How to Size a Market in Mexico Using Free Public Data

A technical SMB in Mexico can build a defensible market-size estimate from public data alone, without a paid research subscription. INEGI publishes the three inputs a first-pass TAM/SAM/SOM needs - the classification system to define the target precisely, an establishment-level directory to count it, and a size-class breakdown of value added to price it - and the method below combines them into a number with honest error bars that you can run this week.

The three datasets, and what each one is for

Three free sources do the work, and they are not interchangeable.

SCIAN México 2023 is the classification layer. According to INEGI’s SCIAN reference sheet, it is organized into 20 sectores, 94 subsectores, 305 ramas, 610 subramas and 1,086 clases de actividad, and it is jointly maintained with Statistics Canada and the US Census Bureau rather than being a translation of NAICS. It is how you pin down exactly which market you are sizing.

DENUE (Directorio Estadístico Nacional de Unidades Económicas) is the counting layer. According to INEGI’s DENUE edition 05/2026 (Comunicado de Prensa 27/26, published 20 May 2026), it lists 6,138,075 establecimientos nationally. It is a directory, not a census: INEGI updates it twice a year, in May and November, with the next edition due 25 November 2026 per that same bulletin. It has classified establishments using SCIAN 2023 since edition 11/2024, per INEGI’s DENUE methodology document.

The Censos Económicos are the value layer. According to INEGI’s Censos Económicos 2024 definitive results (Comunicado de Prensa 79/25, 24 July 2025), the 2024 census - the twentieth edition, conducted every five years since 1930, referring to calendar year 2023 - counted 7,093,631 establecimientos and 36,592,279 personas ocupadas. The census is what lets you turn a raw establishment count into a revenue order of magnitude, because it publishes value added by size class, which DENUE does not carry.

Establishments are not companies

Before you count anything, fix one distinction that most guides skip. DENUE and the census count establecimientos - physical, single-location units - while the Censos Económicos also report unidades económicas, which in some sectors are counted per empresa. The gap is visible in the census itself: the same INEGI Comunicado 79/25 reports 7,093,631 establecimientos nationally but 5,468,180 unidades económicas in the private and parastatal sector for 2023. A single company can run many establishments, so an establishment count is an upper bound on the number of firms you could sell to, not the firm count itself.

Which size band to filter on

The official size definition is the Secretaría de Economía stratification published in the Diario Oficial de la Federación on 30 June 2009, still the governing standard. It sets Micro at up to 10 workers in every sector; Pequeña at 11 to 30 workers in Comercio and 11 to 50 in Industria and Servicios; and Mediana at 31 to 100 in Comercio, 51 to 100 in Servicios and 51 to 250 in Industria, with a combined score of workers weighted 10 percent and annual sales weighted 90 percent. DENUE’s own estrato de personal ocupado bands are close but not identical, so decide which definition you are using and keep it consistent across the estimate.

The method, step by step

Run these seven steps in order. Each one is executable with a browser and the public tools.

1. Define the target with a SCIAN 2023 code. A firm selling industrial IoT sensors to auto-parts makers, for example, sits in the manufacturing sectors 31-33 and should drill down to the clase for “fabricación de partes para vehículos automotores” using INEGI’s SCIAN catalog or the DENUE activity filter. Write the code down; every later step depends on it.

2. Count establishments in DENUE for that code, filtered by state or municipality and by size band. As a real reference point, INEGI’s DENUE 05/2026 (Comunicado 27/26) reports 2,229 establecimientos nationally under “fabricación de partes para vehículos automotores.” Do the count in the DENUE consulta map, or, for a size-filtered total at scale, through INEGI’s DENUE API, which exposes seven query methods including BuscarAreaActEstr and requires a token obtained by registering, per INEGI’s API del DENUE page. DENUE also lets you watch a local market move over time: the same bulletin notes that beauty salons in Jesús María, Aguascalientes grew from 95 to 267 establishments between 2010 and 2025.

3. Check freshness and scope. Confirm the edition date, since the directory is refreshed twice a year, not in real time. Then account for what DENUE structurally omits. Per INEGI’s DENUE methodology document, it excludes units captured only through the census rural sample, the whole agropecuario and forestal sectors, urban and suburban passenger transit, sitio and ruleteo taxis, political associations, households with domestic employees, and diplomatic missions. Its geographic rule covers localities of 2,500 or more inhabitants, all municipal seats regardless of size, and economically important localities such as industrial corridors. The true in-scope count can be modestly higher than DENUE shows, concentrated in the informal and rural tail.

4. Put a revenue order of magnitude on the count using census value added. Per INEGI’s Comunicado 79/25, national valor agregado censal bruto (VACB) was 15,667,990 millones de pesos in 2023 across 5,468,180 private and parastatal units, distributed by census size band as follows: Micro (0 to 10 personas) held 95.4 percent of units and 16.0 percent of VACB; Pymes (11 to 250) held 4.5 percent of units and 29.7 percent of VACB; and Grandes (250 or more) held 0.2 percent of units and 54.3 percent of VACB. Dividing each band’s VACB share by its unit share gives an illustrative national average of roughly 0.48 million pesos per micro unit, 18.9 million per pyme and 779 million per large unit per year. Treat these strictly as an order-of-magnitude sanity check: they are a derivation from published shares, not an INEGI-published per-unit figure, they are all-sector national averages rather than SCIAN-specific, and VACB is production value minus input costs, not sales or profit. When precision matters, pull the SCIAN-specific census tabulados instead.

5. Discount for informality, and name the metric. Two different informality rates exist, and quoting one as the other is the most common error in Mexican market sizing. According to INEGI’s ENOE indicator (Boletín 403/26, published 25 June 2026), the labor-informality rate was 55.2 percent in May 2026, up from 54.9 percent in May 2025 - a share of the occupied population. Separately, according to INEGI’s Comunicado 79/25, informal economic units were 64.3 percent of all units in 2023, up from 62.6 percent in 2018 - a share of economic units, not of people. They measure different things and cannot be used as the same multiplier. Note also that formal units accounted for 96.4 percent of VACB in 2023 and informal units for 3.6 percent, so an informality adjustment cuts your headcount far more than it cuts the addressable value.

6. Cross-check macro consistency against Data México. The Secretaría de Economía runs Data México as a free, open platform of industry and trade visualizations. Use it to sanity-test whether your sector-level totals look plausible, but do not expect an establishment count from it: it is an aggregate visualization layer, not a substitute for DENUE.

7. Present TAM, SAM and SOM as a range, not a point. Given informality, the rural-sample exclusion, the excluded sectors and the twice-yearly staleness, a single figure implies a precision the data does not support. Report a band with the assumptions stated, so a reader can see which discount is doing the work.

What these numbers are not

Three guardrails keep the estimate honest. VACB is value added, not ingresos or profit, so never present it as revenue. DENUE counts establishments, not legal entities, so it is not a company registry and not a counterparty-verification tool. And the labor-informality and unit-informality rates are separate measurements that must always be labeled by which one is in use.

The upside is that none of this requires a subscription. The classification, the directory, the census aggregates and the visualization layer are all public, which means the first-pass estimate that decides whether a market is worth a deeper study costs you time, not budget. Ascend’s free resources collect the checklists and calculators that turn this sequence into a repeatable one-day exercise.

A market study built this way will not be exact. It will be traceable to the edition date and the source of every figure, which is the standard an entry decision actually needs.