A graduate engineer from the Lille Ecole Centrale, Jean-François Verrier’s initial professional experience was in the area of logical security and payments systems. After some time working in exchange systems (messaging systems and directories), he went on to work on decision-support systems and Business Intelligence as a Project Manager, and has remained there for three years now. Most notably, he managed the INDIA project (the French National data center for expenditure and associated information) for the French Treasury Directorate, as part of the roll out of the country’s new financial law LOLF (Loi Organique relative aux Lois de Finances).
The Business Intelligence (BI) systems being deployed throughout the world are giving rise to a proliferation of indicators and dashboards, and these in turn are themselves becoming increasingly abstract and unusable. The ‘chronological layers’ in the structure are stacking up as changes in the law and regulations are taken into account, and systems are re-worked to reflect restructuring operations and the organizational changes that escalate with every merger, acquisition, or policy realignment. In short, strategic information is losing its relevance because BI systems are not being ‘leveled’, or resetting to zero. But because these types of systems are complex by nature, this loss of relevance can originate from multiple, and sometimes incompatible, causes. We will take a look at the three main ones in the paragraphs that follow.
Too much information kills information
Let’s imagine that some military personnel are installing a radar system with such a fine resolution that it will enable them to pinpoint the flight paths of migrating birds. If this data is not filtered out before air-traffic controllers responsible for monitoring aircraft movements in their national airspace receive the incoming data, they will be unable to execute their task effectively, and unfriendly aircraft could easily get confused with the echo cloud generated by birds. What is crucially important is not so much to acquire information, as to protect oneself from it, and filter it so as to extract what is essential and relevant to the act of taking a decision. And the higher up in the hierarchy you go, the more this decision support information has to be synthesized – even if many more precise indicators are still available if a detailed analysis needs to be carried out at a later date.
Withholding information and obstructing information flow
Conversely, a lack of information will, obviously, also impede efficient decision-making. Often this kind of problem stems from a functional directorate that considers this information to be their exclusive property and domain within the business, and refuses to communicate the information. When the IT department is asked by senior management to set up such a project, they need total support from them, so that directorates responsible for the source data do actually hand over the information requested, agree to provide access to their information systems, and accept the setting up of formal procedures for handling this data so it can be transformed into valuable information.
Data quality and refinement
When in the end the right indicators have been defined (not too many, nor too few) the quality of whatever is to be restituted has to be assured. Unreliable sources need to be eliminated, and it’s better for an indicator to be absent than erroneous.
Data that exists in several sources needs to be identified and the repetitions stripped out, giving priority to the most reliable source (for example, taking a member of staff’s telephone number from the file issued by the local phone repository system, rather than any other source, which is at best the same and at worst, out of date to the tune of one or two relocations).
Finally, one needs to be sure that data retrieved has not been ‘adapted’ prior to transmission, applying the old principle that “not every truth needs to be told...” As a last resort, one could even associate a reliability score to all indicators published, to reflect the different data sources identified, and any known processing that has been applied to the data.
In the end, if a sufficient number of relevant indicators from proven, reliable sources are in place, BI system providers will have to deal with any potential issue of multiple meanings for the same word that might be encountered by some groups of users, whose jargon is sometimes incompatible or inconsistent with that of their neighbors, but with whom they will nevertheless share the same indicator tables. Effective training and change management support are essential if everyone is to fully understand where the raw data comes from and the transformations it has undergone to produce the available range of indicators.
Once all these snags have been successfully overcome, we can say that the foundations for an effective and relevant BI system – capable of surviving multiple updates – are truly in place.