A recent report by McKinsey noted that 40,000 exabytes of data will be collected worldwide in 2020. If just five exabytes is equal to all the words ever spoken by mankind, it’s not an understatement to suggest there is a prodigious amount of material for a data scientist to process to glean insights.
To tackle this great mass of data with the hope of finding useful information for an organization, IT leaders need to use and build analytic models.
Sending an analytic model into deployment can be a challenge when the task is relatively simple. When the challenge is complex, the model deployment naturally becomes more involved and considerably more expensive.
Give businesses full marks for trying. Worldwide organizations are investing heavily in analytics and AI — almost $200 billion in 2019. IDC predicts that by 2022 they’ll be spending more than a trillion dollars. While the numbers are impressive, it’s becoming clear that many companies need to reassess where they have been putting their money. New research shows half of all analytics models developed are never deployed.
“Many organizations do not have a repeatable process for operationalizing analytics that quickly moves the actionable insight into production and keeps it current.” said Steve Holder, Head of Strategy and Innovation at SAS Canada. “There is little doubt the value analytics brings to an organization, but what’s frustrating stakeholders are the delays and bottlenecks that exist, which result in a lack of timely and accurate decision-making.
Why are organizations getting such a poor return on their investment? Why do they carry on when the failure rate is so high? Most importantly: How can businesses turn the corner and stop throwing their money (and effort and time) out the window?
The central challenge for organizations, and specifically their data scientists and IT groups, is the all-important “last mile,” where half of analytical models are failing to be deployed. The factors behind the high rate of failure include: bad data; poor governance; drudgerous and time-consuming manual processes; and, perhaps most critically, a lack of collaboration between data sciences and IT.
“The last mile is where hard work and intentions should meet actual results,” adds Holder. “Regardless of industry, the struggle is the same: to put analytics into action.”
Many organizations talk a good game about digital transformation, saying they have adopted a data-driven culture. But adopting data as a driver is not enough. Organizations have to be able to turn data into action — and ultimately an advantage over competitors. Companies are at the “last mile” of their journey when they have done everything except implement, or put their analytics to work. This is where solutions like SAS® ModelOps come in handy.
ModelOps, a new packaged offering combining SAS Model Manager software and advisory services, streamlines the management, deployment, monitoring, retraining, and governance of both SAS and open source analytical models. ModelOps gives companies the tools they need to get their analytics out of the lab and into practical use. Businesses can conquer that last mile of analytics through:
- Data – aligning trusted data to privacy and security standards
- Model creation – creating models with a deployment scenario in mind to avoid rework
- Governance – preserving data lineage and track-back information for audit and governance compliance
- Deployment speed – completing deployments in minutes as opposed to months, with close collaboration between data scientists and IT
- Model monitoring – deploying models with a monitoring mindset so analysts can monitor and retrain models as they degrade
ModelOps also offers tailored consulting services as well as a new standalone service, ModelOps Health Check Assessment, which was designed to help organizations understand how to optimize deployment.
Learn more about how SAS can help organizations conquer the last mile of analytics, and register to attend the Unleash Analytics Event in Toronto on November 6th.
