Wednesday, February 29, 2012

SAS Unveils High Performance Analytics Technology

I spent the early part of this week at SAS’s annual analysts conference, where the company reviewed the past year and presented its vision for 2012. The story this year was simple: “big data”, and SAS’s “high performance analytics” approach to taming it.

Of course, “high performance analytics” is what SAS has always done and, like “big data” itself, the term is relative. What SAS specifically presented was a re-engineering of its core analytical procedures to run in “shared nothing” multi-processor environments.  Each data set is split into pieces that are loaded into separate units, processed independently and simultaneously, and then brought together for a result. SAS cited tremendous performance improvements, such as reducing the time to build a loan default model on a billion rows of records from 11 hours to 50 seconds. This obviously makes possible new, tactical applications.

The high performance architecture is becoming available in stages as each SAS procedure is rewritten to support it.  The change from a customer perspective is purposely minimal: existing SAS procedure calls are simply modified by adding a "HP" prefix.  This will make it easy for clients to take advantage of the new capabilities. 

The company revealed a just a few new products at the conference, most notably a Visual Analytics tool that uses in-memory processing to render billion-row data sets in seconds. But the real benefit of high performance will come less from new products than from using it with existing SAS procedures and tools. The SAS product that may benefit most of all is SAS Decision Management, which creates rule-based decision flows that can call on scoring models and other analytics to help guide tactical processes. The product itself isn’t new, but high-performance analytics will let it do new things.

SAS’s “big data” story also included Hadoop integration and expanded cloud deployments. By the end of March (if I understood the roadmap correctly), SAS will be able to read from and write to Hadoop data sets, embed Hadoop commands within SAS scripts, and send SAS metadata to Hadoop. Over the coming year, it will support cloud deployments through a variety of enhancements related to virtualization, open APIs, and eventually an app marketplace. The cloud-based initiatives also support SAS’s own on-demand business, which grew 57% last year to reach more than $100 million.

These are all positive developments for SAS, which must certainly support "big data" to remain relevant.  The new capabilities will also create some business changes as SAS competes more directly with companies like IBM and Oracle to embed analytics within operational processes. SAS itself noted the company is now more involved in architectural discussions of how its systems interact with the rest of the enterprise infrastructure. Other issues may include educating non-technical users and providing technology to protect privacy.  SAS leaders seem to think they can leave those issues to others, but I’m not so sure.

The conference produced little news directly related to marketing systems.  The company reports 38% growth in marketing applications – which it reports under the label of “customer intelligence” – so that is clearly a healthy business.  But the product road maps showed just incremental improvements of existing products, without any major new offerings. Again, high-performance analytics will make new things possible without other changes in the products themselves. The high performance version of marketing optimization is due by the end of the year.

If you want more evidence of how little attention was paid to marketing systems: SAS's biggest recent piece of marketing-related news, last week’s acquisition of online ad server aiMatch, got exactly one mention during the day-long presentation and was positioned as simply filling a small gap in the marketing product line. The company did announce, very casually, that aiMatch would be extended to include ad buying optimization as well as its current ad-selling optimization. That struck me as a pretty big deal, since ad buying is the heart of an already-huge industry that’s clearly the future of marketing. Then again, it’s also an intensely competitive, heavily-funded space that’s crawling with advanced technologies. Although SAS's high performance analytics could have a huge impact on ad serving, that won't happen unless SAS makes a major commitment of people and money.  We’ll see whether they make one.

No comments: