Identifying a component on an electrical drawing is step one. Knowing the exact part number to order, stock, or price - that is step two. The gap between "480V 20A 3-pole breaker" (what the drawing says) and "QO320L200PG" (what your procurement system needs) is where manual BOM workflows typically stall. AI catalog matching closes that gap automatically.
The Problem: Symbol to SKU Is a Lookup Problem
An electrical drawing annotates components with electrical specifications: voltage, current, phases, interrupting ratings, enclosure types. Your parts catalog stores those same specs under manufacturer part numbers and internal SKUs. The matching problem is connecting the two - reliably, at scale, across the variation in how specs are written on drawings from different customers, different eras, and different drafters.
Manual matching means an estimator reads the drawing spec, searches the catalog, confirms the match, and enters the part number. That step takes 2–5 minutes per line item. A 200-component BOM is a full day of lookup work, and the error rate is non-trivial - a transposed digit or a missed alternate part number adds cost or causes a procurement delay downstream.
How OpenDrawing Catalog Matching Works
When OpenDrawing extracts a component from a drawing, it produces a structured attribute record: component type, voltage rating, current rating, phase count, interrupting capacity, enclosure class, and any manufacturer or model references visible on the drawing. This attribute record is then compared against every entry in your uploaded parts catalog using a similarity scoring model.
The scoring model evaluates attribute overlap: a breaker extracted as "480V / 20A / 3P / 10kAIC" scores against catalog entries on each dimension independently and produces a composite match confidence. Entries above a high-confidence threshold (typically 95%+) are auto-assigned. Entries in a mid-confidence band are flagged for quick human confirmation - a one-click review, not a manual lookup. Entries with no plausible catalog match are flagged as "no match found" with the extracted spec, so a buyer can source the part rather than discovering the gap later in procurement.
Confidence Scores and Why They Matter
A confidence score is not a pass/fail - it is a signal about where to spend review time. In a typical 200-component BOM, 80–85% of items clear the high-confidence threshold automatically. The remaining 15–20% take 10–15 seconds each to review. Total review time: 20–30 minutes, versus a full day of manual matching.
The score also surfaces useful information about your catalog coverage. When a component type consistently produces low-confidence matches, it usually means the relevant specs are missing or inconsistently formatted in your catalog, not that the drawing extraction is wrong. That signal drives catalog improvement, which improves future match rates.
Your Catalog vs. Standard Manufacturer Catalogs
OpenDrawing matches against your uploaded catalog by default. For teams that do not have a structured internal catalog, the platform can also match against standard manufacturer catalogs for major electrical equipment brands. The accuracy difference is meaningful: matching against a generic catalog produces plausible part numbers, while matching against your specific stocking list produces the exact SKUs your procurement team needs.
Your parts catalog is provided once and can be updated as your stocking list changes. Teams that maintain accurate, well-structured internal catalogs see the highest match rates.
Integration with Procurement and ERP
Matched BOMs export with part numbers in a format ready for direct import into SAP MM purchase requisitions, Oracle Purchasing, or standard electrical estimating platforms. For teams using distribution or manufacturing ERPs, the export schema maps to common import templates. The goal is zero manual re-entry between the matched BOM and the purchase order.
For the full picture of how BOM generation works from drawing to matched output, see AI Bill of Materials Generation from Electrical Drawings.