S&W, Smith & Wesson, SMITH&WESSON: why messy data costs you sales

Filter a distributor's site for "Smith & Wesson" and you'll get most of their Smith & Wesson guns. Not all of them. Some are filed under "S&W". A few are under "SMITH AND WESSON". One or two have "SW" in the brand field and the real brand stuck in the model name.

You won't know what you missed, because a filter that comes back with results looks like it worked.

Multiply that by sixteen distributors, each with its own spelling habits, and "show me every 9mm Smith & Wesson in stock" gets a lot harder than it sounds. A big part of what we do is cleaning that up before it reaches your screen.

How bad it is

Here are spellings we've had to map, each one meaning the same thing.

9mm Luger: "9mm", "9MM", "9MM LUGER", "9 mm Luger", "9 MM", "9mm Para", "9mm Parabellum", "9x19", "9X19", "9x19mm", "9 mm (.355-.356)". That's eleven ways to write the caliber you probably sell more of than any other.

.40 S&W: "40 S&W", ".40 S&W", ".40S&W", "40SW", "40 cal", "40 Smith & Wesson (S&W)", ".40 SMITH & WESSON".

.38 Special: "38SP", "38SPL", "38 SPC", "38 Smith & Wesson Special +P".

7.62x39: "7.62X39", "762X39", "7.62 x 39 mm", "7.62x39 (.310)".

Smith & Wesson: "Smith & Wesson", "SMITH AND WESSON", "S&W", "SW", "Smith_wesson". Some rows had the brand typed into the model field, so the model read "S&W" or "Smith" instead of "M&P 15 Sport II". A few used the distributor's own SKU as the model name.

Categories are worse. One distributor's category field reads "rifles,tactical_rifles". In another feed, a bare "cans" means cans of gunpowder, not suppressors.

How we clean it up

For every kind of value we care about, like manufacturer, caliber, category and model, we keep one list of correct names. Each correct name carries every spelling we've seen for it. When a feed comes in, each value gets looked up against that list and swapped for the correct name.

A few things make that work on real data:

• Case and spacing don't matter. "SMITH & WESSON" and "smith&wesson" match without anyone listing both.
• Rules for the hard ones. Some spellings can't be sorted out by a lookup alone. "Rock Island" is the clearest case, below.
• Distributors outvote each other. If more distributors say a gun is one caliber than agree with what we have stored, the stored value gets corrected.
• People review it. We look over the list by hand, merge duplicates, and lock products we've corrected so a later sync can't undo the fix.

The lists aren't finished and never will be. Distributors invent new spellings every month.

Three fixes from this summer

Suppressors got their makers back. In August, only 317 of the 1,071 suppressors in the catalog were linked to their manufacturer. That's 30%. The rest had brand names our list didn't recognize, or matched more than one maker. After we merged duplicate brand entries (things like "Dead Air" and "Dead Air Armament", or "SureFire" and "Surefire") and cleaned out junk aliases, coverage went to 88%. One maker went from 3 suppressors in the Finder to 24.

Categories got a shared vocabulary. Until September, categories were matched against a short list that only covered firearm types. About two-thirds of what distributors sell isn't a firearm, so it had nowhere to go. We built a shared list of 18 product types with 1,092 known spellings behind them. Now 98.2% of active listings land in a real category.

Rock Island Armory got its own name. Distributors put the same "Rock Island - Armscor" label on Armscor ammo and on Rock Island 1911s, and an old cleanup had filed all of them under Armscor. So the Firearms Finder had no Rock Island Armory at all. Now the product type decides. Guns go to Rock Island Armory and ammo stays Armscor. That moved 238 firearms to the right brand.

Why this matters at the counter

Every filter in the Firearms Finder, every brand facet in search, and every catalog count on our distributor pages depends on this. If brands and calibers were left as the distributors spell them, a search would quietly skip guns you could have sold.

If you spot one we got wrong, hit Suggest Edits on that product in the Firearms Finder. We read every one.