When you feel under the weather, you go to the doctor for a diagnosis. On most occasions, the doctor prescribes medication that could lead to undesirable side effects such as drowsiness, weight gain or even an allergic reaction -- the likes of which may discourage you from taking the prescription in the first place.
The same can be true for prescriptive analytics in an IT setting. Like a written Rx from a doctor, prescriptive analytics can be just what the doctor ordered to fix whatever ails your business, but there can also be unintended consequences of a wholesale embrace of data analytics and mining.
During SearchCIO's April tweet jam, we asked participants, "Are there situations where you wouldn't want to take the prescription even though you know it works?" In response, retail giant Target was called out for its analytics faux pas:
A2. Yes! You all know about Target and the pregnant teen? I know other stories too. Just because you can, doesn't mean you should #CIOChat— Andi Mann (@AndiMann) April 30, 2014
Target wasn't the only namedrop during the discussion. Other -- somewhat humorous -- examples of analytics-driven automation gone awry included social media friend-suggestion tools, automated marketing messages and group email:
@AndiMann or your ex-wife— Rich Lauwers (@richlauwers) April 30, 2014
@AndiMann Group mailing; forwarded out of office created a loop, filled the mailbox of 45 participants and STOPPED server; full logs— Rich Lauwers (@richlauwers) April 30, 2014
SearchCIO has reported on the importance of mitigating security risks associated with human error by creating automated processes; but humans shouldn't be completely removed from data analysis. Organizations looking to prevent instances of machine error should consider human intervention somewhere in the analytics process.
More on prescriptive analytics
Prescriptive tips from CRI and Revlon
The future of business analytics
This echoes something Nate Silver, analytics expert and the founder of data-journalism hub FiveThirtyEight, has been talking about for quite some time: turning analytics over to the machines. At the Gartner Master Data Management (MDM) Summit last year, Silver cautioned, "Whatever you do, don't give in to the big data hype."
Our #CIOChat participants advised finding a balance between machine automation and human intervention when it comes to analytics:
A2: Key is good balance of machine algorithms and human judgment. Machine will never have all info, so human can add impt value. #CIOChat— Tom Doub (@tomdoub) April 30, 2014
A2: Technology is so new; balancing act is uncharted territory. Especially impt in domains that rely on lots of judgment #CIOChat— Tom Doub (@tomdoub) April 30, 2014
Humans can't always intervene in order to set analytics straight, but when they can, who should lead the charge? SearchCIO Associate Site Editor Fran Sales asked tweet jammers:
In addition to those mentioned above, are there instances in which IT shouldn't take the analytics prescription, even when the data readily presents itself? When should humans intervene, and who should lead that effort? To read more conversations from this tweet jam, search #CIOChat on Twitter. Follow @SearchCIO to learn more about out next Twitter discussion scheduled for Wednesday, May 28, at 3 p.m. EDT.