The welding & gas attributes buyers filter on — and most catalogs miss
The AWS classification, wire diameter, gas mix, and CGA fitting fields welding buyers filter on — and why a blank one deletes a spool from search results
A welder shopping online doesn't type "high-quality MIG wire." They type .035 ER70S-6 11 lb spool or filter by shielding gas mix, because the spec, not the copy, tells them whether it fits their machine and the job in front of them. Welding & Gas catalogs are full of SKUs that carry the right spec somewhere in a PDF or a paragraph of marketing text, but not in a structured field a filter or an AI answer engine can read. That gap is quiet — nothing errors out — but it removes real, sellable SKUs from the results buyers and machines actually query.
The attributes buyers actually filter on
Filler metals and shielding gas are two different attribute families, and most feeds flatten them into one vague "description" blob. They shouldn't be. Here's the core schema for each.
Filler metal / welding wire:
| Attribute | Why it matters | Typical values |
|---|---|---|
| Welding process | Wire/electrode is process-specific | GMAW (MIG), GTAW (TIG), SMAW (stick), FCAW |
| AWS classification | The single most-searched spec; defines chemistry and use case | ER70S-6, ER70S-3, E71T-1, E6011 |
| Base material | Determines compatibility | mild steel, stainless, aluminum, hardfacing |
| Wire/electrode diameter | Machine-dependent; wrong size won't feed | 0.023", 0.030", 0.035", 0.045" (0.6–1.2mm) |
| Spool weight & format | Drives fit on wire feeder and reorder cadence | 2 lb, 10 lb, 11 lb, 33 lb; plastic vs. wire-basket spool |
| Tensile strength / AWS grade code | Encoded in classification but also filtered independently | 70,000 psi (the "70" in ER70S-6) |
| Diffusible hydrogen designator | Required for structural/code work | H4, H8, H16 |
| Recommended shielding gas | Wire and gas are typically sold as a pair; buyers filter for both | CO2, Ar/CO2 blend |
Shielding gas & cylinders:
| Attribute | Why it matters | Typical values |
|---|---|---|
| Gas type / composition | Determines arc characteristics and weld quality | pure CO2, pure argon, 75/25 Ar/CO2, 90/10, 98/2 |
| Purity grade | Matters for TIG and stainless work especially | industrial, welding, UHP |
| Cylinder size | Capacity, usually cu. ft. or liters | size 80, 125, 300 |
| CGA outlet fitting | Determines regulator compatibility; safety-critical | CGA 580 (argon/Ar-CO2 mixes), CGA 540 (oxygen), CGA 300 (acetylene) |
| Fill pressure | Relevant for regulator selection | rated PSIG |
| Ownership model | Distributors sell both | cylinder exchange, lease, purchase |
The CGA fitting system exists to stop the wrong regulator from going on the wrong gas — which is why buyers filter on it directly rather than trusting a photo to imply the right connection.
Why a missing attribute deletes a SKU, not just hurts its ranking
Faceted search on a distributor site or a marketplace category page runs against structured fields, not the free-text description. If wire diameter lives only inside a title string like "0.035 ER70S6 mig wire", a facet filter for Diameter = 0.035" may not match it, because the filter is querying a discrete attribute value, not doing string search across every field. The SKU still exists. It just never shows up when a buyer narrows by the one spec that decides fit.
The same mechanism now applies to AI answer engines. They compose answers from structured product data across a catalog, not from parsing prose. A product page that lists AWS classification, diameter, and spool weight as explicit fields is answer-engine-quotable. One where the same facts are buried in a sentence is not.
Ask an answer engine: "what shielding gas works with ER70S-6 on rusty mild steel, and what wire diameter for a 200A MIG machine" — a well-structured catalog answers both the gas-mix question (75/25 Ar/CO2 or straight CO2, per AWS A5.18) and the diameter question in one pass, because both live as discrete fields. A catalog with the classification only in a title gets skipped for the half of the query it can't answer.
Worked example: a spool of MIG wire
Raw feed description, as it arrives from most manufacturers:
"WC-030ER70S6-11-8 — Premium MIG welding wire, 11 lb spool, copper coated, great for auto body and general fabrication shops. High quality American-made steel."
Nothing here is wrong. It's just unstructured — every fact a filter needs is folded into a sentence built for reading, not querying.
Enriched attribute table:
| Attribute | Value |
|---|---|
| Product type | Solid MIG welding wire |
| Welding process | GMAW (MIG) |
| AWS classification | ER70S-6 (AWS A5.18) |
| Base material | Mild / carbon steel |
| Wire diameter | 0.030" (0.8mm) |
| Spool weight | 11 lb |
| Spool format | 8 in. wire spool |
| Finish | Copper-coated |
| Tensile strength | 70,000 psi |
| Recommended shielding gas | CO2 or 75/25 Ar/CO2 |
| Diffusible hydrogen | Not designated |
| Typical applications | Auto body repair, general fabrication, rusty/dirty mild steel |
Same product, same facts — but now every one of them is a filterable, quotable field instead of a sentence a search index has to guess at.
How to structure it so it holds up
Two moves matter more than any single attribute:
- Split filler metal and gas/cylinder into separate attribute groups, each with its own controlled value list. Diameter should never be free text — it's a picklist of the handful of real-world sizes, stored in one normalized unit with the other displayed as a conversion, not a second source of truth.
- Encode the classification, don't just store the string.
ER70S-6should decode into process, tensile grade, and chemistry composition as separate fields, because buyers and answer engines query the components independently — "70,000 psi wire" and "ER70S-6" are the same product to a human, but only the second matches a title search.
None of this requires changing how the wire or gas is made or shipped. It's a data structure problem, and it's the kind of gap that accumulates fast across a catalog sourced from dozens of manufacturer feeds, each with its own naming habits.
That's the layer Anglera works on. It plugs into whatever's already storing the catalog — a PIM like Akeneo or Syndigo, or a flat file if there's no PIM at all — and scores, gap-fills, and structures attributes like these directly from supplier documentation, without replacing the system of record. A distributor running manual enrichment at 30-45 minutes a SKU can get a welding and gas catalog into structured, filter-ready shape in weeks, not a multi-year systems project — because the fix isn't a new platform, it's putting the spec where a filter and an answer engine can actually see it.