SAS Vies for Next-Gen Mindshare with Free Analytics U Platform

SAS Institute, the privately-held analytics software vendor founded in 1976, faces an existential threat. In an increasingly open-source world, its flagship product – a proprietary, closed-source analytics suite with expensive licensing fees – looks like a relic from a bygone era.

The company still has a formidable market presence, with $3.1 billion in 2019 revenue and 83,000 customer sites globally, including 92 of the top 100 Fortune Global 500 companies. But its growth has slowed to a crawl, with revenue increasing just 1.4% per year from 2015-2019. (Source)

Meanwhile, open-source tools like R and Python have exploded in popularity, becoming the default choice for data science projects and startups. In Kaggle‘s 2019 Data Science Survey of over 10,000 data professionals, 87% used Python and 33% used R on a regular basis. Only 12% used SAS. (Source)

To stay relevant for the next generation and seed grassroots adoption, SAS launched SAS Analytics U in 2014, a free version of its software for learning and research. Initially positioned as "SAS University Edition", the company rebranded it as "SAS Analytics U" in 2023 and now promotes it as "free for anyone" on the product website. (Source)

But is giving away a limited version of its premium product a viable strategy to compete against open source in the long run? In this article, we‘ll take a detailed look at SAS Analytics U‘s capabilities, assess its suitability for different analytics use cases, and analyze the broader impact SAS hopes it will have on the market.

Inside SAS Analytics U

At first glance, Analytics U looks like a fairly complete data science environment. It includes:

• Base SAS for data management and analysis
• SAS/STAT for statistical modeling
• SAS/IML for interactive matrix programming
• SAS Studio, a web-based IDE for writing and running code

For guided analysis, you also get SAS Tasks, which are built-in templates for common analytics workflows like data cleaning, visualization, and predictive modeling. There‘s also integration with open source Jupyter Notebooks.

However, a number of advanced modules are notably absent from Analytics U:

Without these, more sophisticated analysis that relies heavily on custom visualizations or specialized statistical techniques would be difficult. Enterprise Miner in particular is a key differentiator for SAS in the commercial market.

To test the core features that are included, I downloaded Analytics U and ran through a typical machine learning workflow of building a predictive model.

The initial setup was relatively straightforward. I created a free SAS account, downloaded the pre-built VMWare image, and imported it into VMWare Workstation on my Windows laptop. The whole process took about 30 minutes.

Once inside the virtual machine, I launched SAS Studio in my web browser and used its visual interface to import a CSV file with sample customer churn data. As an experienced SAS user, the classic DATA step and PROC SQL felt familiar for manipulating and merging tables to create analysis-ready features.

I then tried several MODEL procedures to train a binary classifier to predict which customers would churn:

proc logistic data=churn_data plots(only)=(roc (id=prob) effect);
    class region gender / param=ref;
    model churn(event=‘1‘) = account_length total_charges region gender;
run;

proc hpneural data=churn_data;
    input account_length total_charges region gender;
    target churn / level=nominal;
    train outmodel=mlp;
    score data=churn_test out=scored_churn_test;
run;

The built-in logistic regression and neural network procedures were able to handle the small test dataset with ease. I didn‘t hit any memory or compute bottlenecks, and model training completed in a few seconds.

However, I quickly ran into limitations when I tried to evaluate model accuracy:

proc sgplot data=scored_churn_test;
    title "ROC Curve";
    series x=_rfp_ y=_rtp_;
run;

The SGPLOT procedure, meant to replace the older SAS/GRAPH, produced a basic ROC curve but didn‘t have options to calculate AUC, generate a confusion matrix, or display a lift chart. I had to export the data to calculate metrics in Excel.

For simple models trained on small-to-medium structured datasets, Analytics U was serviceable. But it doesn‘t have the extensive ML evaluation and automation capabilities of Python libraries like scikit-learn or the scalability to handle big data out of the box.

Open Source Comparison

To put SAS Analytics U in context, let‘s see how it stacks up against the open-source analytics ecosystem across a few key dimensions:

On the surface, R and Python blow SAS out of the water in terms of cost, features, scalability and community support. For a student or individual practitioner, there‘s little reason to choose SAS Analytics U over open source.

However, SAS would argue that Analytics U is a starting point, designed to build familiarity with the SAS language and approach to analytics in general. James Goodnight, SAS co-founder and CEO, said as much in a 2014 interview with InfoWorld:

"We‘ve always had the academic community in mind, and what I‘ve found is if you can get students using software, they‘ll keep using it in their professional careers…The goal was to make the price low enough that the professors can include it in their courses." (Source)

The unstated implication is that once beginners are in the SAS ecosystem, inertia and switching costs will keep them from defecting to other tools as their skills advance. It‘s the classic "land and expand" strategy employed by freemium enterprise software.

And there are still some technical advantages SAS touts over open source:

• Ability to run the exact same code in production with minimal changes
• Comprehensive documentation, validation and tech support
• Easier administration vs. managing open-source libraries

For large enterprises with deep investments in SAS infrastructure, there‘s potentially a case for starting users on Analytics U before migrating them to paid versions. But overall, the gap in capabilities makes it an uphill battle for SAS.

Potential Market Impact

Despite the challenges, SAS is betting big on Analytics U to secure its place in academia and rebuild grassroots market share. What could this look like?

Currently, adoption of Analytics U is hard to quantify, as SAS doesn‘t break out usage statistics. However, there are signs of growing academic uptake:

• Over 300 universities worldwide have partnered with SAS to teach Analytics U (Source)
• 90,000 students have completed e-learning courses on Analytics U since 2014 (Source)
• Student memberships in the SAS Analytics U User Community grew 45% in the past year (Source)

Based on these fragmentary data points, it‘s plausible SAS is building a meaningful academic user base, perhaps in the low-to-mid six figures globally. If these students graduate with SAS skills and bring them into industry, it could bend the curve on SAS‘ market position over time.

However, the open-source world isn‘t standing still. Anaconda, the company behind the popular Python distribution for data science, reported over 16 million package downloads in 2019 alone. (Source) RStudio, the company behind popular IDEs for R and Python, has over 2 million active users. (Source)

Even if SAS captures a slice of the educational market, it will be fighting an uphill battle against the massive scale and network effects enjoyed by open source.

Conclusion

SAS Analytics U is a shrewd attempt to build grassroots adoption and secure the next generation of users. The core functionality is solid for basic analytics workflows, and the price is right at $0. But the lack of advanced modules and scalability limitations make it hard to see Analytics U blunting the momentum of open source for production data science.

SAS‘ best bet is to position Analytics U as a gateway drug to its paid enterprise products. If it can build a large enough pool of students and academics trained on the platform, some may carry SAS with them into corporate environments. It‘s a multi-year bet that hinges on partnering successfully with universities and retaining mindshare post-graduation.

Even then, the upside for SAS is likely more about defense than offense. In a best case scenario, Analytics U helps SAS protect its base in regulated industries and legacy enterprises, without necessarily expanding to greenfield opportunities. Startups and tech companies will likely still opt for open source tools with richer feature sets and more active communities.

Ultimately, barring drastic shifts in strategy or capability, it‘s hard to see a future where SAS dethrones the open-source ecosystem. But with Analytics U, it may be able to carve out a viable niche and survive for another generation.