Six Steps to Governing Analytics

This approach deals directly with behavioral change and the actions to help ensure that value is being achieved

Information Management Magazine, July/Aug 2010

Greg Todd

Predictive analytics takes the information made available through descriptive analytics (historical facing) and combines it with more sophisticated statistical modeling, forecasting and optimization techniques to derive insights which help to anticipate the impact on business outcomes. Where are organizations when it comes to leveraging predictive analysis? Research shows that while some organizations may analyze data to predict what might happen in the future in terms of competitor activities, market trends, product/service development, risk management, financial/economic trends and skill requirements, many organizations are still using predictive analytics only to a minor extent, if at all.It's clear that companies need to move from descriptive analytics (the "what") to predictive analytics (the "now what?"), from "what happened?" to "what's the best that can happen?"
During previous economic downturns, companies that thrived used data-derived insights made by informed decision-makers to produce lasting competitive advantage...



Today, some companies have this down to a science, while others are just starting to acquire a better understanding of how they can use predictive capabilities to increase their revenue stream and reduce their costs. For them, it often means implementing analytics across a number of business functions, such as supply chain management, customer acquisition and retention, talent and organizational performance, finance and performance management. These are the core domain areas that companies want to analyze and be more predictive in how they forecast information and optimize their existing capabilities.
Organizations considering analytics across the enterprise should follow a six-step approach to governing their analytics that will take them from the "what" through the "so what" and into the "now what." This approach deals directly with behavioral change and the actions to help to ensure that value is being achieved.

1. Identify the key targets and metrics. There has been considerable activity of late surrounding getting data right and making sure the right data is available to make decisions. Identifying key metrics for analytics, though, involves a different set of discussion points. The focus here is on identifying the strategic business issues that can benefit from predictive analytics: How should you be optimizing your supply chain? How should you be forecasting your financial numbers? How can you become more focused on customer acquisition and retention? And what are the metrics that say you're actually doing a good job in those areas?
2. Generate insights. This is where the software automation used helps generate a forecasting capability, an optimization engine, a predictive model or pattern recognition. This is the "aha" moment where you say, "Okay, I now recognize that 80 percent of my turnover problem is occurring in customers who've been with us less than three months."
3. Validate insights. Can I be confident the insight generated is true? Has the data been prepared properly? Have I taken a broad enough selection of data to be able to ensure that the pattern identified is reasonable and accurate? Validating insight is as important as generating it, because once you've validated it you can move on to creating programs that will improve business outcomes.
4. Plan and execute decisions. This is where differentiation in the marketplace occurs. It's where companies can pull ahead of their competitors because they can now take a proactive step toward managing whatever the insight has told them to do. For example, logistical regression can identify the variables that contribute significantly to turnover in new users. This in turn allows a company to monitor its customers and develop strategies to reach out to at-risk customers before they churn.
5. Realize value. If you've identified customers that are likely to purchase from you a second time, for example, you can now target these individuals and make sure you're actually achieving some lifts in your sales figures based on this targeted activity.
6. Monitor performance over time. You want to be able to track over time how well you're doing as compared to your original baselines. So not only do you have to achieve value, but you have to ensure that it's occurring in an ongoing way. And being able to monitor - weekly, monthly, quarterly - becomes a huge part of ensuring the value of the analytics and how it's changing the organization's bottom line.

This six-step approach provides organizations with a good structure to help clients see the end-to-end picture. Most recently, it helped the Royal Shakespeare Company create a new segmentation model using a number of additional variables for the first time. This identified a "Golden Geese" segment that is more likely to buy expensive tickets, attend a Saturday night performance and book more than three months in advance. Less sporadic in terms of attendance than the regulars, these Golden Geese have become an invaluable segment to market the RSC experience to and a prime target for expanded packages.

Greg Todd is Accenture Analytics, Executive Director - Technology. Todd has been a technical architect for over 17 years, with an increasing focus on the reporting and analysis of information, both structured and unstructured. Through leading many enterprise resource planning implementations, he has seen how critical the management of information is to the bottom line of any corporate strategy.