SAS is on the brink of…something.
Never underestimate the power of analyst relations. Each year, SAS invites industry analysts to a posh resort and shows them shiny things that most SAS users will never see. Steamboat Springs used to be the destination; this year, the Ritz-Carlton in Naples, Florida got the nod. Naples is so much nicer in February. The broiled branzino served in the Terrazza, though pricey, is tasty and fresh.
As night follows day, in the weeks that follow the event, analysts publish their reports about SAS. This year, Tony Baer writes that SAS “is on the brink of generation change.”
With the accession of Dr. Oliver Schabenberger from chief of research, product development, and most recently, CTO, to the post of COO (reporting only to Dr. Goodnight), SAS has completed a changing of the guard that began three years ago.
In other words, with Jim Davis’ departure.
With Dr. Schabenberger, SAS has promoted a product researcher with deep academic background (like Dr. Goodnight) to head product development, sales, and operations.
Readers ask me what I think about Schabenberger. I have no opinion. He may be the next Andy Grove, for all I know. I would simply note that everyone thought Jim Davis and Carl Farrell were heirs apparent until they weren’t.
SAS characterizes the newly created COO slot as a plan to “expand” the management team.
That’s nice. But SAS should consider focusing on quality rather than quantity. A management team consisting exclusively of SAS “lifers” lacks the diversity and depth to make smart moves.
In any case, executive succession is less important at this point than ownership succession. Goodnight and his wife own 80% of the company. Unless he has secretly donated his equity to a charitable trust, a sale of the company seems almost inevitable.
The challenge for SAS is that a market where it had unquestioned dominance for the lion’s share of the last 40 years is now getting quite crowded.
With (the) emergence of Viya over the past couple years, SAS smartly concluded that its value add was not in the SAS programming language…With Viya, SAS chose to coexist with the open source environment. Develop in R or Python, use a Jupyter notebook to compose and share your models, but at runtime, send them to Viya where the models run in SAS’s Cloud Analytic Services (CAS) in-memory engine.
Wait. Stop. This is just plain wrong. Viya has client APIs for Java, Lua, Python, and R. This does not mean that you can run Java, Lua, Python, and R code inside Viya. It means that you can invoke SAS code from those languages.
That’s an important difference. Viya does not run Python; it runs SAS. You can connect to it from Python, run SAS commands, and use the Viya result set in a Python program. That’s fine. But if you want to write code that runs in Viya, you’re going to write SAS code.
Pro tip for analysts: read the manual.
SAS has a firm hold on its Global 2000 enterprise base, which isn’t going away anytime soon.
I wouldn’t be so sanguine. In fact, the most consistent message one hears from G2000 CDOs is “we want to reduce our SAS footprint.”
- A top US bank has a project underway to convert its risk models from SAS to Python and Spark.
- A leading US insurer (a “charter” SAS customer dating back to 1976) has migrated more than a hundred SAS users to open data science tools.
- A multinational health information company (and SAS “reference” customer) expects to have a thousand analysts working with open source tools — most of them converts from SAS.
- A trillion-dollar asset management company plans to convert most of its SAS user base to open data science tools
- In the UK, a large government agency has reduced its SAS user headcount by half, all displaced by open data science tools.
The CDO of a Canadian bank wrings his hands in frustration over what he describes as “a $20 million SAS bill.” Right now, he thinks he’s stuck. How loyal will he be once he sees a way to reduce that number?
Two-thirds of recent growth in spending on machine learning software went to new vendors, according to IDC. Innovation drives the machine learning market. Vendor loyalty is dead.
But it is challenged to compete for the hearts and minds of the next generation of data scientists, data engineers, and business analysts who have become empowered by self-service and drawn in by open source.
To say the least. New grads entering the workforce universally favor Python and R over SAS. Organizations that still use SAS send new data scientists off to training, or set up a parallel open data science infrastructure.
This is one reason one shouldn’t count on SAS to hold on to those Global 2000 customers. Surveys show that SAS users skew significantly older than R or Python users. Many large organizations expect SAS headcount to decline, while R and Python headcount grows. Some are more aggressive about pushing for change, but for the rest, it’s only a matter of time.
There are ways for SAS to get cool again. We recall several years back, flying to a conference (actually, it was SAS), where we found ourselves sitting next to a high school senior who was on his way to UT Austin to check out their data science program. Self-taught, he was already proficient in R and Python, but preferred SAS because of the full-featured analytics environment. That’s the scenario that SAS wants to replicate.
Self-taught high school seniors proficient in Python who prefer SAS are as common as twentysomething women of color living on the Upper West Side who prefer Trump.
As for open source, as mentioned above, SAS interoperates with it, mostly through Viya. However, dealing the lack of perception about SAS and ML, SAS should start contributing to open source.
What, exactly, can SAS contribute to open source? To do so, a vendor has to either (a) distribute open source software, or (b) build something in open source software. SAS does neither. The company would have to donate the SAS Programming Language to open source or rebuild its end-user products on TensorFlow or Spark.
Hell will feature ice skating before either one of those things happens. Open sourcing the SAS Programming Language would gut 80% of SAS’ revenue. The dairy farm and rock collection would definitely have to go.
Viya is SAS’ fifth attempt to deliver a scalable platform. It’s possible that SAS has finally nailed it; I know of at least one customer that uses the software, and they seem happy with it.
But customers don’t seem to be voting with their pocketbooks. SAS launched Viya in 2016, and it was generally available for all of 2017. In a market growing at double digits, SAS barely managed 1% revenue growth in 2017. If customers bought Viya, they pulled the plug on legacy SAS rather than expanding the footprint.
Even if SAS can sell Viya as a departmental application, it’s not the enterprise machine learning platform of the future. That will be open source or hybrid open source. Commercial software will sit on top of the stack supporting the last mile to the end user with accessible user interfaces and business solutions. Think DataRobot on top of Python, R, XGBoost, and H2O; StreamSets on top of Spark, Kudu, Solr, and Drill; or Dataiku and Kogentix on top of Hadoop.
SAS has conceived an architecture that is upside down, with a commercial platform on the bottom and open source tooling for end users on the top. Few large organizations will adopt that architecture. It’s too expensive, inflexible, and subject to vendor lock-in.
But in the spa at the Naples Ritz-Carlton, you can purchase a Drift to Sleep massage for $245. And at $60, the dry-aged rib-eye available in The Grill is a bargain.