BI Dashboard

The global operations organization in a multinational hardware and software provider suffered from fragmentation of data sources and diverse communication metrics across multiple teams. Tracking meaningful key indicators was difficult and time consuming, requiring manual collection and redaction of reports and distribution of them by email.

Updates were infrequent and often out of date by the time they were being used. There were further pain points for fine grained security requirements and access levels: some data was intended to be visible to senior executives only, others should be available at various levels below, all the way down to folks involved in project execution. This resulted in blocking the company from executing insightful Business Intelligence and made it extremely difficult to comply with security requirements for already existing reporting, effectively forcing manual management of mailing lists for report distribution.

A further key requirement was to make insights actionable through a dashboard to increase its adoption and the speed of response to critical requests and situations. This is something that requires a dashboard to be a platform for micro apps, more like a phone’s home page. It is something that is not typically found in standard visualization frameworks in reporting tools.

The simplest solution was to create an Extract/Transform/Load system with an extended report authoring and distribution application, one that also persists additional user preferences and enforces all security constraints required by the business.

Even in the first incarnation of this infrastructure we adopted a micro services architecture, when this was novel and still rarely used. We pre-processed all manner of data into a standardized report data model to describe and visualize the most common data aggregations. It was then easy to extend these as needed for new visualizations. On the user side, data was packaged into serialized components to allow for fast interactivity and filtering for users. We used a decoupling strategy in order to track down and understand any potential issue easily and standard Open Source components.

The reporting was done in an active framework, so as to make it easy to drop in simple action controls on them, rather than just being passive in the manner of most standard out-of-the-box reporting technologies. This approach makes dashboards a one stop shop for users to not just track stuff, but action it as needed.

On the security side the solution was to enhance standard SSO layers with permission and preferences to manage access to the dashboard components. This was parametrized and easy to manage for administrators.

The first version was deployed to production within 2 months after finalizing requirements. It was continuously extended with features for an additional 8 months. Deployed in 2014, it is still in use at the company where we developed this; the portfolio of reports has taken on a life of its own, serving up to hundreds of operations managers and engineers.

The requirement to extend interactivity of the reports came subsequently and was quickly added, allowing for smart extended filtering and for the embedding of mini data-driven application. Last but not least, reports can now be grouped and rendered in sections.

Just imagine your regular, business KPI tracking report PDF for a 100,000+ employee company, which used to take days of many people’s time to put together, now generated on demand any time you want, with charts and graphics too :)