The word “nebulous” fits big data perfectly. The patterns are there; the trends rippling; the insights just waiting to be drawn. But they are hazy…unclear….unformed. Many, many trees. Very big forest. A lot of green.
OnlyBoth cuts through it.
This benchmarking engine uses discovery algorithms to ferret out unique aspects of persons, places or things in a data base; compares them to others in the data base; then describes results in natural language. OnlyBoth is the AI-brainchild of Raul Valdes-Perez and Andre Lessa. They developed the discovery algorithms by studying the logic of certain tasks and noting which heuristics led to the best solutions.
As unique in its construction as it is in its ability to associate and assemble facts, OnlyBoth has analyzed – and written up – federal data from 4,813 American hospitals. Going to the OnlyBoth site, I learned that St. Agnes of Fond du Lac, WI, – compared to 14 near-by hospitals –has the lowest spending per Medicare beneficiary, yet achieves the:
- lowest rate of unplanned readmission for chronic obstructive pulmonary disease (COPD) patients (18.8%);
- lowest acute myocardial infarction 30-day readmission rate (15.5%);
- 2nd-lowest pneumonia 30-day mortality rate (11%); and
- lowest death rate for stroke patients (15.2%).
Twenty-six other distinctive facts awaited my click, each one culled and written by AI software, ready to be presented in a fraction of a second.
Benchmarking is nothing new. Googling the term renders 25.5 million hits.
Companies use benchmarking to “compare key metrics to other businesses…to see how well they are performing and identify ways they can become more competitive in the industry,” according to Study.com. Its current practice, however, is complex, expensive and time-consuming; limited in scope; and constrained in economic value, according to Valdez-Perez. Automation is the solution, he says.
“Just like ‘library search’ engendered ‘search engines’, we aspire to have ‘business benchmarking’ lead to ‘benchmarking engines,’” Valdes-Perez told me.
At the core of this engine is a compatriot to the “If…then” statement – the “only..both.” It combines two numeric attributes, Valdes-Perez explained: “Only P has both as much (or little) A and as much B.”
In the hospitals data base, this statement plays out as: “Only Meriter Hospital in Madison, WI has both as low an acute myocardial infarction 30-day mortality rate (11.4%) and as low an acute myocardial infarction 30-day readmission rate (16.2%).
And it is, of course, also the basis of the company’s name.
To appreciate the engine’s potential, consider the background of Valdes-Perez and the man who mentored him. Valdes-Perez, a PhD student at Carnegie Mellon University and later a CMU faculty member, co-founded a search software company, Vivisimo, which was sold to IBM in 2012. His advisor at CMU was Herb Simon, 1978 Nobel laureate in economics and recipient of the A.M. Turing Award and National Medal of Science.
Valdes-Perez distilled their time together into Personal Recollections from 15 Years of Monthly Meetings. Among the key takeaways from that publication: the need for the right framework, when addressing a problem; what the answers to questions would look like; and ensuring that people will care about the result. Each was considered in the development of OnlyBoth, today a laboratory curiosity, but probably not for long.
“The hospitals application is a public service and showcase,” he told me. “We'll be launching another in September.”
Valdes-Perez did not say what this new application would be, only that cloud-based services will be a great new playing field for automated benchmarking, across industries. Like the data it plumbs, OnlyBoth is all about the possibilities.
“If data is available and if organizations want to know how they are doing, where they could improve, and what's best in class, then the engine will generate the answers,” Valdes-Perez told me.
OnlyBoth’s ability to mine and present findings in natural language could be taken as, well, a bit threatening to professional writers. It does, after all, do what we do – find the significance of data and report it. But I’m not worried.
OnlyBoth can knife through a massive data base. It connects the dots nicely. And, yes, it draws insightful conclusions. But…
Can it come up with clever endings for its stories? Riveting questions that keep its readers hanging?
This article is published as part of the IDG Contributor Network. Want to Join?