About a third of development projects fail to achieve satisfactory outcomes, according to agencies' independent evaluation units. To a large extent, these outcomes appear to be baked into projects at their inception due to inadequate project design or relevance. This prompts questions about the diffusion of project design concepts: To what extent are better-designed or better-performing projects more likely to be emulated? Do factors of bureaucratic or political attractiveness -- such as ease of set-up and rapidity of disbursement -- play a role? To address these questions, this paper explores the use of methods from network science. It constructs a network graph of the relationship among the components of all World Bank investment projects initiated from 1996 to 2014, based on the semantic similarity of the component descriptions. It uses the network to assess the characteristics of projects that are more 'prolific' in the sense of having closely related followers, and as tool for visualizing diffusion of design concepts. This illustrative exercise defines a measures of project 'influence' on subsequent projects and tests simple, nonexclusive hypotheses about the determinants of influence. It finds no significant impact of project outcome or quality of entry (as independently rated) on 'influence.' Nor does ease of project preparation (as proxied by time from concept note to effectiveness) have any significant effect. However, very small projects (less than 0 million) have markedly lower 'influence' on average. This finding may have implications for the usefulness of small projects as pilots for subsequent scale up.