IBM SPSS Modeler 18.2
August 2019
“Who are JMP’s competitors? (maybe we have them in narrow areas) — but there is nothing else like it.
Who is as interactive?
Who else does as great a job at DOE?
Who does analysis faster and bigger on the desktop?
Who else does Graph Builder as well?
No one, because JMP is truly unique in the crowded marketplace.” -John Sall
It’s important to remember this as we review these competitive analyses. While many of these competitors will compete on features, for scientists and engineers, there is nothing else JMP in marketplace at this time.
Company
- Global, Service, Software and Hardware provider.
- Since IBM acquired SPSS, they continue to primarily promote predictive modelling capabilities and integrated SPSS modules with the rest of their technology stack.
Product
- IBM SPSS Modeler is a core IBM package for doing data mining/predictive modelling. Originally purchased as Clementine by SPSS, its process flow diagram paradigm lets you visually lay out the steps of your modelling process and is perceived as easy to use. They have Gold, Premium, and Professional versions. The Premium version has everything the Professional version has as well as text mining, entity analytics and social network analysis and costs considerably more. The Gold version has everything the Premium version has plus Collaboration and Deployment Services, real-time and in-batch deployment, SQL pushback, Model management and version control, champion-challenger A/B testing. The themes for version 18 release:
- Big data algorithms
- More extensibility with open source
- Platform flexibility
- Version 18.2 highlights (18.2 was released in December 2018 with these known issues
- A more modern look and feel (though you can still use SPSS standard, classic or Windows interfaces)
- Several enhancements to text analytics, including the ability to import SPSS Text Analytics for Surveys projects (.tas).
- New data views and Chart Builder, providing much more interactive data exploration and new chart types like the relationship chart, theme river chart, etc.
- All of the new modeling nodes are Python-based, including Gaussian Mixture models, Kernal Density Estimation, and Hierarchical Density-Based Spatial Clustering.
- The full list of new features is here.
Target Audience
- They primarily sell to statisticians, social scientists, modelers, market researchers, and other analysts.
- SPSS is used and taught in many schools, especially in the social sciences.
Feature Strength and JMP Response
- IBM can offer faster performance by adding their Server package. JMP can be a full client to SAS on a server. Also, with increasingly more compute power available on PCs as well as the low price of memory, it is reasonably easy to equip key users with powerful hardware, which bridges the gap between what is possible on a desktop and what tasks require the use of server-based software.
- With the new Chart Builder, they now provide a great deal more interactivity as well as some chart types JMP does not provide (e.g., relationship chart) and they have the ability to easily create dashboards similar to JMP. However, JMP is still ahead in many of the innovations and enhancements to Graph Builder.
- SPSS Modeller has support vector machines and now linear SVMs, a classification algorithm now also available in JMP Pro 15.
- Modeler also has XGBoost (Extreme Gradient Boosting) as a Python node. The ability to use a tuning design table and the addition of stochastic gradient descent was added to the partition platform (Bootstrap Forest and Boosted Tree) in JMP Pro 13, making it a more powerful machine-learning method.
- SPSS Modeler Premium includes more text analytics, including entity analytics, as well as social network analysis. While JMP doesn’t have entity analytics or social network analysis, SAS does and there are many open source packages, some very JMP-like such as NodeXL.
- SPSS Modeler streams are .xml files and can generate PMML for other applications to consume models. While JMP does not generate PMML, it does generate SAS scoring code. JMP Pro generates Python, javascript, sql, and C score code in the Formula Depot platform—this is in response to scientists and engineers needing to deploy models in their production systems. Note that unlike SAS, JMP does not address the derived or transformed variables often occurring prior to modelling in the score code that is produced. It also doesn’t deal with other model lifecycle management issues, but SAS does offer Model Manager for these needs.
Feature Weaknesses and JMP Response
- SPSS Modeler does not have partial least squares, however, in IBM SPSS Statistics, it is included as a Python extension. PLS in JMP Pro is a versatile method that should be positioned as a must-have in this era of ever-wider data and we don’t require an additional product to have this method.
- SPSS Modeler does not have as many ways to prepare the data as straight-forward or visually intuitive as they are in JMP, especially with enhanced data exploration capabilities in JMP 15 to see what additional data prep may be useful to do more quickly.
- If a model breaks, it’s broken even though it’s xml and will have to be recreated.
- If IBM SPSS Modeler users also want time series modeling, they have basic time series modeling techniques; JMP 15’s new time series platform adds yet more functionality with more modern forecasting methods (note: the help for this new platform will be in JMP 15.1).
- Is it as easy now to create formula columns, derived columns, script model creation, collect models in a container? Profile models? Compare models? These are strengths of JMP Pro vs. SPSS Modeler.
Opportunities for JMP
- Compete on price, superior customer support, service, and training as well as more dynamic interactive visualization.
- Even with base JMP, you get more functionality. Depending on what the prospect is using or is looking to use, there may be opportunities to sell JMP Pro and still compete on price with just one product versus possibly needing several modules from IBM.
- Consider expanding to other groups to reveal potential for software rationalization. If SPSS is used in one group, Salford Systems, Angoss, Reliasoft, etc. in other groups, the 80-20 rule of getting them most of what they need may work in our favor.
- As of JMP 14, we provided support for Chinese in Text Explorer; IBM SPSS Modeler doesn’t support Chinese (they still require translation into English first).
- JMP Pro 15 has structural equation models; to get the near-equivalent from IBM, customers would have to license IBM SPSS Amos. SEM can be useful in consumer research, so we may have opportunities to position one product versus two or more from IBM.
- With JMP Pro 15’s enhanced Functional Data Explorer Platform, scientists and engineers can more easily make sense of streaming data, often in special forms posing unique challenges to wrangle the data, align curves, build surrogate models and create features. IBM SPSS Modeler would require IBM InfoSphere Streams on a linux machine and other components to be configured before doing anything with streaming data.
- More and varied algorithms with richer exploratory data analysis capabilities in JMP to derive more value from data.
Threats for JMP
- IBM can discount to the point of giving the software away for free. IBM can bundle products (not just analytical modules, but also Cognos and Netezza, as well as services and hardware). They can say the software is “free” and make their margins on services and/or hardware. When they use this strategy, they are often selling more to IT.
- To the best of our knowledge, IBM puts no restrictions on the use of their software for consulting or using with third-party data as SAS does. This may make it more attractive for consultants.
Pricing
- Pricing for Professional/Premium/Gold is not listed on IBM’s web site, however, g2 shows Personal starting at $4,950, Professional starting at $7,430, and Premium starting at $12,300. (As of April 2016, IBM SPSS Modeler Professional was listed with a starting price of $6,800; SPSS Modeler Personal, $4,530; and Premium, $11,300.)
- Student and faculty can obtain a 12-month license for SPSS Modeler from com for free. IBM SPSS Modeler Premium GradPack is available for $89.99
- According to their web site as of August 2019, monthly subscriptions start at $199.
Competitive references where we’ve made inroads, won against them or displaced them.
Frei University
University St. Gallen
Indiana University
