Subscribe to DSC Newsletter

SAS, SPSS and STATISTICA rivalry - could you highlight technical and/or "political" (dis)advantages of each?


Views: 11710

Reply to This

Replies to This Discussion


SAS and SPSS have a very long histories and that is one issue. They have good marketing and support. SAS (the package I use) is very good at attaching to large databases in an enterprise environment, and I find that nowadays companies often have data from various sources (DW, Web, free form text) and want to integrate them within 1 analytics environment. I have worked on projects in which I needed to access terabytes of data from the enterprise. I'm not sure how STATISTICA would be able to handle that.

Having said that, I'm sure STATISTICA has some good features, however I have not used it, so I cannot comment. For me, whatever gets the job done in a quality and efficient manner is fine. However as you imply, there are always other issues involved.

-Ralph Winters
Hi Jiri,

I do not agree in the History of inception of SAS & SPSS apaprt from SPSS (1968) most of the available Statistical Software company came into inception in the mid or late 80's. As per as my observation it is more about focus & forecasting of market dynamics which SAS could forsee & ventureed very well offfering solution to domain specific where advantage of Analytics could deliver benefit from the backofice to front office & millions of people. more or less SPSS did the same focussing more into social & market Research domain enjoying the leadership.

In case of STATISTICA they wanted venture in most of the domain rather taking the market share of leaders they invited competetion only in this observation i am giving example with MInitab.

In my observation more into the area of Business & Revenue generation, Statsoft Offering is one of the best value proposition balancing with perforance & pricing. need to innovate more agression in coming days month & year to maintain space in the Data Analysis Software Market.

Wish Statsoft Team all the best.
I have used all three packages extensively. I reviewed them according to their functionality for CRM data mining in a series of DM-Review Online articles in 2006. The bottom line for all three was that they all have very similar functionality. The price-point of STATISTICA Data Miner ought to drive more business their way. Why has this not happened? I think there are several reasons for this:
1. STATISTICA is not viewed by business management as "serious" software. I served on the analytics team at Fireman's Fund Insurance that selected STATISTICA Data Miner to replace SAS. The upper management decided to stay with SAS. They appeared to be more "comfortable" with SAS. Image is a huge factor in software selection. Lately, StatSoft has begun to appeal to upper managers in several industrial areas: Power generation and Pharmaceuticals. They are acruing much business with this approach. As StatSoft continues to focus more on top-down selling, they will become viewed by management in more traditional businesses (Retail, Banking, Insurance, etc.) as the most integrated and capable data mining package for the money on the market today.

2. STATISTICA was created by scientists and began by appealing to scientists. I started using it in 1989 in my scientific research to replace Systat, largely because of the fabulous graphics. But, SAS and SPSS got the jump on them with an all-out push by both companies into the business market in about the year 2000. The momentum built up by SAS and SPSS in the intervening time is very hard for StatSoft work against. And, the momemtum is self-sustaining, to some degree (e. g. Fireman's Fund Insurance).

StatSoft took a big step forward by providing a 90-day free license on the DVD enclosed in our new book, "Handbook of Statistical Analysis and Data Mining Applications" (Academic Press, 2009), by Robert Nisbet, John Elder, and Gary Miner. This approach to help new data miners work their way up the learning curve quickly is the complement to top-down selling. The union of these two approaches is the way I used to model complex ecosystems. We modeled how trees grow on the ground, using basic chemistry, physics, and mathematical concepts. Then, we modeled from the top-down also, simulating forest influences like climate and fire. We could not directly integrate these two approaches, because there was too much we didn't know and too much chaos in the data we did know. So, we connected the top-down and the bottom up approaches with simple assumptions (engineers call them "transfer functions") to create the most successful forest growth model in the world (JABOWA-II: by Daniel B. Botkin) at UC Santa Barbara. This is the way to model future successes for StatSoft - union of the bottom-up selling approach (where StatSoft excels) with the top-down approach they are following now, patterning the success of StatSoft's Rob Eames in Pharmaceuticals in the Philadelphia office. I am working on a similar approach to the Insurance industry.

I am confident that STATISTICA Data Miner will continue to gain in popularity, being tracked by KDNuggests each year. One day, there may be a "tipping point", where business in general recognize the great value provided by STATISTICA, then watch out! STATISTICA will prevail.
I totally agree with you Dr. Nisbet. As both a PhD candidate and a full time worker, I definitely notice the gap between business world and the academics.

As a Ph.D. candidate in the field of Operations Research, I find that people in academics talk all the time about the importance of optimization. NLP, MIP, and LP are all that they talk about in campus. But once I started working, I can hardly find anyone in the company who knows or even interested in adopting optimization in our daily work. Yes, managers like to talk about "optimization" but they are just using this term to make themselves look sophisticated. But once, I talked to them about really trying it, they just shy away from it.
Thank you very much for your remarks!

I am glad to see, fruitful discusion is emerging...
Let´s continue!

I am also looking forward to read new book by R. Nisbet and coauthors. Do you know the simpliest way to obtain this book? I am writing from StatSoft office of Czech Republic. Thank you very much in advance
The HANDBOOK OF STATISTICAL ANALYSIS & DATA MINING APPLICATIONS will not be in the UK Warehouse for European distribution until August. You can order now from either the ELSEVIER-UK web site or the web site, and then you will be the first to receive once the books are in the warehouse. Of course, you could probably order from the USA Elsevier and, but the shipping charges would be high [the USA has been shipping since June 1. (which site you go to is dependent upon your location) .....(15% discount is available by putting in CODE: 94637.)

As for Amazon, check or the corresponding sites in Czech Republic, Germany, France, etc based on their location.

Hope that helps.
A fruitful discussion indeed!

As a former SAS user that now uses STATISTCA, I can say without a doubt that STATISTICA's graphs are second to none. Aside from that, it is very similar to SAS. (I have not used SPSS enough to comment) There was nothing we did previously in SAS that we aren't currently doing with STATISTICA. Add that with StatSoft's price point and you have a very palatable offering.

I too, would like to know the best way to obtain this book by Dr. Nisbet...Thanks in advance!
See my reply to Jiri in the Czech Republic ...

But, here's a repeat, just in case:

The HANDBOOK OF STATISTICAL ANALYSIS & DATA MINING APPLICATIONS will not be in the UK Warehouse for European distribution until August. You can order now from either the ELSEVIER-UK web site or the web site, and then you will be the first to receive once the books are in the warehouse. Of course, you could probably order from the USA Elsevier and, but the shipping charges would be high [the USA has been shipping since June 1. (which site you go to is dependent upon your location) .....(15% discount is available by putting in CODE: 94637.)

As for Amazon, check or the corresponding sites in Czech Republic, Germany, France, etc based on their location.

Hope that helps.
Desite how much "academics" like anything in their "real life", the business world is less likely to need whatever it is that made those "academics" feel that way. Why? Because the business world, by merit of its economic focus, all too often needs only just as much analytical efficacy to highlight the specific conditions required to meet thier business goal.

Contextual Advertising? No insight into user-context is sought, only the differences in transactional run-up that lead to the 1 fulfillment relative to all the patterns that merely led to clicks.

NLP and search? this vaunted field has given us auction derived keywords that ontinue to be purchased only because those desperate for ANY clicks haven't dicovered their dismal performance yet.

Data Mining? Can we find Osama Bin laden? No, but we can routinely manage to likes of linking Iomega DVD ads with news-stories about "christmas trees causing deadly family fires" via their "Burn Baby Burn" slogan.

Inference is critical to academics, since they are really seeking it. The "business" world you seek to sell your software to cares less about statistical efficacy than it does;

support - SAS is expensive but they support you to the MAX. Their staff will actually write you experimental SAS-code and send it to you if you ask for that level of help.
reputation - nobody ever got fired for buying "IBM"
access to community, peer review, and public code repositories - again, SAS leads here too!

SAS leads this marketplace because they are a business dedicated to working inside and along with transactionally aligned computing platforms, instead of a bunch of scientists looking to prove their graphics are better or mahalonobian distance measure is "more efficient" or "more correctly calculated".

Do your history. SAS was traversing this very minefield back in the 1970's when their CEO, Jim Goodnight, and his cronies were emerging from their professorial offices at NC State in Raleigh, NC to build something better than SPSS, the only real statistical alternative for them at the time.

In the end, if your data is cogent to your predictive horizon, all the discussion about logistic regression vs CART tends to slip away.

So, there's the "analytical" industry's best kept secret; the better your data the less your algorithms matter. So, why all the discussion about algorithms? Could it be that most businesses collect data they find easy to collect rather than collecting data via custom experimental design that is necessarily predictive of the consumer behavior they seek to exploit? Hence, need algorithmic variety to try and make up for the headache of extracting inference where it can't or does't exist?

Actually, I prefer Mathematica to all of them! It is the only High-Performance alternative out there, imho. Try to make any SocialNet calculation in one step with SAS, SPSS, or R as you can within Mathematica!


Sooo, this is obviously an outdated thread which I think deserves rekindling (so, yeah they got Osama) as everyone is probably still questioning  what value each platform brings to them.

I think the Rexer Analytics survey is really beginning to demonstrate some of the dynamic shifts (and some hints to the reasons in these shifts) in platform choice. I would look at these as I think that, despite some of the survey quirks, the information is rather telling.

In my own experience, I have set up SAS shops, open source shops, open source plus Statistica, SPSS, and other more 'data-sciency' (for what that term is worth) shops. For the statistical work, I am fairly agnostic to the feature sets of each platform (though I would far prefer to write code in R or Python vs. SAS) , but I find that as a value-added approach to many businesses for which regulatory validation of analyses are not an issue (such as in clinical trials), Statistica in league with R as a code based extension (particularly at the server level - such as RevolutionR) unequivocally outsteps the competition in my experience. Outside of using specific financial risk model forecasting methods ARCH, GARCH, etc), there is nothing I have found in 20 yrs that SAS can do that Statistica cannot equally as well.

But sometimes it's the 'comfort' that comes with buying the shoe, almost precisely the same in quality and function as another shoe, but which costs 3x or more the price, that will cause brand attraction. (Ask me about the most recent fight I've had to wage....what dreck!)

As far as customer support, I've begun to feel as if SAS has somehow stratified its support ethic, placing more efforts in those companies who are there prime sources of revenue, and frustrating some of the smaller guys as 'acceptable' churn. Nothing really unusual here, but to your SMB folks, don't promise something you don't intend to give...right?

At one time, SAS was bar none an extraordinary support and service organization with their products. But more recently, I've begun to reevaluate their service side.  I'm noticing that rather untrained sales reps are placed in the position to explain how to interpret how SAS does things. To have a sales rep develop and demonstrate to a group of statisticians a forecast using the SAS forecasting suite in such a poor an untrained way is somewhat arrogant by SAS and insulting to me, the person (and statistician) who would be considering their enterprise purchase. I just don't get why SAS would ever let their reputation risk smudging by decisions like these, unless they are arrogant enough to believe there is simply nothing that can touch them based on their brand name alone.  In a recent experience I was engaged in, perhaps they could be right.

I notice that they have also appeared to rely on sales reps (versus technical staff) to visit public schools districts which have plunked down half a million on their EvAAS product. I was able to attend one of the post-mortem sessions of an analysis which evaluated teacher effectiveness. It was truly painful to watch the rep try to tap-dance around some quite impressive questions from teaching staff in order to understand what they were receiving. (I examined the *very* flimsy output as well...and what the district was getting certainly did not add up to the robustness that SAS was selling - but that's another story for another audience).

On the other hand, with StatSoft, I've seen nothing but the highest regard for their clients (they have a  very fanatical hands on - onsite sort of support ethic in my experience - they just don't say no - and they are also world benefactors, given their support of nations in economic crises - this simply cannot be dismissed and should be highly supported). Of course, a robust set of tools which are adapting to take advantage of new technologies in a very easy to use, cost effective, and scalable manner don't hurt either.

While I think that code based technologies help one understand what (and more importantly why) one is doing what they are doing in an analysis, a GUI interface such as Statistica as I've found doesn't  inhibit that understanding. In fact given the emerging spectrum of analytic talent out there (it can be scary sometimes who bills themselves as 'analyst' or 'data scientist'), I would much prefer someone concentrate on conducting a proper analysis than become frustrated with code to the point that the software ends up being used primarily as a VERY EXPENSIVE REPORTING TOOL (yes you all have seen this at one company or another).

I think that with the added graphics capacity of Statistica (it has always been a leader here), you have an extraordinarily powerful tool that can allow experienced statisticians/analysts, develop analysts internally  to a higher level  - yet another portion of the added-value rubric. It is quite nice when a tool, given its method of use can be used as an educational medium as well as a production suite, without obscuring the concepts needed to develop the analysis in code - the point for the analyst is to clearly understand the statistical concepts FIRST, not coding. In business, forcing both is in my opinion a burden ramp up time for getting analysts to a productive level. I say this and I'm quite comfortable in SAS or R and do eventually recommend code based development as a downstream part of an integrated curriculum.

In all, I see each product has its value and niche. But to say that SAS is the king of the giants is simply nonsense. And given where value in total cost of ownership, ease-of-use, scalability, extendibility, and level of service/support are all critical to business in the 'inference of information' age, I think that Statistica is a no-brainer, which is why I have migrated over to them (and leaves change in my pocket to build up hardware for the big-data and heavy visualization/animation work).

Now, once they begin to support CUDA processing and begin research on Dremel/Drill structures for really speedy data I/O , I believe they could begin to consider carrying a scepter at least :).


On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service