How to Search Businesses by Industry (Properly)

"Find me all the manufacturing companies in the region" sounds like one question. It is actually three, and they have three different answers. Ask a government registry and you get everything with a manufacturing code, including a one-person workshop and a chemical plant. Ask a maps platform and you get whoever picked a manufacturing category when they claimed their listing. Read the companies' own websites and half of them call themselves something else entirely.

This is the single biggest reason industry searches return junk. People treat "industry" as a fixed property of a company, when it is really a label applied by different systems for different purposes. Once you understand which system you are querying, you can pick the right one for the job and stop fighting the results.

Industry is three different things

Every list you build comes from one of three sources of truth, and they rarely agree with each other.

1. Official statistical classifications

These are the code systems maintained by governments and statistical agencies. In North America the current standard is NAICS, the North American Industry Classification System, developed jointly by the statistical agencies of the United States, Canada and Mexico, and periodically revised. It replaced SIC, the older Standard Industrial Classification, which you still meet in legacy databases, older filings, and some commercial data products that never migrated.

In the European Union the equivalent is NACE, maintained at EU level, with member states running national versions aligned to it. Most countries outside these blocs run their own classifier, usually harmonised to some degree with the UN's international standard: Russia uses ОКВЭД, Brazil uses CNAE, and there are equivalents across Asia, the Gulf and Latin America.

What they share is structure: a hierarchy running from broad sectors down to narrow activities, each level adding digits and specificity. Broad levels group things like manufacturing, construction or professional services; deeper levels split those into individual trades. The number of levels and the code format differ by system, and every system gets revised, so codes shift meaning between revisions — never assume a code you memorised years ago still means what you think. Look it up in the current official classifier for the country you are targeting.

The important property of these codes: a company is assigned one at registration, usually by its own declaration, and often nobody ever updates it. It reflects what the founders intended to do back then, which may be years out of date.

2. Platform categories

Google Business Profile has its own category list. LinkedIn has its own industry list. Directory sites, marketplaces and app stores each have theirs. None of them map cleanly to statistical codes, because they are built for a different job: helping a consumer or recruiter find something, not helping a statistician count things.

Their advantage: they are self-selected, recently, by someone actively marketing the business. A company that picks its Google category is telling you what it wants to be found for today — usually more current than a registry code from its founding. Their weakness: the lists are short and coarse. A platform covers every kind of business on earth with a limited category set, so anything specialised gets flattened into the nearest available bucket.

3. How the business describes itself

The third source is plain language: the words on the company's homepage, its meta description, its LinkedIn headline, its service pages. "We do Shopify migrations for supplement brands" has no code in any classifier on the planet, and no platform category comes close. But it is the most accurate description of what that company actually sells, and if that is your target market, it is the only signal that will find them.

Self-description is the most current and most precise of the three, and the least structured. You cannot filter on it — you have to search for it.

When to use official codes

Codes are the right tool in four situations.

• Registry pulls. When extracting companies directly from a business register, codes are usually the only filter available — there is no free-text search over a national register.
• Tenders and procurement. Public procurement systems are built on classification codes. If you sell into tenders, you work in codes whether you like it or not.
• Coverage at scale. If you need every company in a sector across a whole country, codes give completeness no maps or search approach can match. Registries contain businesses with no online footprint at all.
• Statistical framing. Market sizing, territory planning and anything you must reconcile with published economic data has to speak the same code language as that data.

The limitation is granularity, and it is severe. A single code routinely covers a solo practitioner and a national chain — one dental code holds both the dentist working alone above a shop and a two-hundred-clinic group. If your product only makes sense above a certain size, the code alone tells you almost nothing. It is a starting universe, not a target list. The second limitation is drift: a company registered as a print shop that pivoted to digital marketing five years ago still carries the print code, because changing it costs paperwork and gains nothing.

When to use platform categories

Platform categories are strongest for local businesses with a physical presence: restaurants, clinics, salons, garages, gyms, retail, trades. These businesses actively maintain their listings because customers find them that way, so the category is fresh, and it comes bundled with the things you actually need — address, phone, opening hours, review count, photos, often a website. Review count and photo activity double as a crude liveness check: recent reviews mean a business that is currently operating, which is more than a registry entry guarantees.

Where platform categories mislead

Businesses without a storefront are badly served. B2B service companies, software firms, wholesalers, consultancies and agencies often have thin or missing listings because walk-in traffic is irrelevant to them. Searching maps categories for "software company" gives you a fraction of the market, skewed towards whoever bothered to optimise a listing. Categories also get gamed: businesses pick adjacent categories to appear in more searches, so any category list contains hopefuls as well as genuine members, and multi-service businesses pick one primary category that hides everything else they do.

When to search by language instead

Use self-description search when your target market is defined by something no classifier has caught up with: a technology, a niche within a niche, a business model, or a customer type. New categories always appear in language years before they appear in a code.

The technique is to search for the phrases the companies themselves use, not the phrases you would use to describe them. This is the step most people skip. Before searching, open five companies you know are perfect fits, read their homepages, and write down the exact terms they use. Companies in a niche converge on shared phrasing, and that phrasing is your query.

Practical operator technique, applicable to most search engines:

• Exact phrases in quotes for the terms of art. Loose keywords match articles about the industry; quoted phrases match companies in it.
• Restrict to a site section — searching within paths like services, solutions or about pages surfaces companies rather than blog posts and directories.
• Restrict to a country domain to keep results geographically relevant, accepting that this misses companies on global domains.
• Search page titles for the industry term. A company that puts the term in its title tag has made it central to its positioning; one that merely mentions it in body text may just be a customer of that industry.
• Exclude the noise — subtract words like jobs, careers, courses, wikipedia, and the names of the directory sites that dominate every industry query.
• Search adjacent artefacts: association member pages, conference exhibitor lists, award shortlists, supplier directories. One good member list beats a hundred search results, because someone else already did the classification and had a reason to get it right.

The weakness of language search is recall. You will never find every company this way, and you will systematically miss the ones with bad websites — in some trades, the majority. Use it for precision, not coverage.

Combining signals into a usable list

No single system produces a good list. Good lists come from stacking two or three signals so the weaknesses cancel out. A workable stack has four layers:

1. Category or code — the broad universe. Keep it deliberately wide: filtering down is cheap, discovering a segment you excluded at step one is not.
2. Geography — city, region or service area. Almost always the strongest single filter and the cheapest to apply reliably.
3. Size proxy — because codes ignore size, and size usually decides whether a company can buy from you. Useful proxies: number of locations, review volume, a careers page with open roles, team page headcount, a physical office versus a residential address.
4. Technology or behaviour signal — what the site runs on, whether they run ads, whether they publish, what booking or payment tool they use. Often the sharpest qualifier of all, because it describes operational maturity rather than sector.

Layer four is where most lists get their edge. "Dental clinics in a metro area" is a category. "Dental clinics in a metro area with three or more locations, an online booking system, and a careers page" is a target list — same starting universe, radically different reply rate. Practically, this means running a broad category-plus-geography search first, then enriching each result with website, contact channels and technology signals, then filtering on the enriched fields. Tools that search maps data, registries and the open web in one pass and return enriched records make this a single operation instead of three; if you want to see what a stacked search returns for your own niche, try it on JustLeadIt and judge the output against companies you already know.

Sanity-checking for classification drift

Before you spend a week contacting a list, spend twenty minutes auditing it. Classification drift is the failure mode where entries are technically in-scope but nothing like your customer, and it is invisible in aggregate — the list looks fine until you start calling.

A fast audit that catches most of it:

• Open twenty at random — not the first twenty, which are usually the best-optimised. Score each as clear fit, wrong size, wrong sub-segment, or not a real company. More than a quarter in the last three means your filters are wrong, not your messaging.
• Check both ends of the size range. If the top of the list is enterprises you cannot serve and the bottom is one-person operations, you need a size filter, not a better message.
• Hunt for the wrong side of the market. Industry searches reliably pull in suppliers to the industry, media covering it, associations, training providers and recruiters serving it. All match the keywords; none are your buyer.
• Check for aggregators and duplicates. Directory pages, franchise head offices listed alongside every branch, and the same company under three trading names inflate a list without adding reachable prospects. Deduplicate by phone number and domain, not by company name.
• Verify a sample is still operating. Registry-sourced lists carry dissolved and dormant companies. Recent website activity, recent reviews, or a working phone number are enough of a check.
• Ask what the list is missing. Name three companies you know are perfect fits and search for them in your own results. If they are absent, your filter is too narrow.

A workable sequence

1. Define the target in plain words first — one sentence including size and situation. If you cannot write it, no code will save you.
2. Pick the primary system that fits: platform categories for physical local businesses, official codes for whole-sector coverage and tenders, self-description search for emerging or hyper-specific niches.
3. Run wide, then narrow: broad category plus geography, then enrich, then filter on size and behaviour.
4. Audit twenty rows before contacting anyone, and fix filters rather than rewriting messages.

Industry classification is a map, not the territory. Official codes tell you what a company registered as, platform categories tell you what it wants to be found as, and its own website tells you what it actually does. Pick the one that matches your purpose, cross-check with a second, and always audit the output against real companies you know. The teams that build good lists are not the ones with the best data source — they are the ones who know which question they are asking.