Yesterday, we learned why predictive anayltics is important.. Today, we'll share what you can do to help your company be predictive.
Best practices to help your company become predictive
The key to becoming a predictive business is accessing the right data and creating the right models to optimize results. The good news? Much of this data already exists and is available to you. The following best practices will help you use it to advance your predictive journey.
- Know where you are now. To project where your organization is going, you need a solid understanding of where it is today. This involves developing both internal benchmarks—is company performance where you want it to be, and where you said it would be?—as well as external benchmarks: are you leading or lagging compared to competitors? Then establish clear metrics or key performance indicators (KPIs) to track your progress and identify corrective actions as needed. “We’re starting to see more development and industry alignments on standard sets of KPIs,” Caseber notes. “It’s very useful to develop your own goals using the same sets of benchmarks that are available externally; this makes it much easier for you to evaluate your performance against peers.” Follow these steps:
1) Standardize metrics: Get internal agreement on the metrics you'll use, and make sure you have access to the same metrics for competitors to establish a basis for comparison.
2) Collect and normalize: Identify where and how you'll obtain the data you need, then eliminate size/scale biases to ensure your assessments are valid. For example, large companies have bigger budgets and more opportunities for savings than small ones, so comparing savings as a straight financial amount isn't useful. Instead, even the playing field by looking at savings relative to the amount of investment.
3) Define segments: Determine the industry or peer groups that will make your comparisons meaningful. "If you're a financial service business, does it make sense to compare your operations to a small manufacturer in China? Probably not," Caseber explains. More relevant selections will strengthen the value of your results.
4) Compare performance: Critically assess your performance to identify desired activity outcomes. This involves 1) understanding what competitors in your defined segments are doing differently to achieve results that exceed yours, and 2) developing an action plan and best practices based on that information to help you attain objectives.
“Value comes only when the insights gained from analysis are put to action to drive improved decisions.” James Taylor, Smart (Enough) Systems
- Use a comprehensive approach. Tucker compares the reliability of results gained from a single-source model to a diversified one like that of Sears Holdings, which uses 500 KPIs fed into a series of predictive tools for understanding risk and revenue. The takeaway? “Don’t put all your eggs in one basket,” he advises. Instead, be comprehensive when creating your model.
- Set goals and outline steps to execute your strategy. Caseber emphasizes the importance of written goals when designing and implementing your predictive model. “It’s amazing how many companies don’t have a clear understanding of their goals when they enter into a program like this,” he says. And be sure to outline a sequential plan for deploying your model—identifying which actions you’ll take first to drive use and get the system live, and which should come later.
- Measure your performance regularly. Once you’ve set your goals, review your performance against them on a periodic basis. “It’s important to basically ask the question every three months, every six months, ‘Where are we against our plan? How are we performing?’” Caseber says. “Michael Hammer made the point that ‘you can’t improve what you don’t measure,’ and you can’t fix what you can’t see. Metrics are what you use to help you see.”
- Continually evaluate and revise. “The first time out of the chute, the system doesn’t always work 100%, so make sure your model outputs are actually producing something reasonable and realistic,” Caseber says. If percentages or results don’t look right, the model may need revision. And avoid slavish clinging to your initial metrics; as conditions change, so should your model.
“At a time when companies offer similar products and use comparable technology, high-performance business processes are among the last remaining points of differentiation.” Thomas Davenport and Jeanne Harris, Competing on Analytics: The New Science of Winning
Where can I learn more?
For additional insights and information on the value of becoming a predictive business, watch the full session on the Ariba Slideshare site or listen to Tucker’s comments in this brief interview from Ariba LIVE.
Go to our Supply Lines group to receive the latest insights and best practices for collaborative business commerce.