Time to switch
R is the most popular analytic language in the world today following an incredible increase in popularity over recent years. If you’re involved in data analytics, you need to be fluent in R. This change has driven a growing number of migrations from SAS to R. There many reasons for migrating from SAS to R, including:
Talent. Most universities have adopted R as their primary analytical programming language in departments such as mathematics, statistics, biology and social science. Companies recruiting new graduates can expect them to be skilled in R and keen advocates of its use.
Cutting Edge Analytics. Businesses looking to outperform their rivals don’t have the time to wait for the latest SAS release. Academia’s adoption of R has led to the efficient, open-source production of new algorithms. Database vendors, such as Microsoft, embed R into their packages and provide high-speed platform integrations
Cost. While money isn’t always a primary driver, cash flow and budget limitations remain ubiquitous pain points. IT budgets have come under increasing pressure over the last few years and software licensing is an obvious opportunity for cost reduction.
Support. R has an inherent support network. Today, there are a range of support options available for users, allowing analytics teams to deploy R in a safer production environment. There is also a wealth of documentation online.
Agility. R is nimble allowing companies to innovate quickly. Cloud based modelling environments allow a shorter time to value and/or fail fast.
Another reason that organisations are looking to migrate from SAS to R right now is the Data-Driven movement. Companies are increasingly looking to Data Science to drive their business forward, and R is a fundamental pillar of the Data Science movement.
The Road from SAS to R
The move to R is not something that can, or should be done overnight. SAS teams have invested years of effort building large libraries of SAS macros and code, not to mention hiring teams skilled in SAS. When moving to R, companies must consider a number of factors, including re-training and intelligent migration of a large, varied code base. But it is possible and increasingly attractive to do so.
Mango’s Data Science team are experts in both SAS and R, enabling them to empathise with analytic teams as they make this transition whilst offering a range of training and consulting services.
But … how do you migrate 20 years’ worth of SAS code to R?
Converting Analytic Code
Mango’s proven approach to code migration is based on 5 steps:
1 – Assessment of Analytical Routines. The first step is to analyse the existing SAS code using technology developed at Mango explicitly for this purpose. In particular, we focus on the analytic code and models to be migrated to design appropriate tests and success criteria. At this stage, we may also recommend improvements based on the wider availability of analytic algorithms in R.
2 – Development of Unit Test Structure. We use a “unit-test-based” approach to code migration to provide documented evidence of successful migrations. Our unit testing framework has been refined over many migrations and includes, in particular, tests for common SAS analytic procedures We also supplement our own test cases with customer data, interpolating additional unit tests as necessary to ensure the most complex migration elements can be thoroughly tested on familiar data.
3 – Migration Infrastructure. Once the scope of work has been defined and agreed, and the existing SAS code fully analysed, the existing code is baselined and tested within Mango’s formal development environment (encompassing a version-controlled repository and Continuous Integration server). This framework is integrated with a reporting mechanism to enable progress tracking and direct feedback on migration progress.
4 – Development of Standardised R Functions. The next task is to migrate the centralised SAS macro libraries. In this scenario, ‘line-for-line’ migration doesn’t provide the best solution due to the significant differences in “style” between the 2 languages. If a “line-for-line” approach was employed it would ignore the fundamental differences between the languages and result in verbose, inefficient R code. Our own migration technique presents an opportunity to streamline existing functionality. Testing in-house allows us to improve code, whilst retaining the comparison of key values.
5 – Training. Relevant training is critical to the success of any software project. Software underpins an analyst’s everyday workflow and, to ensure a successful cultural transition, our trainers are experienced in both SAS and R. Our developer training focuses on package building and code optimisation, ensuring you can maintain and extend your new R code base in the future. The code migration is used as a fundamental part of this training, so users experience an “R-based” coding approach using familiar procedures.
Reducing risk, accelerating value
Choosing to migrate from SAS to R is a significant undertaking for any organisation. Careful planning and implementation are required to ensure everyday activities aren’t disrupted. Mango can take away some of the pain, eliminate much of the risk, minimise discrepancies in code translation and deliver successful projects that empower your company’s analytical capabilities.
