Learn more about some of the other SAS vs. R vs. Python trends we’ve observed over the past five years in our 14-minute video on our extended analysis!
For the past five years we’ve been surveying our network of data scientists and analytics professionals to determine which tool they prefer to use – SAS, R, or Python.
Our most popular annual flash survey began as a battle between heavyweight SAS and up-and-comer R in 2014 and 2015, then we added Python as an option in 2016 and 2017 due to popular demand.
This survey has always been immensely popular with over 1,000 responses every year, and 2018 was no different, climbing to our highest yet with nearly 1,200 respondents!
To keep things simple, we only ask one question: Which do you prefer to use – SAS, R, or Python?
What did we find? An exceptionally close race! Drumroll please…
As you can see from the 5-year trend above, Python has been making strong gains since its inclusion in 2016, and this trend was noticeable in nearly every category that we track.
We also matched responses to demographic information, so we’re able to show how respondent preferences vary by factors like region, industry, years of experience, education, and data scientists vs. other predictive analytics professionals.
SAS, R, or Python Preferences Examined by Demographic Factors
Open source tools like R and Python are overwhelmingly favored by professionals with 5 or less years’ experience. While SAS continues to see strong support among professionals with 16 or more years’ experience, Python made noticeable gains here as well. Those with 6-15 years’ experience slightly favor R, but levels of support are within 5 percentage points among all the tools.
Note: Since Burtch Works is a recruiting firm, we do not ask the age of the professionals that we work with. However, we do know their years of quantitative work experience, which is highly correlated with age, and takes into account how many years since they first entered the analytics or data science fields (this might be after university, or, for those who changed careers, when they switched into the field).
Shifts at the junior end can often be indicative of growing changes in the market, so this year we also looked at how preferences among those with 5 or less years’ experience have shifted over the past three years with Python in the mix.
The data show that Python support among professionals with 5 or less years’ experience has doubled from 24% in 2016 to 48% in 2018. Support for both R and SAS has been declining for this group.
Bachelor’s and Master’s degree holders continue to show the strongest support for SAS, while PhD holders have consistently favored open source tools like R and increasingly Python. This may be due to many reasons, including the use of open source tools in research and academic settings, which establishes a tool preference that they carry with them into business careers, as well as the lack of licensing costs and the ease of putting Python models into production, among others.
As with last year, the largest proportion of Python supporters are on the West Coast, with the Mountain region close behind. Same as in 2017, R support is strongest in the Midwest, and SAS is most dominant in the Southeast. Python gained support in every region when compared to our 2017 survey results.
Python support continues to be highest at Tech/Telecom firms, while SAS maintains strongest in industries like Financial Services and Healthcare/Pharmaceuticals. R continues to see support in the Retail and CPG industries. As with the region breakdown, Python gained support in every industry.
Comparing SAS, R, or Python Preferences: Data Scientists vs. Traditional Predictive Analytics Professionals
At Burtch Works, we separate data scientists from traditional predictive analytics professionals because of the differences in salary bands, tool usage, data volume and structure, and a variety of other factors. The main difference is that we define data scientists as those working primarily with unstructured or streaming data, while traditional predictive analytics professionals work primarily with structured data. You can read more about our definition of data scientists here.
This distinction is important for our salary reports, but it also shows noticeable differences in tool preferences as well. Data scientists overwhelmingly prefer Python, while predictive analytics professionals working with structured data tend to prefer SAS or R.
A Few Lively Comments
Of course, one of our favorite aspects of running a survey that inspires such passionate responses is reading the comments, so we thought we’d share a few:
(R = Python) > SAS
SAS. Python next on my list to get comfortable in.
Here is my preference: Python > SAS > R.
80% R; 20% Python; 0% SAS
SAS first, Python second
Python! 🐍
Rrrrrrrr
SAS, but R for visuals.
I prefer to use Python I am a recovering SAS programmer
R (but truly, these days, whatever works!)
SAS is the tool for me (banking).
R of course 🙂
I prefer Python for most data science work, but R for making visualizations.
SAS by a nose over R.
I prefer to use R for statistical analysis and then Python.
Without any hesitation Python!
If you’d like to see the full results from last year’s SAS, R, or Python flash survey, you can view the 2017 results here. You can also click here to see the results of our other flash surveys, including those on hiring demand, machine learning usage, and what motivates data scientists and analytics professionals change jobs.
Interested in our salary research on data scientists and predictive analytics professionals? Download our studies using the button below.
Click to download our free salary reports
What did you think of this year’s results? If anything surprised you or you have any predictions you’d like to share, please feel free to leave a comment below!
Want to learn more? Check out our extended analysis below, where we go into more depth with additional trends that we found in this year’s data.
Matt Sandy
I think Python is great, but I also use R and Node for just about everything.
MARCO NEVADO
Cost is one of the factor that affect SAS preferences.
Thomas Emerson
What about the others, e.g, Matlab, or S-Plus (both of which I’ve used but not by choice)? Also strange to see Python in here since it’s a general pupose language
Adam
Python is definitely the best – not an opinion at all, just an objective fact! 😉
Kelly
Inspirational content, have achieved a good knowledge of the above content on Data Science training useful for all the aspirants of Data Science training.
Jim Metcalf
SAS provides one significant element that analytic open source initiatives do not. SAS has a large and dedicated staff (100s) of independent numeric validation experts whose job it is to construct a separate technology tower to validate the numbers that come out of an algorithm. These numeric validators are typically PhDs in statistics, econometrics, operations research, data mining and mathematics. Their job is to numerically arm-wrestle with the PhD developer until both of their numbers agree. This can take months if there is a disagreement, especially for a thorny algorithm with many branches. Often, open source advocates suggest just ‘well, just look at the code’ to see if it’s correct. I’ll argue code review simply cannot be done for complicated statistical algorithms to determine if the output is correct or not.
Nearly every analytics developer SAS hires has written an R package or two for their dissertation. If that algorithm proves to be market viable, then it’s put into ‘the machine’ and error checking, numeric validation, scalability, documentation, examples, technical support are all added to the mix to ensure a product that is hardened, annealed and numerically correct. You get what you pay for.
That said….the benefits of R are HUGE. R represents the statistical cutting edge…the Wild West. New algorithms can be developed, proliferated and explored very rapidly. SAS will always lag behind this cutting edge by choice and design.
Python is incredible because of its vibrancy, penetration and (also) wild west nature. SAS is adding Python integration to the platform with more detail here: https://blogs.sas.com/content/sasdummy/2017/04/08/python-to-sas-saspy/
Nope, I don’t work for SAS (anymore); but I’m well-familiar with how they make the sausage.
subhash
thanks for such an information
ynkjohn
Thanks for sharing …..Good information
statmike
and…. they used Excel to show their stats *facepalm*
to correctly interpret these charts it is really necessary to include the n= on each of the x-axis categories
a good next level analysis would be showing how these 3 are used together by the respondents. I use all 3 and many times have processes where one calls another so it would be hard to answer the survey with just 1.
Isom
While I use both Python and R, Python is easier to work with when I am running ML and deep learning algorithms. More so, python enjoys an extremely wide and active community of users worldwide. I can easily get answers to questions and challenges on Python (R as well) on StackOverflow. However, for SAS, I’m not aware of its use in machine learning, let alone deep learning. Maybe I need to be educated on its use.
Karen
This is really informative. Thanks for taking the time to share this with us.
John Stanmeyer
Interesting. Looking at the 5-year trend, it suggests that maybe SAS has stopped the bleeding. But the 3-year deep-dive on younger professionals — who hold the future — tells a different story. Python is king, SAS is dying and even R has peaked. All while SAS is arguably easier to code, understand, and edit, and equally capable. But the cost barrier is ridiculous. A first-year SAS license costs thousands of dollars, versus Python which is completely free. SAS is claiming to play nice with open-source by allowing Python & R in the same studio environment, but that’s not nearly enough. And the SAS cloud solution is proprietary, not available on AWS. The ONLY thing that SAS can do to avert its inevitable demise is to make the BASE SAS and SAS/STAT language completely free and open-source for ALL users (including for-profit corporations), as soon as possible, with no limits on data size. This includes the Linux server version which needs to be made available on AWS at least for serial processing. By becoming truly open source, SAS will ensure its relevance for years to come, and will still make plenty of profit on additional modules and services including cloud and grid processing. Otherwise, SAS will be relegated to a niche language used only in clinical trials, and even there other tools are starting to chip away. Will SAS see the writing on the wall, or remain short-sighted? Are they too blinded by next quarter’s and next year’s revenue stream to see beyond to horizon two and three? Where will SAS be in 5 or 10 years? SAS’s revenue growth is stagnant at 1% while the data analytics industry is growing in double digit multiples. It’s now or never, SAS.