Tackling corruption with transparency and technology

People can’t make sense of data they can’t see or readily understand, and it can be hard to trust automated insights that can’t be verified. This is where tools for visual data analysis, like Power BI (a data visualization solution), can make a difference. By making both input data and derived insights visible and navigable by a broader array of stakeholders – not just data scientists – such data tools can bring unprecedented transparency to the activities behind the data.

Our previous extensions of Power BI to support analysis of unstructured text, media, and meta-data, provided a readymade collection of visuals with which to represent and explore the networks of suppliers awarded public contracts. Similarly, our open source library for graph statistics, developed in collaboration with Johns Hopkins University, offered the algorithms we needed to model these networks, predict unobserved links between suppliers, and identify anomalous changes in activity over time. Using a copy of an actual supplier network dataset, we were therefore able to create an initial proof-of-concept more quickly and easily than in the past, requiring only a few days of development time and several hundred lines of original code.

This kind of speed is important, as it offers the potential for very rapid feedback and many cycles of design iteration. Arriving at the first meeting with a partner after they have shared a dataset with you – not just with a list of questions, but with a proof-of-concept solution – can fundamentally change the nature of the conversation. Although the first version will not be complete, or even correct in many cases, it demonstrates a bias for action, an ability to adapt quickly, and the art of the possible.

Another benefit of working with the same set of capabilities across multiple partners and datasets – like understanding public procurement from the perspective of both the World Bank and IDB – is that it helps to reveal the core structure of the problem, which often doesn’t change across contexts and use cases. This can give us the evidence we need to invest beyond the proof-of-concept stage and develop solutions that are usable and effective in real-world scenarios.

For our current work, potential user groups include governments, journalists, investigators, suppliers themselves, academics, and the general public. Ultimately, these audiences might benefit from different views of the same data, based on distinct roles, responsibilities, and interests. This is where the advantage of using an industry-standard application platform like Power BI comes in – not only does it enable rapid prototyping in the initial stages of development, it also is available for extension and customization by end users – all in a code-free way that may be broadly accessed.