The dataset consists of archaeological object finds recoverd by members of the public, mostly by hobby metal-detecting, and recorded by the Finnish Heritage Agency (FHA) 2000-2023. This is a static research data dump opened to accompany the article: Oksanen, Eljas and Wessman, Anna. 2025. New Horizons in Understanding Finnish Iron Age Material Culture Through Metal-detected Finds, Internet Archaeology 68. https://doi.org/10.11141/ia.68.4 The dataset consists of three chronological parts, or finds recorded that were in: 2000-2015 - Finds reported by the public and redeemed by the FHA before the current recording scheme was initiated. The finds were originally recorded as PDF finds lists in the 'Muinaiskalupäiväkirja' data service by the FHA. These were digitized in the DeepFIN project in 2021 by Taika-Tuuli Kaivo and Eljas Oksanen. Only post-Stone Age public finds were digitized. The original data was recorded using non-controlled vocabularies, and was translated where possible to the current standard Finnish archaeological vocabulary MAO/TAO (Ontology for Museum Domain and Applied Arts) during the DeepFIN digitization process. For MAO/TAO, see: https://finto.fi/maotao/fi/. 2015-2020 - The FHA launched the data service 'Luettelointisovellus' for recording public finds in 2015. The first data dump of its database was obtained 17th November 2020. Many of these finds were recorded before the FHA's internal archaeological vocabulary (which can be directly mapped to MAO/TAO) was developed and taken into use. 2020-2023 - A second data dump was obtained from the 'Luettelointisovellus' database on 8th May 2023. These newer records are in accordance with the current archaeological ontologies. As part of the reserch project, Oksanen and Wessman lightly edited and cleaned the data, e.g., to remove typos and fill in missing date data. New fields were added to assist computational analysis (see Public_arch_finds_Finland_2000_2023_fields.csv). Note therefore that the dataset is composed from multiple sources or datadumps, and the vocabulary (particularly for the object types in the 'find_name' field for the 2015-2020 data dump) is not fully harmonized to MAO/TAO standards. This was not required for the computational needs of the above article project.