With victory in hand, WPL will now have access to a large SAS customer base that had been effectively frozen by the uncertainty surrounding the litigation. Many SAS customers I know initially scoffed at WPL, opining that SAS would simply wear them down by outspending in a protracted legal fight, causing WPL to go broke or otherwise cry uncle. What's $20M to the largest analytics company in the world, they asked? Well, it now seems they have their answer: Who cares? David has tripped up Goliath.
Why all the fuss about WPS? It's simple: economics. SAS is very expensive and has a draconian annual licensing policy. WPS can potentially replace many of those licenses at a fraction of the cost. There's a ton of legacy data step programming at SAS customers currently – work that has much more to do with routine data access/manipulation/reporting than with SAS's advanced statistical capabilities. What was a vanguard programming environment in 1985 is dated today, and yet many companies maintain SAS licenses for straightforward programming tasks unabated. Their legacy programs have been running expensive Base SAS programs for years, helping to fuel SAS's unparalleled growth. That largesse is now threatened by WPS. SAS companies willing to invest in what should be a straightforward migration of many of their data step programs might save a king's ransom in licensing fees.
And those organizations transitioning their mundane SAS programming to WPS for the 80% discount may find added benefit to considering modern BI approaches for their new statistical initiatives. Rather than using SAS (or WPS for that matter) to simply move data around, how about instead an agile open source language like Ruby or, even better, a bonafide ETL tool like Pentaho Data Integration that's optimized for the task? And, while at it, how about taking a look at the R Project for Statistical Computing for new statistical challenges? The Base SAS language, be it either the SAS or WPS dialects, is dated, over 30 years old. After 20 years with SAS, I made the switch to R ten years ago and have never looked back. Object orientation trumps macros every time, just as powerful vector/matrix/array functions/operators trump “do loops”. Even SAS diehards acknowledge there's no comparison between the graphics capabilities of R and SAS. I just completed a predictive modeling project where I experimented with Generalized Additive Models, MARS, Support Vector Machines, Random Forests, Boosted Linear Models and Boosted Trees – all of which were developed by the R community and are available for free. Not sure you'll find all those capabilities in Enterprise Miner.
One area where SAS still has an advantage over R is in handling large data. SAS's virtual memory management operates on data sets of unbounded size, while R's data capacity is limited by physical memory. The R community is working hard on the “size” problem, as is commercial R vendor Revolution Analytics. Lest anyone think R currently works with only “toy” data sets as is often claimed, let me offer a 5M record, 20 attribute data frame on Windows 7 as evidence to the contrary. Expect SAS to continue hammering away with this advantage, however.
Its hard to imagine a CIO of a SAS-intensive company not at least investigating the cost-savings potential of WPS. What would I now do if I headed IT for such an organization? First, since SAS is purchased a la carte, I'd make sure I have a catalog of all current SAS product licenses and can precisely identify what applications are associated with each. Incredibly, when conducting such an exercise, many companies come to realize they've been paying for unused SAS products for years. I'd also ensure that the used licenses are run on the most economical computer hardware. Differences in SAS pricing among competing platforms can be substantial.
Second, I'd systematically look as which SAS applications are prime candidates for WPS. Migrating from SAS to WPS is not a rip-and-replace operation, but should be pretty seamless in the right situations. Data step-intensive programs that perform data movement and reporting tasks are good candidates; esoteric statistical proc programs are not. I'd then commission serious WPS proof of concepts for identified applications. I'd also let my SAS rep know what staff is up to.
Third, if our company were SAS-centric for both analytics and BI as, incredibly, many still are, I'd ignore the wailings of entrenched SAS programmers and identify new development projects for which we'd experiment with modern BI tools, starting with data manipulation. Pentaho DI for ETL comes to mind. I'd further look at the competing but well-developed open source BI suites from Jaspersoft and Pentaho for reporting, OLAP and dashboards. Software cost reductions here could be substantial as well, but some SAS lifers might have to be retrained. Is that such a bad thing?
Finally, I'd identify additional small new projects that might be suitable for R graphics and analytics. R programming is very different from SAS coding, so the organization might have to bring in one or more consultants to help jump-start the effort. I can almost guarantee that the step up in statistical graphics alone would excite the organization's consumers.
I'm not alone in my thinking on SAS/WPS. Analytics and statistical programming expert, ex-SAS and current WPS partner, and MineQuest President Phil Rack is more than willing to share his views on SAS/WPS developments too. According to Rack, “the idea that SAS decided to sue WPL was an indication that the WPS software has gotten pretty darn good.” For new statistical prospects, the Court ruling in favor of WPL “means that there’s literally thousands of companies who have not been able to justify the cost of products put forth by SAS who will now be able to leverage the language.” As for his own consulting practice: “At one time I was a SAS Quality Partner and it became obvious to me (as well as to many other QP’s) that our customers were not looking to embrace SAS products any longer but instead, were looking at ways to lower cost by either dropping certain products or moving the processing to smaller and more affordable platforms. WPS will now provide consultants with a vehicle to move their consulting in directions where their training and expertise can still be used without loss or change of rates.” For those SAS shops that might start paying attention to R, I'd recommend Phil's “Bridge to R” for WPS and SAS to jumpstart their efforts converting data between SAS and R.