Companies are clearly sold on analytics, but not so clear about what it means or what it takes to operationalize them.
The most prolific data collection and model building efforts will only go so far towards delivering valuable business insights at scale.
The vast majority of companies embracing analytics are spending significant time and capital to build analytics models that never deliver on their intended promise or expected value. SAS research found that less than half of the best models get deployed while 90% of models take more than three months to deploy.
As it turns out, model development and data collection might be the less onerous part of the equation. It’s the last mile, the exercise of incorporating analytics into enterprise decision making processes at scale, that’s proving to be the most difficult. Leveraging analytics models for the thousands or even millions of decisions being made each day requires an approach that operationalizes the process so companies can reap the benefits of true data-driven decision making.
“Organizations are investing hundreds of millions of dollars in analytics, data, talent, and tools, but they are running into roadblocks demonstrating the business value of their investments—that’s the number one challenge,” notes Sarah Gates, an analytics platform strategist for SAS. “The only way to do that is by operationalizing the analytics deployed into production.”
Technical Debt is a Speed Bump
There is no easy answer for operationalizing analytics, but there are some common areas where processes can be overhauled to ensure success. Legacy infrastructure, data silos, and manual processes remain the biggest obstacles to successful analytics deployment at scale.
Without a structured process to coordinate resources across IT and the business, organizations find themselves grappling with an uptick in the number of either undeployed or delayed analytics efforts. Even when models make it into production, things can grind to a halt due to the number of handoffs moving models through the analytics lifecycle. Without model management processes, organizations have no way to monitor analytics in production to ensure performance is adequate and that value is sustained as business objectives shift.
“Manual one-off processes and the technical debt accumulated over time drags an organization down,” Gates explains. “It takes so much effort to keep the models performing, organizations don’t have the capacity to scale.”
Experts suggest the following to jumpstart the process of operationalizing analytics:
- Identify key business objectives and sync them to data and analytics efforts throughout the entire lifecycle, from data collection and management to model operationalization and decisioning.
- Promote cross-functional collaboration between IT and the business.
- Adopt rigorous data and model governance.
- Embrace ModelOps practices for developing, testing, deploying, and monitoring models at scale.
- Enlist partners with deep functional experience in analytics and data governance coupled with rich industry-specific domain expertise.
- Build out a bench of talent that goes beyond data scientists. Recruit and develop a team with skills in areas like IT operations, APIs, data integration, machine learning, and open source technologies.
Like any major enterprise initiative, operationalizing analytics requires a significant amount of change management. Yet without a serious effort made, organizations will find their investments in data and analytics falling well shy of expectations and critical business value.
Click here for more information on how SAS can help you operationalize analytics.