Analytics then and now


Big data problems are not new, says Vijitha Kaduwela, founder and CEO of Kavi Associates, an analytic consulting firm. The first case in point that Vijitha referenced in his talk yesterday at Analytics 2011 is a revenue generation model that he developed for United Airlines when he worked there 15 years ago.

At that time, the airline was analyzing daily flight data to estimate revenue. To improve their estimates, United decided to bring in data at the individual passenger itineraries level. This one simple change meant that they went from analyzing 3,000 flights a day to looking at more than a million passenger itineraries a day. to update this 20-year-old revenue management system required accessing booking and flight data every day, performing daily forecasts and optimizing the data within a 10-hour time period everyday. This project started in 1994 and took 3 years to see time to value, says Vijitha.

The second large analytics project that Vijitha discussed was implemented at GE Rail and had a two year time-to-value. There, he took on the financial risk aspects of owning and retiring expensive, heavy equipment. The business for rail cars goes in unpredictable cycles, so deciding when to buy, sell and hold equipment when utilization goes down can be difficult - and having millions of dollars of assets sitting around not in use is dangerous financially. While working at GE Rail, Vijitha led a project that analyzed this fleet of rail cars and other large equipment to not only understand when to buy and sell more equipment - but when and where to have each car serviced.

This optimization and risk analysis solution involved a large data integration process and collaboration between five vendors to develop the system, but Vijitha said won a GE President's award in its first year of implementation.

Ten to fifteen years ago those implementation times of 2-3 years were normal, says Vijitha. By contrast, the complex analytics projects he is implementing today are seeing time-to-value in a matter of months. For example, he has worked recently with a manufacturer that sells sensors that can be attached to delivery trucks to analyze traffic patterns and delivery times for fleet management. The company receives ten million cellullar and satellite messages per month, and analyzes that data for clients to help optimize delivery schedules and fleet sizes. The solution also offers repair event analysis, performance metrics and alternate failure analysis. It was implemented this year and saw time-to-value within 10 months.

What has changed over the last fifteen years? And what trends has Vijitha noticed over that time? Analytics projects are now synonymous with business value, he says. Now that executives have seen analytics projects succeed, the business case is easy, especially when you consider:

  • Analytics awareness has gone through the roof.
  • Technology maturity has increased.
  • Technology cost has gone down.

Analytics technology is mature, its costs have gone down, and it's becoming affordable. However, labor costs are going up and good analytics people are hard to come by.

Vijitha says cutting corners on technology upfront does not pay off in the long term. And, low labor rates do not add up to long-term savings. He recommends staffing a research center that includes training, academic partnerships and an analytics lab.

He also suggests an analytics value acceleration process with a disciplined approach that includes the following steps:

  1. A discovery panel. This step takes a few days and includes talking about analytics opportunities in your organization, formulating a vision, develop a roadmap, providing scope and cost estimates.
  2. Preparation phase. This takes a few weeks and includes digging into data, looking at systems, putting a roadmap in place, developing business process map and providing technical and services estimates.
  3. The foundation phase takes few months and includes delivering data integration in prioritized waves, delivering early value with low-hanging fruits.
  4. Acceleration phase. This can also take a few months where you deliver advanced analytics in prioritized waves.

Finally, Viji recommends choosing a strategic analytics platform that will help hide the complexity of your IT systems and isolate your analytics projects from constant change.


About Author

Alison Bolen

Editor of Blogs and Social Content

+Alison Bolen is an editor at SAS, where she writes and edits content about analytics and emerging topics. Since starting at SAS in 1999, Alison has edited print publications, Web sites, e-newsletters, customer success stories and blogs. She has a bachelor’s degree in magazine journalism from Ohio University and a master’s degree in technical writing from North Carolina State University.

1 Comment

Leave A Reply

Back to Top