Abstract. We look at the effects of skilled migrants and migrant diversity on firm productivity, using a novel worker-firm dataset – leveraging millions of online profiles – to identify impacts and mechanisms. In theory, migration can boost firms’ productivity through improving human capital, improved task specialisation, or via diverse teams’ role in ideas generation/scrutiny. Conversely, discrimination or lack of trust may lead to sub-optimal task allocation or workplace frictions, respectively. Urban location may amplify these effects, via compositional channels and/or interaction with agglomeration economies. A growing body of evidence finds support for migration-productivity effects, although effect sizes vary considerably across countries, industries and workflows. Two major challenges here are the lack of employer-employee datasets in many countries, and limitations in coverage / dimensionality in those that do exist. Data from online platforms and the web may help meet these challenges. We sample individual education/career histories from the Diffbot knowledge graph to the UK company register, alongside financial data from Orbis Historical, 2007-2023. Focusing on larger firms across a range of sectors, we define migrants through name, language and countries of education, using panel fixed effects and GMM/IV settings for identification. We then explore the contributions of seniority, migrant human capital, task specialisation and migrant diversity. Results suggest a U-shaped link from migrant workforce share to firm TFP, which is more pronounced for senior staff, and for a minority of skill- and STEM-intensive firms. Migrant human capital and migrant worker specialisation into technical roles help explain our results.