Agile Analytics, Pt. 2

Discipline and strong focus still basis for new approach

In this post I promised I would give some examples of “agile analytics.” One of the people who turned me onto this topic was Anne Robinson at Cisco Systems. She led a project to implement a new statistical forecasting project at Cisco (http://agendabuilder.gartner.com/bi7/WebPages/SessionDetail.aspx?EventSe...), and she used an agile approach from the beginning. The project was planned for 18 months, but it was broken up into several short-term deliverables.

The first subproject was to test out the SAS forecasting engine, which proved to be a good fit for the project. The second was to determine whether an analytical approach to forecasting improved the accuracy of forecasts; fortunately, it did. The third subproject involved introducing better models. The fourth involved preparation for production scale, and the last technical component was automation of the models. The team took 2 to 3 months for each subproject.

At each stage of the project, the team showed the results to the executive sponsor. He understood analytics and gave a lot of support, and the frequent reviews ensured that the project was consistent with his vision.

As with most analytical models, the new forecasting approach also embodied a new way of thinking and working within Cisco’s supply chain. Robinson found that early on in the discussions of the project, people within the planning community were skeptical that better results could be achieved. “The prototyping process helped them buy into the new approach along the way,” she notes. The new forecasting approach was a big success overall, and Robinson credits the agile approach as a major factor in its success.

Hewlett-Packard is another organization that uses agile analytics. Its SPaM (Strategic Planning and Modeling) group (http://en.wikipedia.org/wiki/HP_SPaM), an internal consulting organization, uses agile and prototyping approaches on many of its projects. Scott Ellis, who used to head the group, told me, “If I had to do SPAM again, I would use prototyping on every project that involved IT.” Many of the projects the group undertakes do involve some sort of tool development, and they find the agile approach to be both faster and more likely to satisfy their internal customers.

I’ve heard a few other leading analysts say they are moving in this direction, and I think we’ll see a lot more in the future. Particularly for analytical projects that involve new systems, processes, and user behaviors, agile methods seem to be the way to go. Of course, their effective use still requires discipline and a strong focus on results.