SAS held its annual Analyst Conference in Steamboat Springs, Colorado last week, an event that drew scant buzz from persons not on the SAS payroll. For a good summary of major news from the event, check Cindi Howson’s post on the BI Scorecard blog (link here).
A few key points:
(1) SAS isn’t talking about SAS High Performance Analytics Server, its marquee in-memory software for predictive analytics. This product went into production fourteen months ago and has no public reference customers to date. Given the full-court marketing press SAS gave to this product last year, the implication is that (a) nobody’s buying it; (b) it doesn’t work; or both.
(2) SAS continues to provide a public speaking platform for a “customer” who sings the praises of High Performance Analytics Server but hasn’t actually purchased the product. Note to analysts: if some guy tells you how much he likes the product, ask him if he bought it.
(3) On the other hand, SAS is aggressively promoting its other in-memory software (Visual Analytics), which seems to be selling smartly. (SAS has a target to sell 1,000 licenses in North America this year). Visual Analytics is a slick in-memory BI tool, that currently does not support predictive analytics.
(4) SAS plans to add some simple predictive analytics to Visual Analytics in 2013.
(5) SAS’ BI revenues grew only 3.2% in 2012, compared to the double digit growth reported by other vendors. This quote from Gartner’s most recent BI Magic Quadrant offers insight into why:
References continue to report that SAS is very difficult to implement and use — it was the No. 3 vendor in both categories. Aggravating this, although it has a worldwide network of support centers and an extensive list of service partners, SAS’s customer experience and product support are in the lower quartile of vendors in the Magic Quadrant. A revision of user interfaces and an enhancement of product integration is under way to help improve the customer experience, but SAS must also improve its level of service — including level of expertise, response time and time to resolution.
(6) As Howson notes, Visual Analytics “…may offer a more modern and appealing interface, but only when data has been loaded into memory on the SAS LASR Server.” And there’s the rub, because it turns out that loading data into Visual Analytics is not exactly a day in the park.
(7) According to SAS product documentation, there are exactly two ways to load data into VA: from a registered table in a relational database or from a SASHDAT file stored in HDFS. According to SAS, the first option is “appropriate for smaller data sets because the data must be transferred over the network. If the table is unloaded or the server stops, the data must be transferred over the network again.” So if you’re working with Big Data, the only way to load data into VA is to first create a file in SAS’ proprietary SASHDAT format, store the file in HDFS, then load it into VA. And by the way, you must use SAS Data Integration Server to create a SASHDAT file. Surprise! More SAS software to buy.
(8) Howson misses the obvious point, though, that SAS Visual Analytics cannot read data from Hadoop unless it has been previously extracted and reloaded in SAS’ closed format. Which misses the point of using Hadoop in the first place.
(9) The other big announcement is that SAS now says it will support public cloud. Yay. I’m reminded of this article from November 2011, in which SAS CEO Jim Goodnight declared that cloud computing is “a lot of hype.” Color me shocked. It seems that when Jim Goodnight makes public statements about SAS’ product direction, we really can’t take him seriously.