What is ModelOps?
The key to faster returns on your analytic investments.
Still struggling to demonstrate value from years of heavy investing in analytics? You’re not alone. On average, only half the analytic models built ever make it into production. That means half of the people, data and analytic technologies involved go to waste. SAS can help you change that.
Deployment – Analytics' Last Mile
The deployment phase of the analytics life cycle – the “last mile” – is where most problems and delays occur. The answer? A well-defined structured approach.
ModelOps: The Way Forward
ModelOps is a holistic approach for rapidly and iteratively moving models through the analytics life cycle so they are deployed faster and deliver expected business value. ModelOps is based on the application development community's DevOps approach. But where DevOps focuses on application development, ModelOps focuses on getting models from the lab through validation, testing and deployment phases as quickly as possible, while ensuring quality results. It also focuses on ongoing monitoring and retraining of models to ensure peak performance.
Crossing the Last Mile of Deployment
To help you cross the "last mile" of deployment much faster, and ensure that your analytic models deliver expected value, ModelOps defines people (or culture), process and technology changes that facilitate smooth, efficient and continuous development and deployment of high-impact analytic models.
Foster dynamic collaboration and improved productivity for analytics and IT operations teams, regardless of what analytic language is used, what data is accessed or where the model will be deployed.
Unify information about models, including metadata, where models are deployed, and performance and accuracy data.
Accelerate how work gets done at every step so analytic models get out of the lab and into everyday business use faster.
Getting Started With ModelOps
Adopt a new level of cross-departmental collaboration with the shared objective of creating analytic models that deliver expected business value. Everyone – including data scientists and IT developers – should have same processes, model expectations and rules. This means uniting disparate teams that scope, build and deploy analytics. Over time, add model-specific metrics to track performance.
A Centralized Repository
Collaborative, cross-functional teams capture and store model information and related metadata in a single location. A centralized repository facilitates the scaling of analytic models by providing a complete model inventory – including their functions, inputs, champion/challenger models, versions and who last worked on them. It also facilitates the measuring and monitoring of models in production. And it retains intellectual property, mitigating the risk from staff turnover.
Automation & Standardization
Every manual task that’s automated and streamlined reduces complexity, technical debt and points of failure. For ModelOps to deliver expected improvements, automation and standardization must include performance monitoring, alerting, and model deployment and retraining.
Enable Model Governance
Gain a holistic view of all your models so you'll know how they are performing and what is being done to them – without a lot of manual effort.
Compare models side-by-side to evaluate champion and challenger models. Make comparisons between models, and select the best model every time, regardless of the language used to create it.
Ensure the champion model is running optimally by understanding how often it is scoring data and what version is running. Automate alerting to know when performance benchmarks are no longer met, so you can easily retrain, revise or retire the model, as appropriate.
Track the history and lineage of each model in your organization – including model details, as well as what data is being used and where it is running across the business.
Increase ROI With ModelOps
ModelOps removes a key friction point in the analytics life cycle, which helps ensure that analytic investments deliver business value faster. A ModelOps approach gets analytics out of the lab and into use, enabling you to conquer the last mile of analytics.
Preserve data lineage and track-back information for governance and audit compliance.
Deploy models in minutes, not months, with close collaboration between data scientists and IT.
Deploy models with a monitoring mindset so analysts can monitor and retrain models as they degrade.
How SAS Enables ModelOps
Overcome your challenges related to operationalizing analytics with a solution that combines SAS Model Manager software with dedicated consulting services to jumpstart your implementation and get ongoing ModelOps guidance – regardless of the language your models are written in.
ModelOps Health Check Assessment Service
Discover what it will take to successfully operationalize analytics in your environment. Our assessment service will help you understand and define your intentions and readiness, with recommendations for consistently deriving value from your SAS and open source models.