Dave Wright, ServiceNow sits down with Dave Vellante & Jeff Frick at ServiceNow Knowledge 2017 at Orlando, FL #theCUBE #ServiceNow #Know17 https://siliconangle.com/2017/05/10/good-predictive-data-analytics-model-never-finished-knowledge17/ Why a good predictive data analytics model is never finished A predictive data analytics model is like a puppy: not a great investment unless one is willing to train and retrain it continually, according to Dave Wright (pictured), chief strategy officer of ServiceNow Inc. Wright outlined how the company uses Machine Learning in the analytics tools, including an Intelligent Automation Engine, it announced at ServiceNow Knowledge17 in Orlando, Florida. Customers sparked the ML integration by asking ServiceNow to help them not just set targets, but predict when and how they might hit them, Wright told Dave Vellante (@dvellante) and Jeff Frick (@JeffFrick), co-hosts of of theCUBE, SiliconANGLE Media’s mobile live streaming studio, during Knowledge17. (*Disclosure below.) “So what we had to do then was augment the whole performance analytics suite to be able to do predictive analytics,” Wright said. Of course, it helps to start with an intelligent model at the outset of training. ServiceNow tested its auto categorization with machine learning, which will appear in its “Kingston” release later this year, inside the company, Wright stated. Out of the box, it automatically categorized ServiceNow’s records with 82 percent accuracy, he added. ServiceNow wanted a truly intelligent model that could tell the user when it simply did not know a particular thing. At 75 percent certainty, for instance, the tool should just say, ‘I don’t know,’ Wright explained. The good news is that “over time, you get to reprocess the things that you don’t know, and that percentage gradually goes up,” he said. Volume of data is another consideration. “If we’re going to train auto categorization, we need between 50 and 100,000 records to be able to get to a degree of accuracy,” Wright stated. ServiceNow’s Jakarta release for governance, risk, and compliance also uses constant iteration and training to make predictions. “The more and more examples you get, the more you can start to predict, so you can say, ‘As soon as I get that precursor, I have a level of confidence of when we’re going to see the next event,'” he said. One-size-fits-all fears Some might be concerned about ServiceNow using their company’s data to train models and then give them to competitors, but Wright said it has no plans of doing so. “You don’t get a generic ML where we look at everyone’s instance and train across that. We can only train for your instance, and that’s because everyone does things differently,” he said. However, in Jakarta’s expanded bench-marking, subscribers may share data, which ServiceNow will render to show how businesses stack up against each other. Watch the complete video interview below, and be sure to check out more of SiliconANGLE’s and theCUBE’s independent editorial coverage of ServiceNow Knowledge17. (* Disclosure: ServiceNow Inc. sponsored this Knowledge17 segment on SiliconANGLE Media’s theCUBE. Neither ServiceNow nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)