Predictive Analytics in the Cloud: Questions and Answers

November 21, 2013

in Analytics, Data Mining, Decision Management

Share

I recently gave a webinar on the results of our Predictive Analytics in the Cloud research (you can watch a replay here on Information Management). During the webinar we had some great questions and, as I did not get to all of them, I promised to write a blog post to address them. Meanwhile the report itself is now published – check it out here. I have paraphrased some to make them work in a blog post.

  • How different do you think business processes will need to be changes if analytics is on the cloud vs locally in house?
    I don’t think that the use of cloud v on-premise predictive analytic solutions makes a difference. Automating decisions using predictive analytic drives significant change in business processes but the deployment mechanism is unlikely to change the “shape” of the process.
  • What are the typical pre-requisites for cloud analytics enablement? How critical is Application Data already being persisted in the Cloud?
    This is a great question. Clearly having data in the cloud is very helpful and accelerates cloud-based adoption. It’s not a pre-requisite though as most cloud-based predictive analytics solutions have good tools and infrastructure for moving data to the cloud. I would say the most important is familiarity with cloud-based approaches. Without that the loss of “control” can seem overwhelming in the context of predictive analytics.
  • What are the future opportunities in predictive analytics?
    Deeply embedding predictive analytic into operations to improve how customers are treated at every interaction.
  • What enabling technology do you see as the most impactful over the past 5 years relative to operationalizing analytic capability?
    I would say it is either business rules management systems or in-database analytics. Both have made a real difference. Check out our technology report for more on business rules and this thought leadership piece on in-database analytics.
  • What is the difference between cloud based modeling and cloud based embedding?
    Cloud-based modeling is about building the predictive analytic models you need. Cloud-based embedding is making sure that the results of those models, the scores, are available to the systems and processes in which those predictions are meant to influence decision-making.
  • Can you give an example of pre-packaged decision-making solutions?
    Loan approval systems, healthcare claims fraud detection, next best action and credit card approval are all examples of pre-packaged decision-making solutions that use predictive analytics and are available in the cloud.
  • When it comes to reducing costs should we also consider the cost associates with skill sets and use prepackage solutions to  therefore reduce costs?
    Yes. Clearly there is less demand for expensive skill sets when using pre-packaged solutions. Given the importance of cost reduction as a driver for predictive analytics in the cloud adoption this is highly relevant.
  • What percentage of customers are storing source data (for analytics) in the (public) cloud today?
    Public cloud did not get much love in the survey so I would think this is still pretty low. Private and vendor managed clouds are much more likely to be where data is managed.
  • Is there any hard evidence that cloud-stored data is at any greater risk from intrusions than on-premise data?
    Most breaches are not of cloud-based solutions but on-premise databases. Someone pointed out that Adobe had a major breach of their cloud services. While this is true I think this too was an on-premise server being managed by Adobe that happened to contain information ABOUT cloud users so I stand by my comment: It is not obvious that cloud-based solutions are more vulnerable than on-premise ones.
  • Does the idea of cloud analytics with locally stored data become impractical? Given security restrictions, that might be attractive on the Public Sector space
    The need to keep control of data for security reasons is a big driver for the importance of private clouds in the infrastructure being adopted. A private cloud might help minimize security concerns while still providing some benefits and being perhaps easier to integrate with external clouds.

Don’t forget you can watch the original webinar here and read the report here for more context on these questions.

Sorry for forgetting to post this until now!

Share

Previous post:

Next post: