There have been various changes within IBMâ€™s Data Science & Machine Learning product portfolio over the last months. This has also brought some confusion around the direction of IBM SPSS Modeler, so I wanted to clarify a few things. If you read no further, I want you to know that IBM SPSS Modeler is NOT going away. In fact, we have significantly invested in SPSS Modeler over the last 12 months and we see this as a very important part of our product portfolio for years to come.
You will also be able to access the classic SPSS Modeler standalone version and its enhancements for many more years to come as we continue investing in it. The short version of the story is a ‘new’ SPSS Modeler will be a part of Watson Studio. We will have three ways to access the â€˜newâ€™ SPSS Modeler: Watson Studio on Public Cloud, Private Cloud and Desktop
We believe having SPSS Modeler in Watson Studio allows us to deliver newer capabilities and ways of working. We also know that we will need to keep supporting our customers with the classic Modeler. We canâ€™t pry the classic SPSS Modeler away from some of you. You know who you are, and we thank you for it!
You may ask whatâ€™s driving our product decisions and what will this â€˜newâ€™ SPSS Modeler become? The product team has looked to historical design principles to guide product direction. In order to understand our design principles, you have to understand a bit of Modeler history.
A brief history of SPSS Modelerâ€¦
There is an interesting story about SPSS Modeler. It was a SPSS acquisition of ISL (Integral Solutions, Ltd) for its product, Clementine (a.k.a. Modeler), in 1998. ISL was a technology enabled service consultancy using its initial product, in 1989, Poplog AI Development Environment, which was a coding intensive platform. Back then, machine learning techniques were used with Poplog for applications such as retail turnover prediction, TV audience prediction and financial sector customer profiling.
Analyzing clients’ data in the Poplog environment meant writing routines (in POP-11) to read the data, perform various manipulations on it, perform exploratory operations such as graph plotting, convert the data into a suitable form for the machine learning systems, apply the machine learning systems, apply the resultant rules or neural nets to test data, and analyze the results. By far, the largest part of the work that went into this was coding, with only a small proportion being focused on the patterns in the data and their interpretation. There was some code re-use within each project, and a lesser amount between projects, but essentially each routine was highly customized to the data involved and the specific task.
As these projects succeeded one another, it became clear that ISL was performing the same coding tasks repeatedly. Then, in 1994, Clementine was born. The initial design combined re-useable versions of the modules ISL had developed for specific projects with a visual programming interface that made it extremely easy to “plug together” these modules to form a data mining process.
One key theme characterized the design of Clementine: drawing attention away from technology and towards the data.
Two basic design principles of Clementine were:
â€¢ Details of data mining and machine learning techniques are hidden unless requested
â€¢ The different techniques within Clementine are highly integrated so users do not need to get distracted by moving between them (visualization, reporting, queries, transformation)
These design principles apply just as much to the data science or machine learning expert as they do to the non-technologist. If the task is to understand patterns in data, it matters little that the user is knowledgeable about the technology used for modelling. This knowledge may be useful when understanding the detailed behavior of the technology, but this is seldom necessary, and then only a small part of the task. If the tools bring such technical issues to the fore, they draw attention away from the main business of understanding the patterns in the data and the reality they represent.
Changes in IBMâ€™s Data Science & Machine Learning Portfolio
Over the last decade weâ€™ve seen a lot of changes in the data science space. This simply relates to the proliferation of data, where it lives, moving it, governing it and computing it. We are increasingly seeing enterprises struggle to keep up with ever increasing demands to extract value out of data. Additionally, and excitingly, open source languages and tools have become part of the ecosystem. Whereas, historically, companies such as SPSS and SAS had to provide proprietary solutions across the full analytics lifecycle because they had to control end to end degrees of freedom, open source languages and tools have decoupled the analytics lifecycle, broken it into parts, and have become trusted. Now, there are specific companies leveraging open source as integration points to focus on data preparation, feature engineering, modeling and deployment as well as a proliferation of open source languages, libraries and tools from an amazing ecosystem.
We are excited because this allows IBM to leverage open source ecosystem with our existing offerings to give our customers the flexibility they need to extract value from their data they ways they want to. These are some of the reasons we have created a data science platform, Watson Studio, with not only IBM product integrations but also open source tools and partner products.
What does this mean for SPSS Modeler?
In short: Nothing. We are sticking to the original design principles of Clementine (Modeler): drawing attention away from technology and towards the data to drive the product roadmap decisions. Excitingly, over the last twelve months we have been recreating the existing SPSS Modeler with newer technologies as well as giving it a nice new facelift by changing the user interface. Weâ€™ve used our award-winning IBM Design team to help us on this path. This â€˜Newâ€™ SPSS Modeler is going to be the visual modeling tool within our Watson Studio offerings. We are on the path to feature parity between the existing SPSS Modeler and this â€˜Newâ€™ SPSS Modeler and will also be modernizing it further to leverage the data science ecosystem. We will constantly ask ourselves â€˜If ISL were to build a new Clementine today, what would it be?â€™
Our No Modeler Left Behind Policy:
We understand that a large portion of our existing customer base are loyal fans of SPSS Modeler. We thank you for it! We still love it and use it too! We promise to you that you can keep using the SPSS Modeler you know for as long as you like. In fact, we will still be investing in it over the years.
For those of you who think you will like to move or try the newer version SPSS Modeler, it will be housed as a data science tool within our Watson Studio offerings, but still called SPSS Modeler. See the blog post here.
If you want to move to any of the Watson Studio offerings you can still run your existing streams.
We hope you stay along for this journey with us. We would love to hear your feedback and stay tuned for more updates!