Great for non-programmer analyst
February 22, 2018
Great for non-programmer analyst
Score 10 out of 10
Vetted Review
Verified User
Overall Satisfaction with IBM Watson Studio (formerly IBM Data Science Experience)
DSx is our area's advanced analytics platform. The product is new and so only our team uses DSx. Specifically, we have built a simulation program that will build financial projections for upcoming changes in our products. The health insurance industry has been in flux for several years, and there is greater value than in previous times on faster analytic turnaround. DSx provides the speed to spin up business concepts, provide financial impacts to the business concept, and spin down without heavy time investment. Therefore, our area can return results to our business areas with a much more robust set of problems (previously unapproachable due to time and resource costs).
- Very low administrative cost. Often, obtaining new technology/software can cause undue burdens on IT administration. IBM handles all of that from the cloud-based server, and so I just get to work. Our IT area is not needed to maintain its infrastructure, software releases, etc.
- Mixes the best of proprietary and open-source benefits. Though all the open-source modules are available for integration into DSx, IBM provisions a large library packages and even sample code that are maintained by IBM. This allows me to have the good "spoon-fed" options for building analytics provided by IBM directly, or to engage github/stackoverflow for any code, modules I might need for a particular situation.
- Lots of user interfaces for difficult coding situations. DSx has a SPSS model builder, and that's a tremendous help in building predictive models without having to know code. Additionally, there are a wide variety of tools for various analytical problems (Data Refinery, Data Catalog, Data Governance) which provider interfaces, rather than code intensive. A user wouldn't need to be a programmer to use, probably just some background in SQL would be sufficient.
- The stability of the application itself, though it has improved greatly, still struggles at times. I'd say about once a month I run into an issue where something is very slow or keeps crashing, typically lasting only a few hours.
- At the moment, pretty much stuck with a Jupyter notebook (unless using RStudio which I don't know much R yet), and I would like to use some of the improved Jupyter environments with enhanced user interfacing (Jupyter Plus). Not available at this time.
- Needs an interface for the server file directory (like windows explorer). Sorta pain to write in a scala notebook from java to access the local present working directory. Python is clearly easier with the "magics." Still I think a file interface would be nice for the server itself, maybe even for established dataframes in-memory.
- DSx has provided for analysis that has influenced business decisions into tens of millions of dollars by impacting the cost of the "unknown." Many business questions simply were infeasible to be understood in our previous analytics platform, took too much time, resources. Now, many of those questions can now be answered, driving business decisions and challenging well-established (incorrect) assumptions.
- I have not seen a negative ROI yet.
The mix of proprietary and open-source benefits that DSx offers gives me more flexibility than any other options I have encountered. I have the custom program building capability of Anaconda with the built-in predictive models of SPSS Modeler. I have more visualization libraries than Cognos as a result. I have a dedicated server for analytics without the cost and administration required for SAS or Oracle Database. DSx provided a happy middle ground for all the analytical problems I encounter.
Honestly, I do think that the best competition for DSx is RStudio because of the same reasons above, however, IBM DSx has open-source RStudio as well. Additionally, my work had on-going contracts with IBM, easier to purchase, no lawyers needed.
Honestly, I do think that the best competition for DSx is RStudio because of the same reasons above, however, IBM DSx has open-source RStudio as well. Additionally, my work had on-going contracts with IBM, easier to purchase, no lawyers needed.