AI · Product matching
AI product matching for lighting: how an equal is found and ranked
Two fixtures can share a description and differ where it matters, or look nothing alike on paper and be true equals. Matching is about attributes, not words.
Short answer
AI product matching treats each spec line as a set of required attributes and searches a verified catalogue in three stages: candidate generation (text and embedding search on descriptions and catalogue numbers), hard filtering (pass/fail attributes such as input voltage, dimming protocol, mounting, location rating and certification), and scoring (delivered lumens and distribution from IES files, CCT, CRI, efficacy, dimensions, lead time). The result is a ranked list of equals, each with the evidence of where it matches and where it differs, for a person to approve.
Key takeaways
- Product matching is entity resolution with engineering constraints.
- Text similarity finds candidates; attributes decide.
- Hard constraints are pass/fail; performance attributes are scored.
- Compare delivered lumens and distribution from IES files, not catalogue headlines.
- Every match carries evidence and a human decision.
Matching is entity resolution with constraints
Product matching is a branch of entity resolution: deciding whether two records describe the same item, or, in procurement, whether a catalogue item satisfies a specification. Research on product offers has long shown why this is hard: many similar but different products, missing or wrong values, and descriptions written differently by every source.
Lighting adds engineering constraints. A fixture that is 99% similar in text but runs on the wrong voltage is not 99% right; it is wrong.
Three stages
1. Candidate generation
Search the verified catalogue for plausible products using text and embedding similarity on descriptions, families and catalogue-number patterns. The goal here is recall: do not miss a good equal.
2. Hard filters (pass/fail)
| Attribute | Why it is pass/fail |
|---|---|
| Input voltage (incl. 347V) | Wrong voltage cannot be installed |
| Dimming protocol | Must work with the control system |
| Mounting type | Must fit the ceiling or wall detail |
| Location rating | Dry, damp, wet |
| Certification for Canada | Required for installation |
| Emergency option | Required where specified |
3. Scoring (how equal)
| Attribute | Compared using |
|---|---|
| Delivered lumens | IES file, same length |
| Distribution | Candela data (direct/indirect split, beam) |
| Colour | CCT, CRI, R9, consistency |
| Glare | Optic type, UGR tables |
| Physical | Dimensions, aperture, section lengths |
| Commercial | Lead time, origin, price |
Weights depend on the spec: in a visible continuous run, colour consistency and optic weigh more; in back-of-house, lead time may decide.
Evidence, then a human decision
Each ranked candidate shows where it matches, where it is better and where it differs. That is what a design team needs to approve an equal under the project’s substitution procedure. The approval, rejection or edit then becomes training signal for better ranking.
Custom linear lighting is matched as a system
For custom linear runs, the “product” is a system: profile, optic, output per foot, colour, driver strategy and the run layout with sections, corners and feeds. Matching works per component and then rebuilds the run. See what drives the cost of custom linear lighting.
Data quality decides
Matching is only as good as the product data behind it. Structured, normalized attributes and photometric files are the foundation; see the data model behind AI matching.
Send a schedule and SuppLux’s AI matches every line to verified products, with the best alternative shown first, then turns the matched set into a quote and submittal package. Send a schedule.
Frequently asked questions
What is entity resolution in product matching?
Entity resolution decides whether two records describe the same thing. In product matching it decides whether a catalogue item satisfies, or is equivalent to, a specified item, even when names and formats differ.
Can AI decide that two lighting fixtures are equal?
AI can rank candidates and show the evidence, but approving an equal is a design-team decision under the project’s substitution procedure. The system’s job is to make that decision fast and well informed.
Why are photometric files important for matching?
Catalogue lumens and beam descriptions are not comparable across manufacturers. IES files (ANSI/IES LM-63 format) give delivered output and distribution, so equals can be compared on what reaches the room.
How does AI handle custom linear lighting?
As a system: profile, optic, output per foot, colour, driver and run layout are matched separately, and the run is rebuilt from sections, corners and feeds.
References
Keep reading
- AI · Document AI How AI reads a lighting specification: from PDF to a structured fixture schedule How AI turns a lighting specification and fixture schedule (PDF, scanned or Excel) into structured, checkable data: layout detection, table extraction, entity extraction with language models, unit normalization, cross-checks against the plans, and confidence-based human review. →
- Guide · Alternates How to get a lighting "or equal" substitution approved How to prepare a lighting substitution or value-engineered alternate that design teams approve, covering what to compare (photometrics, colour, electrical, dimensions, compliance), how to present it, and when not to substitute. →
- AI · Product data Structured product data for lighting: the data model behind AI matching What structured product data looks like for commercial lighting and why AI matching depends on it: attribute schemas (ETIM classification), photometric files (IES LM-63), certification and energy listings, catalogue ordering logic, and commercial data such as lead time and country of origin. →
- Platform · Spec matching AI spec matching for commercial lighting SuppLux matches every line of a lighting specification to verified products and ranks equal or better alternates by photometrics, electrical data, mounting and lead time, with a specialist review before quoting. →