Headlining this week’s Predictive Analytics World conference in New York, “Competing on Analytics” guru and best-selling author Tom Davenport declared results so far to introduce more analytic power across enterprises to be lukewarm at best. What is needed, he said, is a way to embed analytics deeper into day-to-day business operations. 
“Analytics need to be industrialized, rather than a craft,” he told the audience. “Analytics need to be repeatable, rather than a one-off activity.”
Tom Davenport was co-author of the seminal business book of the 2000s decade, Competing on Analytics: The New Science of Winning, as well as a follow-up work, Analytics at Work: Smarter Decisions, Better Results, which carries the message of Competing on Analytics forward with practical advice on becoming an information-driven business.
He observed that while analytics have been on the scene for decades now, they have remained relegated to wonks and quants in organizations, which is not getting the most out of the capabilities a data-driven business can potentially deliver.
“Analytics need to be easier to use,” he said. “A dose of process thinking would be good for analytics.” That’s because up until now, “analytics have been a craft phenomenon; typically a one-time thing addressing one thing or one decision.”
Analyzing data patterns is still “quite labor intensive,” he adds. “That’s good for those who labor in this field, but not good for productivity. We need to become more industrialized, where analytics are more embedded in day-to-day processes.”
He cautions, however, that the automated decision making that analytics brings needs to be self-learning, flexible, and constantly scrutinized. “Financial firms had great automated models, which worked until the financial crisis,” he says. “The models fit the market when real-estate prices were rising, but not for the other way around.”
The best approach to introducing greater analytic capabilities to the organization, he says, it keep things simple and understandable.
First, analytics proponents should outline their capabilities on simple “menu cards” that provide the key offerings. For example, a menu card can offer options including market testing, promotion targeting, promotion effectiveness analysis, and so on. “Analytics is not about the math,” he said. “It’s about the relationships.”
Secondly, Davenport observes that analytics can be offered through simple, single-purpose mobile apps, who’s utility is quickly and easily grasped by business users. “I’ve heard of 50 analytical apps for the iPhone so far,” he points out. Examples include a nursing productivity app, a truck loading analysis app, and a social sentiment analysis app.
“There are so many decisions that can benefit from analytics… when can I turn them into a cute little app?” he asks.
