The Challenge
Our client struggled with its data quality which affected the travel and expense information systems. The systems providing the raw transactional data were never designed to create data for reporting or analytics. This was why the client had issues with billing entities in place of corporate structures and hotel duplications.
The Solution
PredictX provided our client with a data quality management system that helped create a unified end-to-end service that did not separate the outputs from the data that formed them. The system used a variety of processes to improve upon input data, depending on the specific issues with data quality that had to be addressed. For some issues, such as hotel normalisation, where the lack of a known-good master dataset was frequently at the core of the problem, we built one from various sources. For other areas where miskeying by agents was an issue, we developed processes to capture the likely keying errors and autocorrect them where possible.
The other half the problem was capturing transactions that had an issue and isolating them before the numbers appeared to show made it into the outputs. So the first step was to utilise the data sent to look for inconsistencies that point to erroneous data. We also queried inconsistencies with data providers when there was a systematic error, such as duplicating transactions or the wrong data type in a field.
The Impact
The impact of going from a data system with issues to one where there is active control asserted over data quality is hard to overstate for users of travel data:
- Our client saw data quality improve from around 15% of transactions with issues to less than 1%.
- Internal processes have improved due to PredictX’s consistent AI data analysis.
- Travel teams are no longer struggling to get face time with the C-suite.
- Once coupled with other PredictX offerings, such as The Story, our client saw a rapid improvement in their presentation of information.