On July 10, industry experts, data management practitioners and visionaries convened on Twitter to discuss trends in predictive analytics during our bi-weekly #IMChat. Predictive analytics promises great rewards: 55 percent of companies say that predictive analytics increase revenue and 68 percent say they use them to achieve a competitive advantage. But how can you make predictive analytics actionable? And how can you do it without alienating your customers?

Information Management Editor-in-Chief Julie Langenkamp-Muenkel (@JulieLangenkamp) moderated a lively chat that addressed all these questions and more, some of which is encapsulated below:

Q1: What benefits does your organization seek from predictive analytics? 

@DnBUS: Predictive analytics = critical for forecasting, marketing, customer service, product offers, fraud detection. @ventanaresearch

‏@InfoMgmtExec: Looking backwards (BI) or "What just happened?" have been norm for far too long. Predictive #Analytics answers "What is next"?

@JulieLangenkamp: @ventanaresearch also shows competitive advantage, increased productivity and operational efficiencies http://bit.ly/15rT15W

‏@WhitneyAEden: Foreseeing the future! Jk. You can make data-driven, educated decisions. E.G., when should I tweet? What types of stories perform best on Facebook? Those are examples from my vertical (publishing) 

Q2: How do you frame predictive analytic possibilities as “actionable” for business?

‏@JulieLangenkamp: probable outcomes/scenarios are one thing, but what to do about them is quite another

@PeggedSoftware:  Predictive analytic possibilities as “actionable” if you use them to rank possibilities

@Melianthe: You need quick wins that bring revenue for exec approval, but identifying new customers is the long-term prize

@DnBUS: First articulate the question/business problem that needs to be solved. Find and analyze the relevant data.

@IMJustinKern: The biz core of PA models. I like @predictanalytic's push for direct terms (i.e. retailer's likely hat sales, not methodology)

Q3: How do you pitch to customers using predictive models without being “creepy”? 

@WhitneyAEden: Don’t do what Target did and start emailing coupons for diapers. http://onforb.es/15u62vK

@DnBUS: It's interesting that some Twitter users complain about vendors via Tweets, yet take offense when vendors offer help. 

‏@IMJustinKern: At #CAOS2013, many agreed that biz needs MORE customer info, not less, for less "creepy" predictive analytics

@WhitneyAEden: Honestly, consumers should educate themselves about how their info is being collected (social media, geolocation, etc.)

@IMJustinKern: On keeping Obama campaign analytics from being creepy, Dan Wagner/@CivisAnalytics noted model specificity via past voter contact 

Q4: How do you balance starting a predictive analytics initiative with a hunch/curiosity w/out introducing bias inferences? 

@WhitneyAEden: By not ignoring them. If you’re aware of potential “hunches,” you can balance them with data/facts

@NeilRaden: I find EXTREME contradiction in a data scientist with "domain knowledge" and PA without bias. All quant methods have bias. period. 

@JulieLangenkamp: Reminds me of a quote: If you torture the data long enough, it will confess. -Ronald Coase. Need to avoid such a practice.

@DnBUS: Make sure you're not weighing the wrong data elements too heavily. Use clean data to ensure beginning dataset isn't flawed.

Q5: If everything fits nicely into a predictive model, where does experimentation fit into the equation? Is innovation at risk?

@JulieLangenkamp: @infomgmt blogger Steve Miller recently wrote a great commentary about this http://bit.ly/16kC9v3

@WhitneyAEden: It’s crucial. The future is somewhat unpredictable (shocking, I know). So you should be prepared for a variety of outcomes.  

‏@InfoMgmtExec: Managing Lifecycle of Predictive Models is essential to success. A/B Testing, Champion-Challenger, Optimization all required.

@DnBUS: In the Internet of Things, nonlinear effects will be more common (smart grid & real-time process monitoring).

‏@IMJustinKern: Hunches and tinkering - even when it's not the "cleanest" data - can at least show you're on the analytic right track

What industry seems most ripe for advances with predictive analytics?

@DnBUS: Healthcare is ripe for advances with predictive analytics.

@WhitneyAEden: Retail probably has the most ROI. But any industry can benefit from it, from shipping to publishing to law enforcement & beyond

‏@InfoMgmtExec: All industry sectors are deriving benefits from Predictive #Analytics. >>Use Cases. Lack of imagination is the only limitation.

@DnBUS: Another industry that has much to gain from predictive analytics: risk management. 

Join us July 23 at 1PM ET for the next #IMChat on open source.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access