KDnuggets Home » News » 2014 » Mar » Opinions, Interviews » Why Predictive Analytics Marketplaces are not taking off, and how to fix it ( 14:n05 )

Why Predictive Analytics Marketplaces are not taking off, and how to fix it


Three main hurdles holding back Predictive Analytics Marketplaces are a highly fragmented data mining tools market, limited support for customization, and lack of commitment. We examine how to overcome them.



Guest blog by Madhu M. Reddy, Snap Analytx, Mar 1, 2014.

An earlier article on KDnuggets noted that analytics marketplaces have the potential to be the “next big thing in Big Data”. However, analytics marketplaces are yet to take off; their impact minimal. I believe there are three main hurdles holding back analytics marketplaces from taking off:

  • A highly fragmented data mining / predictive analytics tools market
  • Limited support for the customization of predictive analytics  models
  • Lack of focus and commitment by marketplace operators

Size Matters
Like any other marketplace a critical mass of consumers and producers is required to foster a healthy trade. Analytical marketplaces have to be big enough to reach a substantial fraction of the total data science market to attract model creators. Puzzle That’s one of the reasons app developers flock to iOS and Android marketplaces but not so much to the Windows phone or Blackberry marketplaces.   However, the predictive analytics and data mining tools market is highly fragmented. Even the biggest player (SAS) owns less than a third of the total market and the long tail of vendors have market shares in the low single digits. One can see the extent of fragmentation in the recent poll on this site.

As a result of this fragmentation, marketplaces affiliated exclusively with low-share vendors are unable to attract the critical mass of developers required to jump-start the virtuous cycle necessary for success.  It is feasible that over time market consolidation will address this issue. Virtuous cycle of multi-sided marketplaces Meanwhile,Snap Analytx, a new startup, is addressing this issue by being open: Snap Analytx supports a number of tools and technologies (including SAS, SPSS, Mahout, Java, C++, Matlab, PMML, R and Python) that together add up to a sizable market share. The current set of models in the Snap Analytx catalog showcases a wide variety of business use cases such as: customer churn, item recommendation, customer’s propensity to buy products and services,recognizing hand-drawn digital images and resolving author-name ambiguity. More models are in the pipeline to be added to the catalog.

Customization – One Size Doesn’t Fit All
Apps in a typical digital marketplace are more or less commodities. Predictive Analytics models on the other hand, almost always have to be customized to meet the specific business needs, data sets, missing attributes and performance requirements of each customer scenario. Inability of any existing marketplace to enable easy customization of predictive analytic solutions has been a significant barrier to the adoption of predictive analytic marketplaces. With this realization, Snap Analytx is designed, ground-up, with customizability built-in. Snap Analytx enables – in fact encourages – direct collaboration between model authors and users to facilitate customization.

Focus and Commitment
Marketplaces can succeed only if all the stakeholders – buyers, sellers and the marketplace operators – have an "all- in-this-together" attitude.  Model authors want to invest their time and energy in marketplaces committed to their long-term success and to ensure that their modeling skills and efforts find longevity and visibility.  They can sense, and will avoid, marketplaces that are ancillary to the operator’s main business.  I believe this has been one of the factors inhibiting the adoption of marketplaces. For this reason Snap Analytx has taken a ‘pure-play’ approach – the marketplace is its only business and its raison d'etre.

Predictive Analytics Marketplace: A higher goal
All of us in the predictive analytics and data science community should be rooting for the success of predictive analytics marketplaces. In addition to all the obvious benefits, I believe that marketplaces can have a much bigger impact on the society at large – marketplaces have a role to play in addressing the well documented, and much discussed, issue of the shortage of data scientists. However, that’s a topic for another day!

Madhu M. Reddy Madhu M. Reddy is the CEO & Founder of Snap Analytx. Previously, he held senior positions at Microsoft, Oracle, and HP.

Sign Up

By subscribing you accept KDnuggets Privacy Policy