"Forget it, Jake. It's Chinatown."
"What I don’t understand is what they are doing now. If Viya is not selling, wouldn’t you either make an aggressive investment to fix it, or cancel it to shift developers onto stronger products?"
This comment and flavors of it have come up in multiple threads. It's a good question that requires difficult decisions to be made. Only one person can make those decisions and he has historically shown that he can't, or won't, make them. Product analysis has been done multiple times with the intent to get out of losing products. Suggestions were made and each time went nowhere. Until the one person decides to call it quits, it won't happen.
I understand that in 2012, it was difficult for management to forecast the growth of R and Python and related infrastructure.
What I don’t understand is what they are doing now. If Viya is not selling, wouldn’t you either make an aggressive investment to fix it, or cancel it to shift developers onto stronger products?
JG did make such a statement. Seems like it was also accompanied by a proviso (or at least a tacit understanding, based on his other comments about having a succession plan of “from here to the ground” ) that major.minor V9 releases could go on and on.
CAS is a major architectural/functional evolution of LASR, designed to be far more democratic with respect to who could develop actions (i.e. pretty much anybody in R&D) with an attempt to be “more open” insofar as open-source client languages are supported.
It is important to recognize that CAS development occurred sm--k dab in the middle of the exponential growth of open source analytics and the explosion of “data science“ careers. It was pretty difficult to predict in 2012 just how pervasive the R and Python ecosystems would be by 2020. Sometime around 2010 JG stated that he did not want to embrace R, he wanted to ki-l it.
Here is some anecdotal backstory on the growth of data science in that era. SAS only gets brief mentioned as something she used in NCSU graduate school.
https://towardsdatascience.com/analytics-is-not-storytelling-a1fe61b1ab6c
Referring back to the early comments in this post, before CAS and Viya were on the drawing board, didn’t the Big Guy handicap any effort to design an outright successor to V9 by saying he was going to retire when SAS got to V10? And thus everyone was terrified of the great unknown void after the exit of the benevolent dictator?
I’m surprised by this assessment, although I don’t know the constraints.
What is the current plan? Is SAS trying to achieve backward compatibility? Or helping customers migrate to Viya? Or hoping that they want both platforms?
“It's easy to make armchair arguments that SAS should have factored backward compatibility into Viya. However, doing so is an organizational and technical problem set of such complexity and rigor that SAS culture was and very likely still is not up to task for.”
They did that same stuff with VA. Bundled free within other contracts and booked as a new install.
"Coming into the new year, we've got a really strong pipeline with a lot of upcoming deals..." - past quarterly updates.
"The Big Guy announced at the Quarterly webcast that they have 3,000 new Viya deals during 2023."
I think it was mentioned on another thread (or I heard it from a current employee) that all SAS 9 sales included a "free" version of Viya to try, or something like that. And then they counted that as a Viya sale.
may not used to do.
Should read:
may not choose to do.
A "ratcheting down" strategy for version 9 customers who, depending on their current product mix, don't have a smooth path to Viya seems like the only realistic option at this point This would include a strong analysis of each individual customer to see what it would take to take them forward to Viya, including some selective backfilling of V9 compatibility into Viya. For some customers, this will involve a costly manual migration effort that they may not used to do.
The bottom line: SAS did not have a clear-cut compatibility/migration plan when CAS was conceived (remember "Viya" is a branding term concocted in the 11th hour a couple of months before the new Architecture was publicly announced). As previous posts have pointed out, such backward compatibility is an incredibly complex problem -- one that frankly was beyond the capability of most of R&D leadership throughout the past 10 years.
Oliver did the best he could to move the company forward to a modern architecture, knowing the constraints JG drew around getting there. Anyone who does not understand the business and technical relationship between these two tall men (going back to weeks of world travel on the SAS jet during the Analytics Road shows of ~2007-2010), does not understand the degree to which such constraints were clear to Oliver.
It's easy to make armchair arguments that SAS should have factored backward compatibility into Viya. However, doing so is an organizational and technical problem set of such complexity and rigor that SAS culture was and very likely still is not up to task for.
The Big Guy announced at the Quarterly webcast that they have 3,000 new Viya deals during 2023.
"the ultimate objective was to migrate all ongoing/profitable SAS solutions (running on 9) and build new ones using CAS and Viya microservices."
Is this the current objective? It seems to require customers to rewrite their SAS code in CAS -- or else pay for two platforms.
"having a small, highly focus team to support Version 9 as it winds down and building out an expertly supporting new enterprise computing solutions"
I'm sure I remember the Big German announcing this at one point, or maybe his predecessor (or maybe even Dave Mac), and it being greeted with a gasp of disbelief. Like, how on earth could we have so few people supporting the cash cow, and so many people working on the successor to SAS9 that no one is buying?
But it never happened. In fact, I'm pretty confident now that it was the Big German, and the scuttlebutt was that Dr. G wouldn't let him do it because it would threaten the cash cow. The Big German never mentioned it again. IIRC, the next Town Hall was the one where he said "as my time here at SAS draws to a close" which sent everyone into such a tizzy.
Given design discussions around various solutions use of PostgreSQL and other real time streaming inputs (ESP was/is involved but that's a whole nother wrinkle), any directives he made were "fluid" although the ultimate objective was to migrate all ongoing/profitable SAS solutions (running on 9) and build new ones using CAS and Viya microservices.
Internals of the CAS Table Architecture evolved over time as performance and functionality dictated. The initial design worked especially well for multipass algorithms if data could remain RAM resident once loaded. Over time, additional database features were needed. SAS has long lacked the patience and expertise to build such correctly and in a comprehensive way. Witnessed the disaster of TKTS in the early 2000s.
Please correct me if I'm wrong. In the mid-to-late teens, didn't the Big German mandate that all solutions were to consume data from the CAS system, piped first through the data preparation solution?
This is the direction I recall, which I followed as he was the HMFIC for daily operations. Other mid-managers ignored this directive, which resulted in conflicts about technical direction. Am I remembering that directive correctly? It's maddening to think that I was doing what was asked while other managers weren't. Maybe I was wrong in my interpretation, or in following that directive?
As someone who has both a long-term insider view of Version 9 and Viya/CAS , I support your rejoinder.
However, @5bel+1pcUX1Ev does contain kernels of truth that expose two decades of growing incompetence within R&D, including layers of management. Many of us who were/are proud of the innovative heritage we built with SAS going back to the 80s, spent much of those two decades frustrated due to constantly trying to "outrun said incompetence".
Also, CAS is an evolution of LASR and I think some here either don't understand or recognize that. SAS needed a new multithreaded/multiuser engine to support not only VA/VA, etc but also highly scaled ML and multipass analytics algorithms. The highly parallelized in-memory computing architecture provided in CAS began in the fall of 2013 to address these growing needs that LASR (and certainly Version 9) could not.
The only real long term win I see for SAS has been documented in earlier posts about significantly cutting the current headcount, having a small, highly focus team to support Version 9 as it winds down and building out an expertly supporting new enterprise computing solutions, including radical modernization of Risk and Fraud based on whatever they can salvage from Viya as well as a commitment to open source, Innovation, contribution and integration.
I don't think it will be profitable to try to make Viya backward-compatible to Version 9. Too much water has passed under the proverbial bridge.
I don’t have the time or desire to address all the inaccuracies. And I’ll ad hominem where and how I please. Just as you clearly do.
Address the "inaccuracies" or keep the ad hominem to yourself. Simple as.
@5bel+1pcUX1Ev Talk about a false narrative. So much is inaccurate in your diatribe. Hints of truth here and there amidst all the BS.
"That's an interesting narrative, but it's not true. It was announced that Viya and SAS9 would meet at some point in the future, at least publicly. Then things changed."
It was true pretty early on. All the way up to the big man, not the big German, it was said that Viya wasn't for everyone and there would be two different things. Not too long after the message was altered.
"I don't think that was always the case. I was just a grunt but I had many conversations with managers about two competing ideas about Viya: v9 replacement and it's own thing."
That's an interesting narrative, but it's not true. It was announced that Viya and SAS9 would meet at some point in the future, at least publicly. Then things changed.
But you are describing the exact problem that prevented that from happening: the development teams and their management didn't want to work in the "constrained" environment of legacy support and decided to build something that is almost, but not quite, entirely unlike SAS9. The implementation of Viya was an exercise in self-advancement, with little thought given to the actual problems of the company.
Oh, of course someone re-implemented a microservice in Go because that's the new hotness, but the API? A development team may have said that it should be backward-compatible, but none of that matters because how could you tell? The API isn't versioned!
So lots of developers got to play with Go. And lots of the build pipeline team got to play with Jenkins. And lots of the testers got to play with various frameworks. In the end none of it matters because what they built cannot replace the product that is slowly aging out of the market, losing more and more revenue every year, and which is rapidly approaching the point at which not enough people at SAS will understand it well enough to maintain it.
All those clever people who thought they were going to re-implement the build pipeline only succeeded in making something which supports six ship events a year, one every two months. Compare and contrast to the previous build pipeline which supported arbitrary ship events. They replaced the build pipeline with something that isn't as functional by redefining the problem they were solving!
And that's very common at SAS: redefining the problem to make it easier to solve, moving the goalposts retroactively as it were. When we started working with VA it was always because Dr. G had talked to a highly-placed someone at one of SAS's customers, who said they would consider VA if it had this feature or that feature. So then that became a priority for the development teams, pushing other features into "future". Now they're just doing it to keep themselves from failing to deliver, out of self-preservation.
So if you ever wondered why Viya doesn't have a feature that SAS9 has, or why a known problem, grievance, or annoyance of SAS9 was never addressed, it's because it was too hard to do, not because, as you say, Viya is "it's own thing". There's no management pressure to address those issues because they are hard, and failure is the quickest way to find yourself working for someone less competent, but way better at managing expectations, at SAS.
"Viya was intended to replaced V9."
I don't think that was always the case. I was just a grunt but I had many conversations with managers about two competing ideas about Viya: v9 replacement and it's own thing.
Having Viya be it's own thing made a lot of sense to me. Probably Less than 1% of existing customers cared about CAS actions and the arcane results handling, etc., so don't bother trying to get them to use it. Maybe new money will like it. And you keep 2 revenue streams (but at the cost of having 2 code bases).
But having 2 divergent product lines didn't fit in with the goal of world domination held by egotistical upper management, principally The Big German.
I agree. In my own experience, opportunities to do deep and unique software development at SAS attracted recruiters from several top tech companies. I was ultimately recruited away for a significant compensation increase in a strategic role.
It is everyone’s full-time job to manage their own career. If you work/worked in a role that isn’t very marketable outside of SAS, then it is on you to manage your time outside of work in pursuit of education/self-directed, study, etc. necessary to “do the next thing”.
SAS with all that’s foibles is not ultimately responsible for the choices of its employees.
If you actually answered “Tell me about your role at SAS” with what you said then you are kneecapping your own career…
"...SAS is historically bad at knowing what products are installed and being used in production. Shelfware is rampant and finding out what is being used could open a can of worms. They would see some 'successful' products at thousands of sites are never installed. My guess nobody has a firm count. I've worked on exactly 1 that is fully migrated..."
Shelfware. This problem kneecapped my career after SAS.
"Tell me about your role at SAS..."
"Yes, I worked in a Potemkin village that produced glorious shelfware. We don't know how well it worked for people, as we had no details about installations or usage. But it was all great development theater."
"Hmmm...Okay...Well...we still have some other people to interview, so we'll let you know if you move to the next round." (click - crickets)
I’m not sure if the founders, principally JG, can envision SAS as anything, but a global analytics leader, with advanced historical language(s)/toolsets that a significant number of users consume directly while others use enterprise components built mostly upon a SAS-proprietary base.
Some of SAS’ largest customers have built massively complex enterprise systems on those historical language(s)/toolsets, and therefore will have a difficult time migrating off Version 9. Perhaps an AI could be constructed that would analyze all aspects of their current system and morph into something new and modern. That would be quite a technological feat!
Otherwise, a whole bunch of contractors and third-party service providers are going to be spending many years figuring out how to migrate these customers off Version 9 and on to some admixture of open-source and modern available products.
I don’t think that there is enough will on the part of SAS to make Viya a complete/viable target platform for such a migration. That’s going to leave customers looking for other alternatives and many already are.
This discussion is depressing.
Why? If I understand correctly, SAS seems to be in a real pickle, without any quick solutions.
On the bright side, SAS has v9, a proven renewal based cash cow. That's good, but not sustainable good because the V9 usage seems to have peaked and more and more customers are looking for something to replace V9. At best, V9 will be an ever declining asset. The most firmly entrenched V9 customers will likely be the last to pull the v9 plug. Will SAS have an alternative to v9 for them that soon? Or are customers more likely exploring non SAS options to replace SAS?
Viya was intended to replaced V9. But Viya, after 8 years of intense development effort, is not yet a V9 replacement. Will it ever be? And if so, will that happen soon enough to capture a fading v9 audience?
One has to wonder if Viya is proving to NOT be the next big thing for SAS. If not, how does SAS continue to carve out their $3B market share? And if they can not continue to carve out $3B(and grow) then the founders are going to have to accept the realization that their company may not be worth anywhere close to what they believe it is worth.
The forced giveaway software is a funny dynamic. It brings to mind this little dramedy, which I've named "Viya Con Dios".
"Hi, yes. I'd like to buy...." Customer.
"Oh, we know what you need. You need this special recipe that one of our big personalities cooked up. You'll love it! We promise!" - SAS
"Uh, no thanks. It will make me sick." - Customer
"Oh no, it won't. You'll love it!" - SAS
"I don't want that." - Customer
"How can you not want this great recipe? We'll just add it to your lunch box. It's no problem, really. Here! Take it!" SAS
"I said I don't want it. Can I just have what I came here for?" - Customer
"Of course you can! But first, you need to try this. It's great!" - SAS
[This cycle repeats a few times.]
"Oh....all right. I'll give it a try..." - Customer (now exasperated).
"Great! We knew you'd come around!" - SAS
[Customer begrudgingly tries new recipe]
"Cough, cough. Your new recipe is gross! I told you it would make me sick." - Customer
"No, no, it must be something else. Here, we'll add a little more of this. Now try it..." - SAS
[coughing intensifies]
"I don't like it. Can I have what I asked for?" - Customer
"Of course you can. But why don't you like it? All our internal people love it!" - SAS
"It's making me sick! I told you it would!" - Customer
"Well, here. Let's change this part of the recipe. Now you should love it!" - SAS
[coughing further intensifies]
"CAN I JUST HAVE WHAT I ASKED FOR TO BEGIN WITH!" - Customer, now shouting.
"Well...there's no need to raise your voice at us. We'll get you what you want." SAS
"Great! Finally!" - Customer
"But you should really try this again. It's like coffee. It tastes bad, but you'll come to depend on it!" - SAS
[coughing further intensifies]
"You know what? I'm no longer hungry. I'll go somewhere else since you aren't listening to me." - Customer
"We are always listening. Here, we've changed our recipe based on your feedback. Try it now. You'll love it!" - SAS
"Viya Con Dios" - Customer, while leaving.
"Hello?" - SAS
"Hello? Are you there?" - SAS
"Hello?" - SAS
"Come back! We've made a new change. You'll love it this time! We promise!" - SAS
It wouldn't surprise me if the forced VA and the other forced but not used applications had similar dramedies.
Yes, it is most certainly happening. SAS Technical Support expends an inordinate amount of time hand-holding those deployments at NO COST to the customers who subsequently continue to use SAS9 because Viya is what it is.
"And to get their big customers to buy Viya, SAS has "gifted" Viya to them and given free technical support to the ones willing to deploy Viya for "evaluation purposes" because those customers aren't willing to pay outright for something that can't replace what they already have. Those engagements aren't so successful because Viya is not a replacement for SAS 9."
Having to gift Viya to a V9 customer to get them to "evaluate" because the customer is not gonna buy Viya due to it not being a V9 replacement. Seriously...is that happening? That makes as little sense for SAS as trying to convert a paying vegetarian to a meat eater by giving them free beef. Basically giving away something undesired to replace a desired product and expecting payments to just roll along. Isn't there a stronger hand to play than that?
Has 8 years of effort produced any paying Viya customers?
IMO, SAS has wasted effort on multiple fronts for the past decade. Lack of a clear product direction for what comes after Version 9 being the chief mistake. Done right this would require teams of highly competent people researching open-source integration, as well as how to migrate existing customers to something new with minimal effort and maximum value to their businesses. For that matter, it might’ve signaled a carefully measured transition between being historically an analytic tools vendor to being an enterprise software company with products, strongly based in Analytics, ML and AI.
Instead, we had infighting between The Photographer and the Big German with scores of bland, semi-committed employees going on with “business as usual” whilst enjoying the “wonderful benefits” of working for SAS. Hubris was strong at SAS between 2010 and 2020. Goodnight (and by extension much of the company ) believed too much of our press.
The severe opportunity cost of all of this is now manifest!
@3jfx+1oW1I0bm points out that a search for Viya on LinkedIn yields exactly three job postings in the entire United States.
That, and similar results on other job sites, are probably the best metrics we can get.
@fda+1pcUX1Ev - I have never read a more accurate statement in my life. We keep "end of life" -ing products that are still in use and have no viable replacement on Viya. Heck, even a lot of our solutions still wont even support the latest SAS 9 release yet.
You're asking that here, of course, because the company is never going to answer those questions, even internally. They like to talk up the "big wins" internally (a million here, a couple million there...) but real customer numbers for any product are extremely hard to get, even if you're trying to determine where you should focus your tester hours on the inside.
And to get their big customers to buy Viya, SAS has "gifted" Viya to them and given free technical support to the ones willing to deploy Viya for "evaluation purposes" because those customers aren't willing to pay outright for something that can't replace what they already have. Those engagements aren't so successful because Viya is not a replacement for SAS 9. And it's still expensive.