Robin Way will be going deeper into the discussion of model deployment during his session at Analytics 2013 in Orlando next week, but first he shares his insight as a gust blogger on The SAS Training Post.
Big Data seems to be all the rage these days, and businesses everywhere are hopping on board the analytics train looking for improved performance and a competitive edge. What you may not realize is that many companies are falling short in their big data efforts, and are failing to realize the full economic potential of their analytic investments.
Far from an attempt to make analytics sound dramatic and theatrical, the title of this post is a testament to the importance of a single step in the model process that often gets overlooked (and underutilized) – Model Deployment.
Building analytic models for the sake of identifying analytics-driven business insight is a worthwhile exercise, but the moment of truth occurs when the resulting models are used to drive customers’ behavior change. In this blog post, I spend some time dissecting the glaring gap in model management, provide a framework for moving from model development to execution, and provide a taste of my Analytics2013 presentation and free Redpaper.
What’s Missing?
Companies of all sizes across all industries are failing to realize the full financial benefits of their analytic model assets due to lack of adoption (largely the result of failing to deploy them properly). Robust models involve multiple constituencies in analytics, IT, information security, and the customer-facing field – constituents that work in siloes with their own processes, frameworks, and often even their own language for tackling the same projects. It’s no wonder that efficiency and accuracy suffer at every turn.
What’s missing are the common language and processes for understanding and implementing analytic models that businesses can share across constituencies – the infrastructure to build a comprehensive strategy between model development and execution.
While some common processes (e.g. SEMMA, CRISP-DM) now exist in the practice of model development, few common processes exist in the world of model deployment. This gap leads to long project cycles, increased risk, lower quality, and reduced adoption of model-based insights.
What Will it Take?
To succeed, businesses need consistent model conventions and processes with buy-in from all constituent parties. These common understandings will facilitate essential information sharing about how a model works, who built it and approved it, which decisions it informs, and who in the field should use it. Firms must also manage their model portfolio as it ages so they are calibrated, enhanced, and replaced as their value decays.
At Corios, we’ve put together a set of best practices for analytic model deployment, organized into five phases, which we’ve nicknamed the “Five D’s.”
What it Means for You…
What the Five D’s shows is that there exists a great deal of work that must occur in order for you to make the most from your model assets – work that must happen in addition to model development and well beyond the jurisdiction of your model development team.
Establishing a model deployment framework with the Five D’s will facilitate the common language and processes that are integral to proper model development, management, execution and deployment. A business-wide approach that secures adoption from every set of constituencies in analytics, IT, information security, and the customer-facing field sets the foundation for optimizing model performance and increasing your analytics ROI.
How to Learn More
If I’ve managed to grab your attention, I would be pleased to meet you at my upcoming Analytics2013 presentation on October 21st in Orlando. In the presentation and the corresponding RedPaper, I’ll dive into a deeper discussion about the Five D’s of Model Deployment, drawing on real case studies to provide practical lessons from businesses who’ve successfully translated their analytic models into the field, supported their governance, and monitored their performance.
Click here to download the free RedPaper now.
