do not establish equivalent scope.

03Define the comparison quantity, technical requirement and delivery basis before entering prices into a ranking.

04KoSIT publishes reusable country and currency code lists associated with standards including XRechnung and XBestellung.

05[1] That is a useful example of consistent field meaning.

06Your internal comparison still needs its own agreed treatment of quantities, units and commercial assumptions.

07Keep the original supplier wording accessible.

08Normalisation should make offers comparable without erasing qualifications.

09A field saying delivery included needs the location and scope behind it.

10An engineering reviewer should be able to return from the comparison row to the precise statement in the source document.

02Then apply reviewed calculation rules.

03Finally, present missing information as questions.

04Combining these steps in one generated summary makes it difficult to identify whether an error came from reading, arithmetic or interpretation.

05Supplier documents may include personal contact details and confidential commercial information.

06DSK's 2024 guidance discusses purpose definition and AI application selection.

07[2] Use the approved processing route for the intended data rather than uploading every quotation to whichever tool is most convenient.

08Give technical and commercial reviewers different responsibilities.

09Engineering should determine whether a proposed alternative meets the requirement.

10Purchasing should assess commercial conditions and request clarifications.

11AI may organise the evidence for both, but should not turn an unreviewed technical substitution into an apparently comparable price.

02Each row should contain the quoted value, its source and any normalisation rule.

03Missing information should remain visible until the relevant owner resolves it.

04Avoid a single overall score until critical differences are resolved.

05A weighted score can make a missing acceptance requirement look like a small disadvantage.

06Some gaps should block comparison, while others can be priced or consciously accepted by the authorised buyer.

02Supplier A quotes EUR 18 per component plus EUR 4,000 for tooling.

03Supplier B quotes EUR 20 per component with tooling included.

04Both totals are EUR 40,000: 2,000 multiplied by the unit price, plus any separate tooling charge.

05That equality does not settle the purchase.

06Supplier A may transfer tooling ownership while Supplier B retains it, or the offers may have different inspection scope.

07These are hypothetical possibilities that require evidence, not assumptions to enter silently into a model.

08The buyer requests clarification and records the answers beside the original quotations.

09If engineering confirms equivalent scope and commercial terms become comparable, the team can evaluate delivery and supplier capacity.

10The AI's useful contribution is showing why the apparent EUR 2 unit-price difference did not by itself establish a lower total cost.

02Review whether the system preserves those distinctions.

03A technically neat spreadsheet can still contain a commercially false comparison.

04Check that revisions replace the correct offer state without deleting the history.

05A supplier may change price while leaving the technical scope unchanged, or revise both.

06The comparison should show what changed and which previous approvals need another review.

07There is a tradeoff between speed and the depth of normalisation.

08Routine repeat purchases may need only a few fields.

09Bespoke tooling or equipment may require a more substantial review.

10Use separate comparison modes so an uncomplicated reorder does not inherit unnecessary bureaucracy and an engineered purchase does not bypass important questions.

02The purchase order should carry the agreed scope and references, rather than relying on a summary that never reaches the supplier.

03Review a completed comparison when the first invoice or delivery arrives.

04If unexpected tooling charges or missing inspections appear, trace them to the comparison fields.

05This closes the learning loop between procurement, engineering and goods receiving.

06Start with a recurring purchase category whose offers are difficult but manageable to compare.

07Build a small set of reviewed examples and explicit calculations.

08Expand when buyers can explain the selected offer from the stored evidence, without reconstructing the reasoning from private notes or trusting an unsupported AI recommendation.