Treating analytic discovery and model development as a production environment


IT support levels for different processes and business applications depend on many factors. So how are analytics projects generally supported?

It's common for IT to view the implementation of advanced analytics as part of a production job or process, especially when analytic models are deployed into different operational systems that are viewed as production level systems.  However, it is common for IT to view the analytic discovery and model building process as a non-production or a development process.

Since the importance of using advanced analytics for strategic business purposes continues to be embraced by all types of industries, treating the analytic development process and related jobs as non-production is a mistake.

It might make sense to treat operational reports as development projects, because these generally deliver information about what has happened in the past.   Analytics, on the other hand, provides insight into what will happen and even when it will most likely occur.  As a result, interactive analytic discovery and model building should not be treated as a development process in the historical way IT views development activities.

Analytical models degrade over time and must be monitored and updated in order for them to remain effective and deliver meaningful business value.  As a result, the analytic discovery and model building process should be viewed as part of the production processes. Consider a model that predicts when a multi-million dollar piece of equipment, like an electrical submersible pump, will stop working. Or what about a model that helps detect electrical theft or predict credit card fraud. The longer it takes you to update or replace this type of model with an updated (better) version, the more money, time and resources you lose.

Your interactive analytic discovery and model building environment is just as valuable, if not more valuable, than the implementation of the models that run in other production level systems.   This means IT should provide the same high-available, disaster recovery processes for analytic development areas instead of viewing and treating them with the support levels associated with development environments.

What might this switch entail? It will probably mean a higher, short-term cost associated with setting up your analytic IT environment. However, the long-term benefits of having a system that is continuously updating models in production will more than make up for any short-term investment costs.


About Author

David Pope

Technical Leader, Senior Manager US Energy

David leads the pre-sales technical team for SAS US Energy which solves business problems in the Oil & Gas and Utilities industries using advanced analytics. He is a lifetime learner who enjoys sharing information and helping others to grow their careers. He earned a BS in Industry Engineering and a Computer Programming Certificate from North Carolina State University. Furthermore, he has over 29 years of business experience working with SAS across R&D, IT, Sales and Marketing in the Americas and Europe. He is an expert in working with data and producing insights through the use of analytics. David has presented at SAS Global Forum, the 2012 SAS Government Leadership Summit, IBM’s Information on Demand(IOD), EMC World, CTO Summit Conferences, is the author of the book: "Big Data Analytics with SAS", and he currently holds 14 patents for SAS in several countries: US, CA, Norway, UK, China, Mexico, and Hong Kong.


  1. Robby Powell

    Hey David Your post did an great job of assembling some thoughts that have been rattling around in my head. I have been studying DevOps, which focuses on Development and IT Ops collaborating and working together transparently to create and deliver software. From that research, I have been toying with applying the same principles to Development and Analytics - "DevAlytics" - yeah I just made up that word. I recently posted
    "What's with all the fuss about DevOps?" and am interested to hear what parallels you see. Thanks, Robby

    • David Pope

      I see quite a lot of parallels between DevOps and IT support of analytics (or business). One key is getting executive support that analytics are a strategic value to the future of the company and to get the business and IT teams to work together. Which actually ties nicely into a recent blog written by Aiman Zeid on Only leaders can transform organizations.

Leave A Reply

Back to Top