The rules of data governance, defined as the management of the availability, usability, integrity and security of enterprises’ data, are critical for effective business intelligence (BI) tools, says Yellowfin South Africa distributor AIGS sales and marketing director Gustav Piater.

Many modern BI tools pull data straight from source applications without passing the data-warehouse layer, and then make the mistake of allowing users to export it into Excel reports without defined business rules, leading to faulty and inaccurate reports. This constitutes a breach of data governance, which exposes the organisation to many ‘versions of the truth’.

“Spreadsheets are often the culprit. They sprout like weeds in every department, each with its own way of calculating the sales figures, each with its own formula, and more than a few with their own hidden calculation and transcription bugs,” he says.

Business must have centralised information architecture that can provide in-system governance with models governing user-, group-, content-, function- and data-level security to underpin accurate and efficient BI tools.

“In this scenario, BI is done in a closed, centralised security environment in which data is extracted and collaborated on, but only by users with rule-based access and sharing rights (as agreed between information technology administrators and the business).”

To attain one version of the truth subscribed to by the business, the platform must have a comprehensive store of business and technical metadata, which is based on shared, reused business logic, and standardised definitions in a single system.

“The difference between the basic approaches is architectural. The principles of data governance must not only be observed, they must also be implemented centrally in a single integrated environment, applied to everyone equally and ensuring a single version of the truth,” says Piater.

A single-platform approach offers several clear data governance advantages. Complete data lineage is ensured as business users and data experts alike may modify existing data and create new data. A single system reveals how it was calculated, by whom and when.

A single system also provides complete activity lineage by revealing actions that do not necessarily alter data, such as collaborations, messages and implementation.

Further, an integrated BI platform also recognises and supports two fundamentally different approaches to BI-supported decision- making namely self-service and a well- governed, centralised setup.

“These approaches occur in every organi- sation. Individual users tend to undertake highly innovative, ad hoc exploration of data to address immediate questions. Often performed in spreadsheets with minimal governance, it has long operated under the radar of effective BI governance.”

The second type is the formal or regular reporting and analysis based on preapproved data and using agreed tools to ensure repeatability and quality of results, usually associated with a data warehouse and centralised BI implementation.

The secret to BI success is to combine and enhance both approaches in a centralised process, thereby both encouraging creative use of data and analysis, and – as appropriate – to bring the results of such creativity into a more formal and governed environment, concludes Piater.