Is Viya the single biggest reason for SAS's downfall?
Agree by casting a thumbs up vote. Disagree by casting a thumbs down vote. Please...no pontification. Just let the numbers tell the story.
78 replies (most recent on top)
Many software companies find Agile problematic. Some SAS competitors use it; others don't. On this thread, I think you’ll find a consensus that, while Agile is not the cause of SAS’ troubles, it hasn’t helped.
https://www.thelayoff.com/t/1u122bGi
Ironically, in its early days, SAS listened to its customers quite well: that’s how it grew. Nowadays, SAS has Product Managers whose job is to listen to customers. Like Agile, this hasn’t helped.
If it's an industry-wide problem, how does it explain that many of SAS's competitors are doing just fine, and rolling out features that their customer want, whereas SAS is completely lost and dying a slow death?
There's nothing wrong with agile methodologies and good product managers don't need to be software engineers - they need to listen to their customers. Something that SAS has never been good at.
I know one good product manager who remains at SAS.
Sadly, most of the ones I knew had no training in software engineering, UX design, or anything related to the construction of the product. They could only define a list of desired features, and advocate for those.
Combined with Agile methodology, this tends to turn development into a “feature factory”. The development team, incentivized only to produce features, naturally tends to neglect everything else — user experience, system architecture, performance, and cost.
This results in mediocre products — even though they're managed to meet perceived customer requirements.
This is an industry-wide problem, occurring at SAS.
https://www.productplan.com/glossary/feature-factory/
wow this guy you talk about stipulated the 'Workbench'[1] and 'Visual Studio Code'[2] in 2019. If it took five years to push one's ideas through, I'd be frustrated to death! If Viya the analytics platform was built out as he imagined we might have a chance. Question is are there any good product managers left at SAS? The ones in my corner are more of pre-marketing and post-sales supports devoid of visions and guts.
[1] https://www.drmindle.com/analytics-platform/#focal-points
[2] https://www.drmindle.com/analytics-platform/#single-ide
Thanks so much for posting those links!
I read his SAS blogs, and rooted for him, and so much admired his courage.
It's no surprise he left.
Remember the guy who went gaga on internal blog? Is it you @7txa+1tPWcYil ? Anyhow I have been following his personal blog and his ideas on Causal Analytics[1] and Machine Learning Experiences[2] can definitely be SAS' new endeavors you're talking about.
In the old days, on the south east corner of fifth floor of bldg S, OS graced the corner office, two offices up north was this guy, across the hallway of this guy's office was JP's office when he was still a mere UI developer. How I wished in a bottom up engineers driven culture like Meta and Google, the ideation, execution and political skills among those three guys combined could've created something much much bigger than $3 billions. Yet the engineers all left SAS and the politician stayed... and the day when analytics dies... sigh...
[1] https://www.drmindle.com/ab-analytica/
[2] https://www.drmindle.com/ml-experience/
“reduce forces by at least 50%”
Also, in addition to those eliminations, more highly-qualified H1Bs and sending work to Asia would be progress.
@7txa+1tPWcYil very well said. SAS, are you listening? WAKE UP!
agree in theory. In practice, doesn’t seem likely for now. Maybe a prospective buyer would try.
SAS does not have to be “a leader” (initially) in anything to reimagine itself with a radically scaled back product line featuring substantially refreshed (e.g. Risk + Fraud) plus new verticals. What SAS needs is a sensible business direction in a handful of growing niche markets where they have a chance at succeeding.
On the practical side, SAS headquarters has a lot going for it. Short driving distance to three major universities which continue to turn out considerable technical, analytical and business talent, located in a highly desirable place to live, but still has reasonable CoL, etc.. this could be a draw for a lot of significant new talent.
The company does not have to be sold or even go public. Instead, the very brutal decision to reduce forces by at least 50% (this is guaranteed to happen as soon as all SAS sells, if it ever does), especially those who have been in management long-term is needed to free up the cash from the declining revenue stream to fund these new products — that don’t even have to be based on existing SAS infrastructure.
SAS this not have to be an innovation leader in ML/AI or anything else to do this. They just have to make savvy business decisions about realistic markets and hire the right product and engineering talent to implement. There are startups beholden to the VC game, who are doing this right now. there are initiatives within big tech companies doing this right now. Within a year, SAS could free up $1 billion annually to fund their new endeavors with a much financially healthier environment than VC. Offer equity incentives and higher the best!
So, keep enough of the existing best SMEs to support currently profitable SAS products. Ki-l the losers, including Viya if necessary. Be brutally honest about the existing technology that might have some chance of being brought forward. If not, leave it behind. Then pick a small number of niche markets to rebuild in. Use mostly an OSS model. Figure out how to leapfrog the existing incumbents there. Otherwise it’s a slow death, no matter who the owner is.
If he had the will, JG could initiate this in the next 6 to 9 months. I wager several of us who spent our lives building SAS would love to see it ongoing, yet realize it must initially make painful cuts and become a reimagined company for that to occur.
I can't believe this poll is TIED at 29-29.
Perhaps we need more pontification 😂.
“ When the business school case studies are written, they will fairly say that OS did not cause the decline; he accelerated it.”
Yeah ok
Last I checked JG, not OS is/was ultimately responsible for what SAS has built and become over the past nearly 50 years. Oliver, who did not rise to executive level until 2015, did what he could given circumstances over the past 20 years. For a singular, non-founder individual, his net positive contributions to SAS are geometrically lager than anyone else’s. Again, so easy to armchair quarterback now.
“SAS is where SAS is in spite of OS. He did not cause it but he did not help it either.”
By all accounts, early in his SAS career, OS did good work on Mixed Models. No one doubts his intelligence or work ethic.
But SAS revenues show that HPA, LASR, VA and CAS/Viya have not slowed the decline. I doubt any of these efforts are profitable.
That’s how OS damaged SAS: by opportunity cost.
Instead of those efforts he led, we could have built products that grew revenues and made profits. Other companies did — and ran away with what should have been our market.
When the business school case studies are written, they will fairly say that OS did not cause the decline; he accelerated it.
“serious product and engineering talent who could salvage some good from Viya and create a platform for building successful industry verticals based on lead edge analytics/ML/AI — including a heaping helping of OSS were appropriate.”
You’re right; that plan could turn the company around. Sadly, there’s nothing to suggest it will happen.
"People who use the term spot on are to be ignored.
His tenure at post SAS gig could also have been short because he ate too much rhubarb pie. Since we are throwing sh-t out there I thought ai would pull my weight."
People who use the term 'pull my weight' are to be ignored.
"...who could salvage some good from Viya and create a platform for building successful industry verticals based on lead edge analytics/ML/AI "
The problem is that SAS has never been a leader in ML/AI, only a follower and a slow one at that. And it's never been successful building industry vertical solutions either, after years of trying.
The only time SAS can truly claim to have led anything was over 25 years ago, when it was the leader in statistical software with SPSS a distant second. Since then it's been running to catch up, but getting left further and further behind.
“ SAS is where SAS is in spite of OS. He did not cause it but he did not help it either.”
FACTS: OS contributed significantly to the current ongoing renewable revenue stream before CAS/Viya even came into being. Since then (I.e. for the past 12 years), SAS has continued to collect renewals on Mixed Models, HPA, LASR/VA, and related product components that OS was the primary lead contributor or significant force in their development and introduction to the customer base. OS worked tirelessly, often seven days a week to get these things to quality, timely production even in the face of considerable lethargy within R&D. So there’s that part of the revenue stream that he contributed to for over 20 years now. So much for him not doing much to help SAS’ current situation.
Then, there is CAS/Viya itself, which, in spite of several haters on this forum, has helped generated several hundred million dollars of revenue over its lifetime. Sure, it cost a lot to build, but that is the risk and gamble that all technology vendors take when attempting to advance their platforms.
Maybe SAS has too many business/sales/“marketing strategy“ types falling all over themselves to justify their existence, rather than replacing them with serious product and engineering talent who could salvage some good from Viya and create a platform for building successful industry verticals based on lead edge analytics/ML/AI — including a heaping helping of OSS were appropriate. This would require action all the way up to JG.
Doing so should be the highest of priorities. It’s the one thing that could turn the company around.
“ Absolutely spot on! Could also be why the post SAS gig was so brief.”
People who use the term spot on are to be ignored.
His tenure at post SAS gig could also have been short because he ate too much rhubarb pie. Since we are throwing sh-t out there I thought ai would pull my weight.
"projecting their personal blaming of SAS’ current state on to a departed scapegoat"
Two separate issues but loosely related. SAS' currently state is mostly on SAS. Without OS and his hours of toiling on stuff few wanted, SAS would a) have more money in the bank and b) have fewer V9 customers in distress and c) still would be for sale today with a declining V9 revenue stream as the strongest asset. SAS is where SAS is in spite of OS. He did not cause it but he did not help it either.
On the other hand, OS sliding back to academia is all on him. Some might see that as a sweet comeuppance for his arrogance and inability/unwillingness to listen. A scant few might even believe that he figured out(after a many year hiatus) that teaching is his true love.
"questioning the effectiveness of OS and his ability to function competently there is suspicious"
Yeah, suspicious indeed. Suspicious that such a highly competent person did not land an even better job.
Company representing the “post-SAS gig” has managed to significantly trend down in the DB Engines Rankings (https://db-engines.com/en/ranking_trend) in the 21 months following OS’ departure in fall 2022. That place plays by cutthroat Silicon Valley rules and therefore very beholden to its board and VC interests. Many other luminaries employees, including its founders, have departed from said company in the same time frame. In fact, another executive, with an absolutely world-class stellar reputation/background was let go a few months prior. Quite a few sharp former SAS employees remain.
One could draw many conclusions from this, but questioning the effectiveness of OS and his ability to function competently there is suspicious —- kinda reeks of people projecting their personal blaming of SAS’ current state on to a departed scapegoat.
also why academia seems a better fit. be as clever as you want without putting revenue and a company's future at stake. ironically, companies like sas, open source projects, even openAI, probably start that way. no worries about revenue and employees and survival and legacy and heirs.
"being technically great at producing things the market does not want doesn't move any needle. those things can even be technically superior in some dimension. it takes a lot of hubris to miss this obvious fact. possibly where some of the criticism comes from."
Absolutely spot on! Could also be why the post SAS gig was so brief.
Sorry for the confusion. By "SAS compatibility", I meant making Viya compatible with SAS.
It seems to me that would have been an appropriate strategic reaction to Spark, by offering a marketable feature that Spark does not.
Can you reply with some details about what “SAS compatibility with Spark” would look like?
Does that mean the CAS engine would seamlessly run Spark SQL and other procedures native to that ecosystem? Alternatively, does it mean that Spark would be extended somehow to run SAS code?
Doesn’t Viya/CAS Already interoperate with Spark via the embedded process data connector?
Which one or combination of these capabilities makes the most sense, given your understanding of the market?
Also, are you suggesting that VA be able to run with Spark as its backend analytics engine? Are you aware what that would involve from an engineering standpoint? Do you know about all the SAS specific micro services required to run VA in addition to CAS?
Just some things to consider. “Compatibility” with Spark sounds really nice on the surface but implementing it effectively is a very difficult nut to crack. Not to mention that many at SAS probably already feel that Viya interoperates just fine with Spark.
I’m sure OS and others would have been reluctant to abandon CAS, but that was not the only possible strategic reaction to Spark.
SAS compatibility was another possible reaction. That would have given CAS a marketable feature that FOSS lacked.
Perhaps an enhanced VA, or cheaper pricing, or other strategic changes would also have been appropriate.
It seems that SAS saw the threat, but made no strategic response.
The takeaway from these illuminating discussions?
Prior to being fully baked, it became evident to some that CAS/Viya was not going to be a game changer, yet we doubled down on that bet. Why? We knew it was not the answer yet leadership was too unwilling or too unable to find a better answer.
If that is a correct conclusion, then staying on the CAS/Viya train might not be the top reason for the company's decline but surely is in the top 5.
SAS provided an embedded process “data connector” for Spark early on. Initially, it worked with the SAS9 In-database tech and eventually CAS. At one point a few years into CAS/Viya someone from the SAS patent team remarked something to the effect “should we just concede that Spark has won and pivot away from CAS?”.
Certainly in retrospect that seems like a great idea. However, Oliver, many of the key engineers building CAS, and others in R&D already had lots of experience trying to bring SAS analytics “to external databases/data platforms” via established SAS in-database embedded process technology. There were a specific set of technical requirements for CAS (efficiently making multiple passes over the data being a major one) that made it difficult for those involved in its design to consider other alternatives.
Then there is the “SAS way” — the already decades-in-the-making, obsessively believed company cultural meme that analytics is the Royal and primary domain of SAS. And that we build the necessary, high-performance infrastructure and data plumbing (often as a “pass-thru” or round-trip data feeder working cooperatively with, yet not properly part of an external database or data platform) to make it work “our way”.
I’m no expert on Spark, but my guess is that Oliver would have chaffed at and JG would not have underwritten the wholesale abandoning of CAS to move to open source Spark as our primary platform infrastructure underpinning the future of SAS analytics. At that point SAS already had 10+ years of experience directly embedding analytics and SAS specific data management capabilities into other data platforms. Each of those external platforms had their way of allocating memory, handling exceptions, scheduling parallel activities, etc. these often conflicted with similar systems-level facilities within SAS. The often painful lessons learned from this were significant motivation for CAS, where the behavior of critical systems level capabilities would be consistent, known, controllable because they are defined by SAS — without having to negotiate with external organizations (including open source communities ) or vendors to make changes that might not be in their interest.
Again, much of the historical dynamics and motivations for CAS/Viya have been discussed ad nauseam and in great detail on other threads. It’s very easy to “Monday morning quarterback“ software design decisions made a decade ago — when it would’ve taken a radical shift in cultural thinking from JG down for SAS to embrace a nascent open source framework like Spark as the platform evolution in support of scalable SAS analytics. It would require a massive engineering effort with extensive participation in and acceptance by the Apache Spark open source community. From either that community’s or SAS’ perspective, such is not likely to be politically possible now and it certainly would not have been a decade ago.
I find the pontifications helpful, because I’m trying to understand what went wrong.
In particular, Spark 1.0 was released 10 years ago. So, well before the exodus of engineering talent, it must have become clear that Viya was competing against FOSS, and therefore its strategy had to change.
Why was there no change in strategy? No reaction to the competitive threat?
At least SAS compatibility would have given Viya a marketable feature that FOSS lacked,
Well then. I’m sure glad everyone understood the assignment and refrained from pontification. 😐
Correct … however, you likely agreel the market is far different now than it was when CAS (the infrastructural platbeginnings of what was ultimately branded as “Viya”) was conceived a dozen years ago. SAS is also a very different company today than it was then.
It could be observed that things could be very different had OS and other key technologists and engineering, talent remained at SAS, fully committed to evolving the components that comprise Viya, doing much deeper FOSS integration, building modern industry solutions thereon, etc. It appears that considerable momentum has been lost in the last five years while the market has dramatically changed.
These conditions are not unique to SAS. Worldwide and across the industry, billions and billions of dollars have been spent on failed technology development and related initiatives — simply “gone up in smoke” over the same time. It’s very difficult to predict market dynamics and steer a ship as large as SAS (whose traditional technology has been exposed to market risk for well over a decade) while evolving to a new platform that is more amenable to modern cloud computing infrastructure. There are lots of ways this can go sideways. No one should be all that surprised this continues to be the case at SAS given the variety of factors that have been discussed on this thread alone, much less many others here.
being technically great at producing things the market does not want doesn't move any needle. those things can even be technically superior in some dimension. it takes a lot of hubris to miss this obvious fact. possibly where some of the criticism comes from.
I have a long history in R&D and know plenty about what I speak having worked directly with OS for years. You and I must know each other from SAS then — if you have accomplishments approaching his, then you and I have worked together and known each other for a long time … decades.
Curious what you’ve built at a scale that approaches the complexity off MIXED Model PROCS, HPA, In-DB, LASR, CAS, and literally tens of thousands of lines of supporting TK code, etc. Of course, Oliver did not do all those things alone, but he was a driving force, leader, and significant if not principal contributor in all of them. I can think of maybe two or three other people in the history of SAS who have come close to this. AFAIK none of them are still at SAS.
Can you please give us some hints at your accomplishments? Perhaps you can explain why your influence was unable to persuade JG, Oliver and others to go in a more successful direction that would have SAS thriving today.
Various nuances surrounding issues with Viya, and many other aspects of SAS decline have been discussed ad nauseam on this and other threads.
I apologize if I mischaracterized your viewpoints and intentions, However, there were factors in play outside of Oliver’s control, bearing directly on why SAS is dealing with the issues it is today. He was/is not perfect, but for a long time he did more to move the needle at SAS and attempt to secure its future than any other singular individual I’m aware of.
@5wqh+1tPWcYil “ Of even higher probability is that you have some kind of seeming pathological angst against OS and have likely never been tasked with building anything approaching the scale and requiring the intelligence and grit that he has.”
Not really sure what you are talking about. Nothing I said was anti OS. Pathological angst? Wtf?
I knew him quite well. He has his warts but is quite good.
And I have built things of similar scale and get by just fine.
Keep on spewing BS about things you don’t know.
The CASL syntax is just atrocious!
There's no doubt that OS had considerable strengths or he wouldn't have done what he did. The problem was that he also had considerable weaknesses too, as I (and many others) evidenced firsthand, that made him the wrong choice for the job that needed to be done.
He may have been technically brilliant, but he made several major blunders from a broader business perspective. Here's a starter...
- Build a "cloud ready" software architecture without building a business capable of delivering a cloud product.
- Solve a problem (big data / massive parallelism) that was only relevant to a fraction of the customer base.
- Sit on the fence with regard to supporting SAS syntax, which was the worst of both worlds - still wedded to an archaic language that younger folk have no desire to learn, but not deliver backward compatibility for existing customers.
Anyone fancy adding to this list?
"never been tasked with building anything approaching the scale and requiring the intelligence and grit that he has."
Uh huh. And you probably also believe Viya is moving the needle? Hahahahahaaaaa
Funny thing about probability, even when assumed to be low, is that subject outcomes CAN be positive — as is the case with the poster you reference regarding their knowledge of said dynamics. Anonymity and the disclosure of minimal details is maintained for the privacy of many.
Of even higher probability is that you have some kind of seeming pathological angst against OS and have likely never been tasked with building anything approaching the scale and requiring the intelligence and grit that he has.
@5oyr+1tPWcYil “Are you sure you understand all of the actual dynamics in play at his post-SAS gig?"
The chances that poster knew the dynamics of OS and his departure from SS are negligible. But making sh-t up to fit the argument is the way in here.
"Enlighten us!!!"
He pretty much did:
"They’ve been on the edge of IPO for several years (still has not occurred), while managing to run off much of their best talent."