Hardly anyone uses Viya. Even the existing customers to whom it was given to gratis to evaluate. Eight years later after inception, there still no clear migration path from V9 to Viya.
Call me pessimistic, but that doesn't sound like a wonderful selling feature, whether "selling" is defined as:
Perhaps the Big German is going to have the last laugh.
@4qbs+1pwgAngm …. Excellent points with one correction.
VA was initially developed/released ~15 years ago. At that point open source analytics were not nearly as pervasive as they are today and term “data scientist” was not well defined and had not yet become mainstream. Being a company known for advanced analytics, SAS was naturally more focused on those used cases and likely more traditional statisticians working in business/industry who needed a powerful visualization tool based in a compute server where they could share big data. Hadoop was gaining ground so
the LASR storage infrastructure model was built around it.
Several posts observe that SAS itself came from Academia and was more focused on traditional statistics. Designing software for simple/common business used cases was not a natural thing for SAS R&D, at least not during the beginnings of VA.
Was it not the case that SAS targeted VA to be rich in advanced analytics from the very beginning?
Wasn't this was done because at that time the BI visualization competition lacked many of these advanced statistical capabilities?
So, the thinking was: provide the advanced features to capture that segment of the market, and then backfill to get the more commonplace BI features.
WRT VA performance and Scaleability -- : @4bhi+1pwgAngm provides the backstory. World class platform systems programming + mid m-tier infrastructure (java. At that point, Go had not yet been released ) talent is needed to make this happen. By the time LASR/VA emerged (~2008), SAS had maybe 10 such developers,. How many of these does anyone here think could have passed a Google technical infrastructure engineering interview without 6 to 9 months of full-time preparation? From my experience that number is approaching zero.
Rewind to the late 80s when the platform development game was all about systems programming talent on specific host hardware architecture and OSes. SAS had several such experts who were being headhunted for jobs by the tech Titans of that
@4bhi+1pwgAngm explained why that was no longer the case by 2005.
Thanks all for your responses; you’ve given me a bit of closure. It’s most disappointing that the feature set wasn’t right. I knew some of the VA developers, and they would have happily built whatever management wanted.
The Pre-Sales Perspective is always valuable. If pricing was determined by architecture, does that apply to other products on top of Viya?
This sentence below encapsulates how we ceded our core market, and where it stands now. And these products are all old. So somebody new will innovate, and grab a chunk of this market.
“Data scientists were well served by open source, and analysts / business users were happy with Tableau, Qlik and PowerBI.”
Yes, the Big German's only prior experience to SAS was in academia...that is telling!
VA felt like an academic's pet project. So he would prioritize building a correlation matrix visualization, over basic aggregated calculations. It was nuts, the priorities were all completely wrong. Release after release we saw new wizzy features that might appeal to an academic, but which your general analyst has absolutely no need for.
The Big German was no doubt a brilliant data scientist, but he clearly had no clue about what your typical business user needs from an interactive visualization tool. That became really evident when he did the big VA/LASR World Tour and he was up on stage doing a demo to managers and executives and he spent half the time showing code! We spent the next several weeks apologising to our customers for wasting their time!
He'd travelled half way around the world for this big launch event, and then he refused to meet any of our customers in 1:1 meetings. It was utterly bizarre. No wonder he built stuff no customers wanted - he never wanted to meet them and ask them.
And much of this is the result of the decline to downright denial of doing basic research among the true Principals and others (even some entry-level luminaries) within platform and data management R&D. A handful of Director to VP level power players presided over significant decisions and their ideas nearly always ruled. The micromanaging was so bad under one of these folks that I as a principal developer was, for a time prohibited from talking to another principal developer on my team about design ideas.
In contrast, during this 20+ year decline in the R of R&D, Ami first and then Oliver arose as “Rockstars”. Each had a very dominant workplace personality, worked 60+ hours a week and gained visibility/influence with JG within a few short years of joining SAS. Ami (RIP) was sufficiently abrasive that virtually none of the top R&D talent would work under/with him if they could possibly avoid, it. Truth be told, in the HR climate of the past decade he would’ve not survived as long as he did. Oliver on the other hand, was able to attract small teams of the best programmers to work with him.
From my perspective, there was an imbalance between the micromanaged rank and file R&D and the elite Developers working under the rockstars. Don’t get me wrong, there was plenty of excellent talent within the rank and file, it was just being held back by micromanaging/controlling division heads in their mandate to keep $$$ flowing from the V9 (which includes its MVA predecessor versions) cash cow. I don’t think JG understood that he was motivating them to do the wrong thing. From all appearances, they were acting in their own self interest and not in the longer-term interest of the company. This can be seen in the lack of professional development, budget to send rank and file talent who had earned the privilege to relevant industry conferences, leadership in and contribution to open-source, and the need for pervasive cultivation of platform integration talent as Linux and ultimately public cloud platforms became center stage in modern computing.
So, at the end of the day we had a relative handful of R&D division heads who throttled down by their own self%interest to keep JG happy in the matter of the classic SAS cash cow (after all that big bonus to help pay for the beach house was depending on it), and the fact that they had micromanaged their divisions, in a way that very few of even the brightest and most motivated developers were able to keep up with the key skills necessary to build world class software as the hardware and computing platforms continue to evolve. To be fair, some of this skill development occurred in small pockets , but not at the level required to keep SAS relevant.
All of this is one reason why very few effective Architects (newsflash, product managers are rarely qualified to be software architects) emerged with R&D post year 2000. There was too many politics and not enough skill development. In total this, and not doing correct initial product research is precisely why VA, Viya and other products have not been more successful. From where I set, this is all a direct result of ineffective executive and mid-level management. No amount of virtue signaling about how wonderful your workplace is will change this.
What applications are part of Visual Studio?
"...was lacking in so many basic features..."
This is a common problem when developers lack business experience and domain expertise. Technical expertise alone did not guarantee a home run for Viya. Recall, the Big German's forte and only experience prior to SAS was academia. Academia is great but it is not operationally reflective of the business world.
This comes as no surprise to those readers who came to SAS with significant business experience and domain expertise.
The Big German and the sappy photographer shared that same deficiency. Having said that, both appeared to have good intentions. Sadly, both were in way over their head for their respective roles.
I can give you a pre-sales point of view on why VA failed.
We actually sold quite a bit of it initially because I had some slick demos with some wiz-bang features like the correlation matrix and multi-variate forecasting that competing tools (Tableau, Qlik, Power BI) at the time couldn't do. Combine that with a business story around actionable insights and we had an interesting value proposition.
But scratch beneath the surface of the shiny demo, and VA was lacking in so many basic features (which I won't go into detail on) that it was extremely difficult to actually use it for finding answers to business questions. Some of those features came later on, but they were too little too late. So early adopters of VA, quickly realised it was missing some key features and therefore failed to get business value from it...and word spread in the market.
Also, the architecture of VA meant it was extremely expensive to run (and license). The SAS marketing machine loved to claim it could handle "a billion rows...that's with a B! ". Aside from the fact that a billion rows may sound like a lot but is actually quite common, it would typically require at least 256 cores to handle that. So on top of several $100k for license, it was another several $100k for hardware, and throw in another $100k of services just to install and set it up. And if you had more than a few dozen concurrent users, it would be even more than that. The cost per user could be enormous...and really unpredictable. It was very difficult to estimate how many cores would be required, involving filling out a ridiculously long form asking all sorts of questions around the types of interactions, frequency, etc. that the customer could never reasonably be able to provide an estimate on. So we would guess, generally low balling it to keep the costs from being too outrageous, then the customer would find it would perform like a dog.
By way of price comparison....this was a typical scenario.....150 users with moderate data volumes....$500k per year in licenses, and $200k per year in AWS instances. $4666 per year per user. Microsoft (Power BI) was charging $180 per user per year. So VA was utterly, hopelessly, outrageously expensive. And not a good product either.
By the time VA had all the features one would expect of a modern BI tool (and had been moved off Flash and onto HTML5), Power BI, Tableau and Qlik completely dominated the market. We realised we had lost that battle so tried positioning it as an advanced data discovery tool / light data science tool, when combined with VS and sometimes VDMML...but it was still too expensive and it was quite simply a tool that nobody wanted. Data scientists were well served by open source, and analysts / business users were happy with Tableau, Qlik and PowerBI.
There's no point in anyone working on VA anymore...it's dead....but then I guess that applies to all of the Viya products. :-(
VA does not allow, in any elegant way, incorporating SAS code within the interface, so programmers are still needed for data manipulation prior to loading
I asked that question because SAS/VA was exactly in our wheelhouse. It should have succeeded, because analytics is our core competency.
I've heard it suggested that because SAS/VA was tied to SAS/Viya, that made the price too high. It might have done better on SAS V9; I don't know.
But if SAS can build one thing well, one would expect that one thing to be software for analytics. So I'd love to know why SAS/VA did not succeed.
Is anyone buying Visual Investigator (VI)?
@3tfi+1pwgAngm .... now this could be a very deep rabbit hole. Here's a summary.
VA needed a high performance multithreaded/multi user data+compute server that did not exist in V9. To remedy this, Oliver and his initial very small team built LASR which, after a few years, evolved to CAS, requiring a larger team, and a foundational iarchitectural that simplified/democratized analytics and data management action development.
Prior to CAS/Viya, I'm pretty sure there was limited LASR/VA integration with V9.
What would it mean for an interactive/visual based tool like VA to be compatible with V9? (not asking in an acerbic or rhetorical way -- just don't know the answer because it's not my area of expertise)
Why did Viya fail?
No compatibility with V9 for starters. That was and still is a big turn off.
Others can scribe additional reasons as they see fit.
Why did VA fail?
If I remember correctly, when VA came out nobody wanted it. Customers were using Power BI or other tools. So SAS bundled it gratis in new sales contracts. Minitrue claimed VA was a great success.
Try as I might, I couldn't find a customer who used VA. Customers used some other BI tool, but not VA, even though they had VA for free. But then again, maybe I'm not remembering it correctly.
'Oceania has always been at war with Eastasia. Since the beginning of your life, since the beginning of the Party, since the beginning of history, the war has continued without a break, always the same war. Do you remember that?'
'Yes.'
' Eleven years ago you created a legend about three men who had been condemned to death for treachery. You pretended that you had seen a piece of paper which proved them innocent. No such piece of paper ever existed. You invented it, and later you grew to believe in it. You remember now the very moment at which you first invented it. Do you remember that?'
'Yes.'
'Just now I held up the fingers of my hand to you. You saw five fingers. Do you remember that?'
'Yes.'
O'Brien held up the fingers of his left hand, with the thumb concealed.
'There are five fingers there. Do you see five fingers?'
'Yes.'
Excerpt from "1984" by George Orwell
https://6sense.com/tech/analytics-and-bi-platform/sas-viya-market-share#
Pride, arrogance and vanity are the Achilles Heels of SAS. Viya, sales losses, and other longstanding messes are merely the symptoms of these problems.
"Forget IPO, they should be focused on survival."
SAS recently kicked the IPO can further into the future(now 2025). Perhaps that kicking was a face saving excuse, but under the covers, the emphasis is now on survival?
No significant Viya sales coupled with an ever declining v9 revenue stream has to mean fewer interested buyers. Those who might still be interested won't want to pay the price the founders have in mind. All because the SAS cow is producing less milk each day. Sad.
A migration path seems unlikely to help. That requires customers who want to migrate. Few customers want to learn CAS, but many want to continue running SAS.
If Viya can be made to run SAS code, it may be worth salvaging; but in another thread that was described as a hard problem. If it can't be solved at reasonable cost, it seems better to cut the losses on Viya, and shift developers back to improve V9, or create V10.
What is management's current plan?
"The years wasted rebuilding Viya multiple times could have been years/money spent on changing 9.x to be highly scalable, cloud-native, etc"
That's a very big "coulda woulda shoulda". Attempting to radically morph 9.x would likely have been an even bigger clustedFu-k, especially when compatibility issues between SAS releases going back to pre-MVA (Version 5) are considered.
The Big German used the term CF more than once to explain the dynamics and conflicts going on within R&D surrounding these interoperability, compatibility, etc. issues.
Earlier attempts to parallelize/scale MVA (predating version 9, but the same basic Architecture prior to TK) was an exercise in futility that created the SPDS spinoff -- which is earlier incarnations required its progenitor, Ami, to actually go to customer sites and configure it.
Then came TKTS, a cluster fu-k of epic proportions that clearly demonstrated the difficulty of building database technology correctly. The remnants of that train wreck remain in Fed Server -- and how much revenue is that generating?
Together, these data management fiascoes no doubt, further convinced JG of what he already deeply believed -- that simply su-king data out of databases, or bringing Analytics to the Database (some thing the Big German was integral in making successful) became the cultural imperative.
Then the big cloud vendors started building, massively, scaled, distributed databases with very low cost of entry, especially for average size workloads.
Wouldn't you know it, they figured out they could integrate analytics pretty easily with their native data platforms, so, with their massive market, capitalization and profits they went out and hired the best Analytics PhD's on the planet, offering them 2-4x what SAS pays.
Then open source took over the world ...
To answer your question, yes. They shouldn't have released Viya without a clear migration path. Or they should have planned it as an add-on to 9.x rather than a replacement. The years wasted rebuilding Viya multiple times could have been years/money spent on changing 9.x to be highly scalable, cloud-native, etc. Instead, they've continued to invest in a product that has never taken off. Such a waste. So long as they continue down that path, they will continue to see sales decrease. Forget IPO, they should be focused on survival.