Things may not appear to move that quickly at 41-year-old SAS Institute, with its myriad analytics, BI and other products. But that might be changing as it ports its portfolio over to its Viya platform, but perhaps more significantly because it is injecting machine learning into most of its tools and applications. That said, it's something the company has been doing for some time.

At its recent European analyst summit, it took quite a bit of time to explain its progress with machine learning.

The 451 Take

SAS has many smart people who have been doing machine learning for a long time – decades, in fact – and as such, have an entrenched advantage over rivals that have more recently discovered that machine learning will in fact change everything in the software industry. However, entrenched positions are subject to erosion and SAS must battle against some of its older ideas that now seem a bit outdated; i.e., the near-religiosity of its anti-open-source message, which seems jarring to us, although we understand why it takes this stance.

Context

SAS has begun the process of reorganizing its teams by vertical industry in Europe, starting with the Nordic region in 2016 and then will do the same in the DACH (German-speaking countries) this year. The US has been organized like this for some time.

The government market grew particularly strongly in 2016 (although at 15%, contributed the same proportion of all revenues as it did in 2015), which SAS put down in part to desire for government departments to invest in technology where its application can help fund themselves. This was also one of the main drivers behind a 36% jump in fraud – or preventing it – as a growth area too. SAS says government tax departments are using its software to track down non-payers and thus recoup lost tax revenues. Fraud was also a strong area for key partner Accenture. Risk management was also strong for SAS, particularly in Europe.

CEO Jim Goodnight is 75, but there's no hint of retirement or mention of succession planning, although we're sure the latter is happening behind the scenes – at least we hope it is because nobody can go on forever. The company currently has about 14,000 employees and is regularly high on the lists of best places to work in the US and elsewhere. An earlier report on SAS documented its 2016 finances.

Products

At its recent European analyst summit, SAS highlighted three high-level areas of product investment: platform, openness – which it markedly distinguished from open source – and public APIs.

The platform investment is centered on Viya, the in-memory cloud-ready environment for data scientists, business analysts, application developers and executives, as SAS pitches it, which will eventually take over from SAS's 9.x branch as the platform on which all SAS offerings ultimately run. However, at the moment and for the next couple of years, the company stresses that this is a 9.x and Viya strategy, not a 9.x or Viya one. There was a major release of Viya (3.2) in March 2017 with 3.3 scheduled for Q4 2017. Gradually, all SAS products are being ported to Viya, with the latest major offerings being Visual Analytics 8.1 (8.2 comes with Viya 3.3 in the second half of 2017), Visual Investigator 10.2 (10.3 also in the second half of 2017) and Visual Forecasting, a new product. The 'Visual' theme is a big one for SAS because it has focused a lot more on the user experience in the last couple of years than it did in the past.

Openness and public APIs seem to us like the same thing and are SAS's answer to those that accuse of it being a closed and proprietary platform. But SAS has REST, Python, Java, Lua and R APIs for those wanting to program Viya's Cloud Analytic Services (CAS), which it describes as the centerpiece of the SAS Viya framework. So developers and data scientists don't have to know SAS to use Viya. And while some of those languages may have disadvantages versus SAS in terms of pure performance, those are countered to various extents by the size of developer communities and thus availability of skills. SAS has invested in university programs to have people graduate with SAS skills and continues to do so.

Machine learning

SAS has many smart people who have been doing machine learning for a long time – decades, in fact. But of course, the machine learning they were doing back in the 1980s and '90s is what might be called classic machine learning; the algorithms and models that drove many text analytics products over the past 20 or so years (which we started covering in earnest in 2001). Those sorts of algorithms required smaller data volumes than today's Deep Learning algorithms feed on. But they are also less flexible, in that they work well when future data resembles past data. With Deep Learning's multilayered neural networks, the role of the data is to program the model – the model hasn't been pre-chosen, as it were. So the model fits the data, not the other way around. SAS says Deep Learning is a huge area of investment for the company and its first Deep Learning programming toolkit will ship with Viya in the fall.

In its messaging, SAS refers to machine learning, Deep Learning, artificial intelligence and cognitive computing as distinct things. In this era of AI hype, we think it would be advisable to narrow that list down and drop cognitive computing, partly because it was IBM's term and partly because we feel it doesn't really add anything. Just for the record, when SAS talks about cognitive computing, it is currently talking about machine-learning driven voice interfaces. At the recent European analyst summit, Goodnight demonstrated an Amazon Echo device running queries via his voice commands to Visual Analytics running on Viya, which mostly worked – never easy in a packed room with background noise.

One area where the appetite for machine learning has been strongest is fraud and security intelligence. Here, the company reports customers are much more informed now than they were even a year ago. Some are asking for things like prescriptive analytics and neural networks. Customers are realizing, says SAS, that in the constantly evolving threat landscape that fraud represents, retuning models constantly to try to keep up is a losing proposition. So it's better to hand control over to the software and let it retune in near-real time.

Another area is customer intelligence. Here, SAS sees use cases such as multivariate testing, but taking a full factorial approach to optimize such tests compared with simpler serial A/B testing. This optimized approach results in a reduced number of variants and volumes of traffic, according to SAS. Other use cases include contextual marketing, next best action segment discovery and customer journey optimization.

A key part of the upsurge in interest in machine learning has been due to the somewhat accidental discovery a while ago that graphics processing units (GPUs) are great for accelerating machine learning, and especially the parallelization of Deep Learning multilayered neural networks. SAS's risk products can already take advantage of GPUs and specifically works with Cuda, the language that runs on NVIDIA GPUs, although there is no official partnership between the two companies as yet. Not all SAS products are suitable for GPUs, which has to do with the single versus double precision nature of some GPUs and CPUs and how that affects the parallelization of code. SAS says it is not currently looking at field-programmable gate arrays (FPGAs) as an alternative to GPUs.

Competition

SAS's competition is obviously broad and depends on the product area. In the customer intelligence and marketing technology space, SAS competes most directly with the vendors of broad-based cloud suites: SAP, IBM, Salesforce, Oracle and Adobe. In the fraud and security intelligence space it would be companies like IBM, Microsoft and Oracle but also more focused ones like Palantir Technologies and StatPro.

In the modeling, analytics and machine learning, it sees IBM, SAP and Oracle most often. Smaller vendors up against the analytics products include Tableau, Qlik, TIBCO Spotfire and Domo.

In terms of machine learning, IBM Watson and Microsoft's Machine Learning are potential competitors as well, along with too many startups to mention here exhaustively, but some names might include Alteryx, BigML, H2O.ai, RapidMiner, Ripjar, Skytree and Yottamine Analytics. Microsoft, Google and AWS would be other machine-learning platforms developers would be considering.

SWOT Analysis

Strengths

SAS has deep knowledge of machine-learning-powered analytics and a massive customer base to target with machine-learning-driven upgrades, before it ever will have to look for new customers.

Weaknesses

The anti-open-source message might come back to haunt the company one day. It eventually acknowledged the presence of the R language long after it became popular. Open source is a key element of machine learning – and we can't believe SAS doesn't use some in its machine learning, but we don't know for sure. But the strength of its anti-open-source message seems jarring to us, though we understand why it takes this stance. In addition, the overlapping use of machine learning, Deep Learning, artificial intelligence and cognitive computing as distinct things could use some rationalization.

Opportunities

We believe the company will experience continued strong growth in areas that look to exploit large volumes of unstructured data, where investigators of all kinds trawl through large volumes of data looking for patterns and anomalies. In general, up- and cross-selling opportunities abound with its customer base.

Threats

Developers and data scientists coming up now that have not been exposed that much to SAS at university may be opting for tools, or at least platforms, from the likes of AWS and Google. The company may need to do more developer outreach.
Nick Patience
Research Vice President, Software

Nick Patience leads 451 Research’s coverage in two key areas: digital transformation and artificial intelligence/machine learning. Nick is a cofounder of 451 Research and Research Vice President, Software. He oversees the company’s coverage of the software industry spanning four research channels: Customer Experience and Commerce, Workforce Productivity and Compliance, Data Platforms and Analytics, and Development, DevOps and ITOps.

Patrick Daly
Senior Research Associate, Information Security

As a Senior Research Associate in 451 Research’s Information Security Channel, Patrick Daly covers emerging technologies in Internet of Things (IoT) security. His research focuses on different industrial disciplines of IoT security, including the protection of critical infrastructure, transportation and medical devices. In addition, Patrick’s coverage spans technological domains, including security for IoT devices, applications, platforms and networks.

Keith Dawson
Principal Analyst

Keith Dawson is a principal analyst in 451 Research's Customer Experience & Commerce practice, primarily covering marketing technology. Keith has been covering the intersection of communications and enterprise software for 25 years, mainly looking at how to influence and optimize the customer experience.

Want to read more? Request a trial now.