Why advanced analytics projects are different from traditional IT

Senior Managing Consultant, IBM

From our recent analytics project experiences (optimization and predictive modeling) in Canada, we have learned that a company’s senior leadership’s ability to acknowledge the unique challenges, support an "agile-with-discipline" delivery approach and deliver with excellence will be the true market differentiator for the next decade as enterprises start to invest in advanced analytics to transform their organizations.

To further illustrate, here are the top seven reasons why delivering an advanced analytics project is different from a traditional IT project:

  1. advanced analytics.jpgAnalytics project leadership (both on vendor and client side) must have a multi-disciplinary background to be able to orchestrate daily delivery challenges arising from business domain (function and industry), operations research modeling techniques, advanced analytics technologies and traditional IT development & project management areas.
  2. Agile development with frequent business user check-points must be the delivery approach. The inherent challenge in analytics projects is asking the right questions, not necessarily finding the answers. A small and focused team with iterative approach works best to define and redefine questions, document the requirements, build the models and validate results in a short iterative cycle.
  3. Data profiling must be completed prior to developing the first iteration of the model algorithms, and then continue to narrow down and enhance based on input data-set to establish pre-condition checks for the continuous variables to find optimal solution sets for the objective functions.
  4. Visualization of the model output early is very important to continue to confirm requirements, keep business engaged, force tough business trade-off decisions and show progress of the project to senior executives and other key stakeholders.
  5. Non-deterministic outcomes need to be managed with interim and ultimate success criteria, which must be objective, measurable and logically tied to the initiative purpose and anticipated benefits. This is critical to manage executive stakeholders on all side and report progress.
  6. Testing as means to change management to account for user knowledge gap, cultural shock, related resistance and time needed to be super-users of the application post go-live.
  7. Optimization resource management is quite tricky due to the scarcity and PhD personalities. Project leadership has to recognize and accommodate differences in work style, interactions and recognition preferences of the team members in remote locations.

We are soon to publish a series of white papers that focus on this notion of uniqueness in overall implementation strategy, gathering business requirements, data profiling, testing approaches and delivery project team capabilities. We are very curious to learn about your experiences and feedback on this topic; please leave a comment below to keep the discussion going!