Industry Classification: NIC, NACE and the Search’s First Filter

Definition

Industry classification is the standard coding of an activity into a hierarchy of classes — the scheme the [comparable search] (/docs/benchmarking/search-design) uses to say where the tested party’s comparables live in the database. The two schemes in play for an Indian study: NIC-2008 (the National Industrial Classification — the Indian statistical office’s hierarchy, 5-digit down to the class, ~1,297 classes) and NACE (the EU’s Nomenclature of Economic Activities — the code most global benchmarking databases key on, with a published NIC↔NACE crosswalk). The classification’s role in the [search string] (/docs/glossary/search-string) is the first structural filter: the classification codes select the activity population (the companies that do what the tested party does), the keywords select the function within it, and the [size filter] (/docs/glossary/size-filter) selects the scale — three orthogonal restrictions, and the classification is the one that is standardized (the code is a shared vocabulary between the study, the database, and the examiner, in a way a keyword list is not).

Industry classification, in one use:
  1. The tested party's activity (the FAR — what it does, in its own language)
  2. The code family (the NIC/NACE classes that cover the activity — the 2–3 sibling codes)
  3. The database's key (the code the database indexes on — NIC for Indian sources, NACE for global)
  4. The candidate population (the companies coded into the family)

The element	The content
The scheme	NIC-2008 (Indian) and NACE (global) — crosswalked where the database keys on the other
The level	The code family for the tested party’s activity (the sibling classes — broad enough to catch the segment, narrow enough to stay relevant)
The function test	The code selects the activity; the FAR keywords select the function within the activity (a NIC 312 plant may toll-manufacture or own the product — the keyword splits it)
The record	The codes used, in the search design, with the family rationale (why these classes cover the tested party)

The working read (the search design guide): the classification is a map of the activity space, and its failure modes are the two map errors — the code too coarse (the 2-digit division: “manufacturing” pulls in the chemical plants into a consumer-goods search) and the code too fine (the 5-digit class: the exact product class may contain a handful of Indian listed companies, and the [comparable set] (/docs/glossary/comparable-set) cannot reach the [statistical significance] (/docs/glossary/statistical-significance) floor). The practice answer is the family — the tested party’s class plus its siblings (the 3-digit group level, typically), with the function narrowed by the keywords rather than by the code. The NIC code term has the Indian scheme’s structure and the NIC finder the lookup; the classification decision belongs to the [search design] (/docs/benchmarking/search-design), documented in the study before the results are seen — the same frozen-first discipline as the string.

Example

An Indian contract manufacturer of electronic assemblies. The classification decision:

The question	The answer
The activity	Manufacture of electronic assemblies (the tested party’s function — the manufacturing, not the design)
The NIC family	NIC 26/319 family — the electronics and “manufacturing n.e.c.” classes (the 3-digit siblings that cover the assembly activity)
The NACE key (global database)	NACE 26/28 family (the crosswalk — the database indexes on NACE)
The function split (keywords)	“contract manufacturing”, “electronic assembly”, “toll” — the keyword that separates the contract manufacturer from the own-brand producer in the same code family
The level check	The family returns a usable population; the single 5-digit class does not (the thin-market check)

The point is the split: the NIC family alone would return both the contract manufacturers and the own-brand electronics producers (same activity code, different FAR — the own-brand producer bears the market and inventory risk the tested party does not). The classification says activity; the keyword says function; the [accept-reject matrix] (/docs/glossary/accept-reject-matrix) then records the qualitative decisions the two together could not make.

See also

• NIC Code
• Benchmark Search Design
• Search String
• Functional Analysis (FAR)

FAQ

NIC or NACE — which does an Indian study use? Both, at different layers: the NIC-2008 code is the Indian vocabulary (the statutory classification — the Local File describes the tested party’s activity in NIC terms, and the Indian database sources key on it); the NACE code is the database key for the global benchmarking sources (the Prowess/global feeds index on NACE, with the published crosswalk). The search design records both (the NIC family and its NACE mapping) — the examiner reads the NIC, the database answers on the NACE, and the crosswalk is the bridge stated in the file.

Is the industry code the same as the tested party’s “industry” in the report? No — the code is the search’s classification (the activity population’s key); the report’s “industry” is the narrative label (the “consumer packaged foods distribution”, the human phrase). They should agree (the code family covers the narrative industry — a mismatch is a design defect), but they are different artifacts: the code is machine-readable and standardized, the narrative is human-readable and specific. The [industry diagnostics] (/docs/product/results/industry-diagnostics) in Quartyl key on the same classification, which is why the study’s code choice and the diagnostics’ industry view line up.

Can the classification be wrong and the set still be fine? It can be incomplete and the set still fine — the keyword and the qualitative screening catch much of what a coarse code lets in. It cannot be exclusively wrong: a code family that misses the tested party’s activity (the wrong division entirely) is not repaired by keywords — the candidates were never in the population, and the comparable set is built from what the search returned. The level check (the family’s population size, the thin-market test) is what catches the exclusive error at design time — before the results, per the frozen-string discipline.