We wrote here an introduction to the Open Contracting Innovation Challenge, which is a competition run by the Open Contracting Partnership and the Open Data Institute to recognise ground-breaking data-driven ideas for improving public procurement. The Innovation Challenge honours original ideas for managing, analysing, and monitoring how governments buy goods and services, as well as cutting-edge approaches to publishing what gets bought, when, from whom, and for how much.

The competition looks for innovators who understand the transformational impact better public contracting can have on people’s lives. 88 teams from 40 countries responded to the challenge:  How would you use data to strengthen the integrity and effectiveness of public procurement? Six finalists have been chosen, and we promised you a synopsis of each innovation. We start today with the first two.

SpaceshipLabs

SpaceshipLabs is a civic tech and web development firm based out of the state of Quintana Roo in Mexico. The current Contratobook platform has been live since October 2016 and, using machine learning, aims to detect and flag potential contracts and contractors that may be involved in scandal and corruption.

Here’s their idea:

Corruption is a common problem for government procurement in Mexico: many suppliers have been exposed in the national press for corruption or conflict of interest. The submission team estimates that in the datasets they have used, from the national e-procurement platform CompraNet by the Secretaria de Función Pública, there was a total of $70 million USD in contracts awarded to these companies. This may not be the upper limit, with a potential of $1.5 billion given to fraudulent companies.

The team are looking to create a software that uses machine learning to automatically detect bids from companies that have been mentioned in media stories about corruption. Automatic notification of key stakeholders, such as government departments or journalists, will enable these bids to be scrutinised more closely.

The current version of the Contractobook platform has been cited in around 20 journalist investigations since October 2016, including national press outlets. They estimate that as a result of this project, that number could be significantly higher. Given the team’s estimate that contracting may cause up to $1.5 billion given to fraudulent companies, there is huge potential for the Contractobook product to address this corruption and lessen it through close scrutiny.

OpenOpps

OpenOpps is a civic tech firm operating out of London that focuses on publishing and analysing government procurement data. Its aim is to have joined-up procurement data, linked and useable. OpenOpps will unlock new analysis opportunities of procurement data by linking and matching the data of buyers and government entities for its database of over 450 different sources from 76 countries.

Here’s their idea:

Almost all of the contracting data published is unlinked – all that binds data together is the source, or sometimes the use of a common procurement classification (e.g. CPV in Europe). This makes the data difficult to compare and analyse, as users must first invest time in matching entities in the data (like the buyer names).

The team at OpenOpps wants to link as much data as they can, starting with linking buyer and supplier for open contracting data around the world. This will allow users to search for documents for a specific entity, but also by a specific sector and location (e.g. healthcare in Europe) where thresholds may be different.

This search function will allow the comparison of tenders, spotting ones that have comparatively high or low value taking into account the specific sectors. Access to this information will rectify the current imbalance in information which favours incumbent suppliers and reduces competition in contracting. This will benefit businesses who bid for government work as well as government, to help them find savings through transparent procurement.

Tune in for more finalists tomorrow.

0

No comments

You must be logged in to post a comment.