Fashion and accessories networks share the same postal code 3.7 times more often than expected

The address of each establishment in the CNPJ (Brazilian company registry) makes it possible to verify which franchise networks occupy the same places. For the 140,185 pairs formed by the 530 networks with 30 or more active units, the indicator counts in how many postal codes (CEP) both networks have a unit, and compares that count with the one expected from the size of each network, the municipalities where it is present and the concentration of networks in each postal code. Of the 36,920 pairs with at least one co-occurrence, 8,392 have 5 or more, and 3,777 exceed what chance would explain.

Co-location is concentrated in commercial centers. In 83.6% of the 153,453 co-occurrences both networks are in the same shopping center, commercial gallery or airport, although 29.3% of the measured units are in these places, and in 15.2% both are outside them. Within the municipality, the commercial center chosen does little to differentiate networks. The median lift among the 7,447 pairs with 5 or more of these places in common is 1.21, and 5 pairs show a preference for the same place above chance.

Among the segments, fashion and accessories networks share a postal code with one another 3.66 times more often than expected, and the strongest cross-segment pairing is that of food with fashion and accessories, at 2.25. Education, gyms and services remain below 1 with all segments, which indicates more dispersed units, although part of this comes from their absence in commercial centers, since education goes from 0.44 to 1.41 in the street subset. The strongest pairs within the same segment are Intimissimi and Maria Filó (20 postal codes, lift of 24.7), Calzedonia and Dudalina (17 postal codes, lift of 25.6) and Intimissimi and New Era (32 postal codes, lift of 21.4), and between different segments, Johnny Joy and New Era (20 postal codes, lift of 16.9), Intimissimi and Yogoberry (10 postal codes, lift of 21.2) and Biscoitê and Dudalina (16 postal codes, lift of 17.3). Outside these places, among pairs in the same segment with 10 or more postal codes, the strongest are Instituto Visão Solidária and Óticas Diniz (25 postal codes, lift of 10.7) and Óticas Carol and Óticas Diniz (56 postal codes, lift of 6.6).

How we calculate

Units are the active establishments (14-digit CNPJ) that the observatory's census attributes to each network, in the open CNPJ data of Receita Federal (the federal tax authority), reference month October 2026. The 530 networks are those with 30 or more active units in the census list, a set that goes beyond the 503 measured networks of the other indicators because it includes networks with no Receita Federal count published in the network profile. A co-occurrence exists when two networks have a unit in the same postal code, counted once per pair and per postal code. Postal codes where both networks appear only under the same CNPJ root are excluded, because they reflect the decision of a single company. Also removed are general city postal codes, identified by the ending 000 or 001 with units on different streets, where 7,750 of the 59,809 units with a postal code of these networks were located.

The main level is the postal code. At the municipal level, 89,218 pairs share 5 or more cities, because large networks are present in almost all large cities, and the municipal lift, adjusted for the number of units in each city, falls between 0.38 and 1.45 in 90% of them, with a median of 0.83. At the postal code level, the same limits range from 0.56 to 10.87. The lift by municipality is included in the file for reference, with no weight in the conclusion.

Lift is the ratio between observed and expected. To calculate the expected value, each network occupies in each municipality the number of postal codes it actually occupies, and each postal code receives a probability proportional to the number of networks it hosts, so that a highly contested location does not count as affinity between the networks present there. The result is rescaled so that the average lift of the set is 1, so a value below 1 means less co-occurrence than the average pair, and not that the networks avoid each other. The split between shopping center and street repeats the calculation in each subset. A pair counts as shopping center when it is at the same postal code and number, at an address with 3 or more networks, or with the address mentioning shopping, mall, outlet or LUC (the leasable unit designation used in shopping centers), under the criterion of the indicator of units in shopping centers, in addition to the postal code with 5 or more networks in which most units are in that condition or share the same number, which includes airports and commercial galleries. The same postal code and number at an address with 3 or more networks alone accounts for 64.0% of co-occurrences. The remaining cases are street cases, and a postal code with one network in a shopping center and another outside, which accounts for 1.2% of co-occurrences, is not included in either subset.

A high lift in a small pair may be a coincidence, since 5 co-occurrences over an expected value of 0.3 result in a lift above 16. For this reason each pair undergoes a test that calculates the probability of reaching the observed count by chance alone, with a correction for the large number of pairs tested, under which up to 1% of the accepted pairs may be false positives. In practice, a pair with 5 co-occurrences needed a lift of at least 7.1, while a pair with 100 or more passed with 1.21. The lists are ordered by the lower bound of the lift in the 95% interval, which penalizes small counts. In 20 simulated draws using the model itself, with no affinity between networks, an average of 9.6 pairs passed, compared with 3,777 in the database.

Calculated on November 8, 2026. Method version 1.0. Sample: 530 networks with 30 or more units, 140,185 possible pairs, 8,392 with 5 or more co-occurrences.

Report an error on this page (in Portuguese)