What’s happened over the past 18 months?
SAS Viya on Azure is still a fraction of SAS’s core business, but growth is starting to pick up. For calendar Q4 2021, Viya on Azure sales were up 30% over a year ago as the company closed 20 seven-figure deals, one of them topping out at $6 million for a three-year legacy SAS 9.4 migration to Viya. Over the past year, the largest was a similar $15 million multi-year migration engagement. SAS has been building integrations with Viya services to fit in natively with Azure analytics, streaming, and machine learning services. Obviously, with such an ambitious game plan, it’s currently a work in progress. SAS currently offers Viya plus a dozen classic services on Azure Marketplace. Let’s drill down on the solutions and where SAS is in working Viya into the Azure fabric. In general, you can access SAS models and model management from various Azure analytics services and vice versa, although not surprisingly, the degree of integration currently varies by service. It starts with Viya integration into core Azure infrastructure, including Azure Active Directory for single sign-on; Azure Key Vault for certificate management; Azure monitor for logging; and Azure Policy for governance. Integration with Azure Kubernetes Service (AKS) allows Viya services to run on Azure K8s clusters, so that, for instance, SAS Event Stream Processing (ESP) can elastically scale without having to reinvent the wheel. Other basics include the ability for developers to version code from SAS Studio in GitHub. And at the database level, SAS Viya has connectors to all Azure databases; this is not exceptional because, like any other analytics suite, Viya has connectors to other popular databases such as Oracle. One of the best developed integrations is with chatbots; you can build your chatbot in SAS Conversation Designer and run it in Teams. SAS models can now be registered to run in Azure Machine Learning, while Azure Synapse Analytics can run SAS scoring models in-database. For anyone who follows SAS, there’s a small irony here. Running the models in Synapse requires running Spark, a switch from the LASR engine that was originally SAS’s preferred high-performance analytics compute engine in the days before Viya. You can register models developed in Azure Machine Learning (Azure ML) in SAS Model Manager for lifecycle management, and in turn register and deploy SAS models as containers in Azure ML. There are other connections in the works. They include running SAS Intelligent Decisioning (managing decisions and tracking their effectiveness) with Microsoft Power Apps and Dynamics 365 with connectors that are now generally available. In the long run, the payoff of SAS integrations on Azure will be gated by the breadth of the Viya platform. Not surprisingly, the development of Viya has followed the 80/20 path, with the crux of SAS capabilities in statistics, risk management, decisioning, and so on covered. But SAS has a rich trove of solutions in domains such as fraud and anti-money laundering; supply chain; risk management; optimization, and so on. There is also SAS’s broad portfolio of vertical solutions, and on Azure, SAS is focusing on four key verticals including banking, healthcare, manufacturing, and retail.
Why Azure?
SAS’s strategy is a gamble that departs from the usual script of third-party technology providers, especially for those whose data and analytics platforms and solutions compete with counterpart offerings from each of the Big Three. So why go all in on Azure? In part, it’s to provide a highly functional bridge to the cloud for its current installed base. But there’s another reason, we suspect, for getting close with Azure: It’s the shortest path to appealing to a new generation of customers. This is a generation that got inspired over the past decade by Tom Davenport’s extolling of data science as the sexiest job of the 21st century. Over 40 years, SAS may have built an analytics portfolio practically unrivaled in depth and breadth, but that means relatively little to grads coming out of data science university programs accustomed to working with open source languages, technologies, notebooks, and AutoML cloud services. Even though SAS has made peace, and works with all that open source stuff, that may be a secret to up and coming data scientists that’s been kept all too well. With a focused relationship with Microsoft, SAS could enrich Azure analytics with assets such as model life cycle management; functional depth in disciplines such as decisioning, risk management, and optimization; and of course, its deep portfolio of modeling libraries. Some of this is here today; we would like to see SAS bring more of the richness and breadth of its capabilities into Viya, and in effect, hide in plain sight as they operate as extensions to Azure Synapse or Machine Learning. We would proffer that Azure provides the onramp to a new customer base to take advantage of SAS capabilities without having to make big enterprise software commitments. So, here’s our suggestion to SAS: clear up your Viya on Azure branding because it’s downright confusing. Right now, you call it SAS Viya on SAS Cloud, implying that customers have to run on a separate cloud. Why not borrow a page from Databricks, which currently enjoys an obscene valuation, and call it something like Azure Viya?
