We use cookies and other similar technologies (Cookies) to enhance your experience and to provide you with relevant content and ads. By using our website, you are agreeing to the use of Cookies. You can change your settings at any time. Cookie Policy.
Start a Discussion

Will Predictive Analytics Be the Next Big Thing For BI?

Vote 0 Votes
According to this article on TechTarget (the company that now owns ebizQ), predictive analytics is the next battle ground for BI.  Do you agree, or would you say it's something else?

16 Replies

| Add a Reply
  • Somehow I am biased about this whole concept of predictive analytics. In my opinion it’s more towards speculation rather than anything else! This is more so because you never know what’s waiting next around the corner – at hindsight everything looks so easy and meaningful!

    I think that given today’s volatile business environment and with moving targets it would be a real challenge to ascertain a workable model which is cost effective from a business perspective.

    Consider, the latest volcanic eruption where the Airline industry lost millions of dollars – the question is would a high tech nerve center for monitoring have saved the Airline industry? If yes, would it be cost effective? – knowing well after years of R&D in the weather prediction department the Met still falters big time!

  • Predictive Analytics has a fancier name than it what is actually needed and currently in demand in the marketplace. When we think of Predictive Analytics we have this vision of a Delphi Oracle sitting right next to the Boardroom giving out predictions like "Go Forth and Concentrate on 46 inch 3-D LCD TVs. That's the next big market!" :-)

    Reality is that predictive analytics constitutes an experimental platform where companies want to test associations between Dependent Variables (Sales Performance of a product in a region, Margins and Profit Contributions of products, Average Handle Time in a Call Center, etc) with Independent Variables (Demographic Information, Economic Indicators, Training Courses for a Call Center agent), etc. Before you can build a predictive model, you need to test and see if there are associations between variables. Once you do that then you have a predictive model that is reliable and then it can be on auto-pilot, spitting out predictions as to what will happen.

    We are at this stage of evolution where datamarts have lots of data but now need tools that can associate different variables, test if there are correlations and a build a predictive model.

    It is taking off as we speak - companies are using demographic data, charitable contributions, even political affiliations to predict whether a 42inch or 36 inch LCD tv is better poised to sell in a particular zipcode!

    An area to watch definitely!

  • Oh forgot to add that seismologists still can't predict a possibility of an earthquake or a volcanic eruption to a specific day or a month!

  • I’m a big believer in Predicative Analytics. PA has already made its mark in many industries and use cases where massive amounts of data need to be synthesized and mined. As we see more BI applications being used in the operational aspects of an organization, we will see more models being applied that will allow managers to make better choices.

    So what is stopping the wide use of PA in Business Intelligence? I still think there is a fundamental communication issue. Statisticians and to a lesser extent, data miners, have a language unto themselves and tend to speak at the function or algorithm level. This is not the effective means of communication with a line of business manager. Debates on the merits of gradient boosting, least angular regression splines, neural networks, linear and logistic regression, partial least squares regression...put line of business managers to sleep at best, at worst create a barrier to trust and acceptance.

    The way around this issue is to bridge the applied technology and techniques with two basic but very important questions. The first; what is the goal of the additional analytics and can it be linked to a strategic objective of the organization. For instance, do we expect to gain operational efficiencies by reducing costs, to detect fraud, quantify risk, create new opportunities to generate revenue through up sell or cross sell, provide better insight into the behavior of your most important customers? This might seem obvious, but is often overlooked.

    The second and more challenging question is the following; if you had additional insight from the data, what actions would you take to take advantage of this new insight? Simply put, will the organization accept the recommendations “from a computer? and make it part of the operation? IMHO, this is the organizational challenge that still holds back the widespread use of PA.

  • Gartner had identified "Advanced Analytics" replacing the term business intelligence as one of the top ten strategic technologies and trends in 2010. This included what is described as the third step in decision making; which is the ability to predict and optimize based on this.

    While I do see predictive analytics as being extremely important in the evolution of BI and providing deeper value; two drivers will dictate this; user adoption and vendor capabilities. Until now PA has been performed through sophisticated modeling and forecasting tools (SAS, SPSS, R); which is not designed for the inline business user. For general user adoption to take hold; this complexity must be transparent along with a layer of abstraction or continuity between determining "what happened" and "What will happen".

  • user-pic

    Predictive Analytics can indeed provide some value, but in terms of being the "next big thing", I don't think so. PA has been around for some time and several of the big BI players have tools or point solutions around this, and for a variety of reasons have not gained traction beyond the "wonk-set".

    Before PA can become the "next big thing", BI itself must become much more mainstream - as mainstream as the spreadsheet for corporate decision making and decision-communicating.

  • Predictive analytics (in one name or another) has been "the next big thing in BI" for at least two decades. I am quite certain it will remain "the next big thing in BI" for the next two decades.

    Meanwhile people are still not getting their basic reports. Their is data is either locked up and inaccessible, or spread across the internet (wait till EHRs become common). And IT standards (i.e. Microsoft sycophantism) trumps all the good ideas.

  • Simulation optimization is one of the predictive analytics technologies that has been in use for many years outside of what is formally called BI. That BI is just now becoming aware of PA is a positive factor, albeit a bit slow in adoption. Change and acceptance of "something new" is typically resisted initially but eventually occurs after the pioneers publicize the value received.

  • I don't see this being a front in any BI battle per se, as a BI tool is just an enabler for analytical capabilities; the fact of the matter is that companies still struggle with fundamental issues such as data quality, completeness, and integration, let alone the ability to contextualize and correlate that data. The early adopters and pioneers in what we call predictive analytics - i.e. the folks at the Tom Davenport level of thinking - aren't sitting back waiting for BI vendors to add these capabilities into off-the-shelf services, they're creating the capabilities either themselves or in concert with BI vendors to create competitive advantage through the use of advanced analytics and modeling. For the rest of the universe, simply getting BI implemented and providing value by adding context to historical data is a victory in and of itself; by the time BI vendors have predictive capabilities in a commercial offering, the service will have been commodified anyway (the Nick Carr / IT Doesn't Matter paradigm).

  • Predictive analytics may be next, but it's running against lots of other candidates. It probably will be big -- someday. It'll definitely be a thing -- another set of tools that are slightly better than a finger in the wind to tell us which way the wind's blowing.

  • I think given the current encumbrances with businesses such as technology/culture conflict with BI, cost-effectiveness of the cloud, ROI of social presence and networks, PA is going to take a while to come into mainstream market. PA is still struggling to cross borders between incubation labs and the marketplace either because of application feasibilities or simply market demand. It will be interesting to wait and watch how this story unfurls. Shall i just say, 'crystal ball gazing' for now?

  • If Insights from BI systems can be categorized into Past, Present & Future, Traditional BI has done a great job in helping analyze historical (past) data. It is in the present (actionable BI) & future (predictive analytics) that the next battle-lines are drawn.

    For actionable BI, insights from analytical systems have to be integrated with Business Process Management tools, ERPs etc. so that these insights can be converted automatically to appropriate changes in the business process, leading to better ways of managing business transactions.

    The other area of focus will be 'Predictive Analytics' which exists to answer the question of 'How will the business be in the future'. This involves modeling Human Behavioral systems using statistical techniques and requires precise interpretation of the output.

    Whether actionable analytics (present) or predictive analytics (future) will be the next immediate battleground for BI is debatable but whichever it is, interesting times are ahead for BI practitioners.

  • user-pic

    I see some very good insights here! While I agree with most of the posts, I would like to say predictive analytics (PA) depends on the BI maturity of an organization. While many organizations are struggling with their basic data (completeness, correctness etc.) there are others who have a fairly stable environment and have mastered the basics of operational BI. The first step in PA would be to extend the historical trending to the future. The more detailed analysis would be based on specific areas - for e.g. trying to analyze the usage pattern of a credit card holder. If the card usage drops dramatically over a fairly long period, the chances are that the user will close the account or will leave the account dormant. Based on this, the credit card company can pro-actively come up with incentives for the user to more actively use the card. This scenario can be extended to more complex events and a neat working model of a PA will emerge.

    While it will be impossible to predict how the market conditions will be or acts of God etc, the basic business scenarios can be predicted fairly accurately, based on customized models.

    I definitely think PA will be the next big thing in BI. However, it may not be available as a "off the shelf" tool or "one size fits all" tool. It may be more customized for individual businesses.

  • Interesting question. In my view, predictive analytics implies the starting point of BI. Whatever we have had so far isn't really intelligence by definition, its simple information delivery. Intelligence means you can potentially influence the outcome, else its MIS.

  • Predictive analytics has emerged as a critical business platform across various industries. The economic downturn has pushed decision makers to ask for even more analtyics. With the proliforation of "standard" transaction reporting systems, the only true IP will in fact be the mathematical models built for decision support using both internal and external data.

    In recovery, our world is different and our organizations need to be different too. With
    predictive analytics, we look far beyond current conditions, opportunities and risks. By doing so, we can make the changes needed to enhance our efficiency and reach our business potential.

    To steal a phrase "The only thing to fear is fear itself"; it seems like the only thing that impeeds the progress of Advanced Analytics at large corporations is the belief in mathematics itself.

  • Am completely new one to Predictive Analytics ??

    Q1 -
    Is it something that I need to perform statistical analysis such as linear regression,multivariable regression , Polynomial regression etc,. to do PA ??

    let's say my r square is .99 that doesn't mean my prediction/speculation is accurate ?? As we aware innovation is a key game changer, if someone brings major differentiator definitely my business can nose dive .then whatever prediction that I made with the statistical tools will not be applicable ?? In this case, can I say prediction will not be a driving engine ??

Add a Reply

Recently Commented On

Monthly Archives