From startup registry to observatory: data quality in southern Brazil

Resumen

Objective: to diagnose the quality of a public state registry of startups and to propose, and test on the same data, a longitudinal observatory able to monitor startup ecosystems over time.

Method: the study analysed all 1,514 records of the startup registry of Rio Grande do Sul, Brazil, extracted from its public report in September 2026. A protocol written before the analyses defined indicators for six dimensions of data quality: completeness, consistency, uniqueness, validity, coverage and timeliness. Candidate duplicates and the mapping of market segments were coded independently by two coders, and the registration dates of 693 website domains were retrieved from the Brazilian domain registry.

Results: half of the records had all descriptive fields filled, and business model was missing in 44.6%. Missingness varied sharply across municipalities, from none to 81.0% for business model, which points to differences between the organizations that register startups. The registry had no identifier, no dates and no record of the information source. Only 31.6% of the records would pass ten validation rules proposed for the observatory, and 31.5% could be found again in a future wave only by name.

Conclusions: the registry is a reasonable snapshot but cannot yet support longitudinal analysis. The proposed observatory links each diagnosed failure to a design component: persistent identifier, dated observations with source, controlled vocabularies, validation at entry and a six-monthly wave calendar. The procedure can be replicated by other regional programmes that maintain startup registries.