Evaluative Study: Projects of BMZ Commissioning Value below EUR 3 Million - Dataset of 490 Reports on GIZ Development Projects

The attached datasets provide the underlying data of two evaluative studies commissioned by the Corporate Evaluation Unit of the Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ). Both studies examine the effectiveness of projects implemented by GIZ on behalf of the German Federal Ministry for Economic Cooperation and Development (BMZ) with a commissioning value below EUR 3 million—projects that are not covered by regular Central Project Evaluations (CPE). The studies respond to a recommendation by the German Federal Audit Office (Bundesrechnungshof) to ensure that the effectiveness of smaller BMZ-funded projects is systematically assessed. The first study, conducted in 2023, served as a pilot to examine both the effectiveness of such projects and the feasibility of applying computational text analysis methods for evaluative purposes. The second study, completed in 2025, builds on the pilot by applying the same methodological approach to a new cohort of projects and by validating and refining the initial findings. The empirical basis consists of project metadata from GIZ’s central project database as well as project progress reports and final reports. In total, 490 reports from 159 projects were included, covering projects with end dates ranging from October 2018 to December 2023. Using these sources, the studies assess project effectiveness in line with the OECD DAC evaluation criteria, with a particular focus on the criterion of effectiveness, defined as the extent to which projects achieve their stated objectives. Methodologically, the studies combine automated extraction of project success indicators from report tables with computational text mining techniques applied to narrative report content. This includes the use of a fine-tuned BERT language model to identify and classify evaluative statements in project reports as positive or negative, as well as complementary methods such as dictionary-based analysis, topic modelling, and word frequency analysis. Together, these approaches allow for a systematic, comparable assessment of project performance and the identification of supporting and hindering factors across a large number of projects.