Leveraging JMP to Improve Business Decisions

Derek Wilson, Director Business Intelligence and Analytics, Just Energy

Derek Wilson, Director Business Intelligence and Analytics, Just Energy

Businesses looking to gain new insights and optimize business processes can leverage JMP from SAS. JMP is interactive statistical software that enables you to explore your data using various techniques. n order to turn your data into actionable information you need to understand your data. This includes data quality and completeness of the data available in your organization. Transforming your organization to a data driven culture starts with a fundamental understanding of your data. One method to driving data mining insights is to leverage Cross Industry Standard Process for Data Mining otherwise known as CRISP-DM. This process consists of six phases that enable you document and explore your data mining question.

Business Understanding

In this phase, you are focused on understanding the true business problem. This includes determining the objectives and requirements of the data mining problem you are trying to understand. During this phase, you will gather requirements from your business partner. Including documenting and learning about analysis that may have already been done. In addition, you will determine potential data that you can leverage for your analysis. This may be internal data assets that are easily available or external data that may need to be obtained.

  Transforming your organization to a data driven culture starts with a fundamental understanding of your data 

Data Understanding

Once you have a defined project and initial list of require data elements you can begin to collect and analyze your data. During this phase, you want to understand the range and quality of your data. You can leverage JMP analyze the data and quickly create histograms and perform outlier analysis. Both of these enable you to visually see data that out of ranges or skewing your data. You can leverage the Scatterplot Matrix to graphically display the correlation between many elements at once. In addition, you can build a Treemap graph to display the relative size of data to the total data. All of these abilities allow you to gain a comprehensive look at the data you are using for your analysis.

Data Preparation

After you have your data selected and removed missing values, bad values, or outliers. You then transform and build the data into the sets required to begin analyzing your projects objectives. Some activities during this phase are cleaning up poor data, such as commonly misspelled words or words with extra spaces or hyphens. Users leverage JMP will be able to access a wide variety of functions that enable data transformations and cleanup. Another common activity is to transform columns or values as required. For example, changing columns with the terms Active and Inactive to integers such as 1 or 0. Focus on the critical data elements that your model will need and ensure you understand the raw data meanings and document any transformations you make in the process.


Most of your project time will be spent in the data understanding and data preparation phases. Only after you have a solid grasp of the available data can you begin to apply the data to various models. Users of JMP will be able to build supervised and unsupervised data mining models against their data. Depending upon the project you may need to build model to predict future outcomes such as when a part is likely to fail, or when a customer is likely to churn. In addition, you can build models that classify data into groups to perform segmentation analysis.


The evaluation phase consists of two primary goals. Review of the current process and testing the performance of the model. First, review the steps that lead to the creation of the model to the business objective. If you excluded certain data from the model, is it still appropriate? Were transformations performed on the data that skewed the data from the original population. Do the results make business sense? For instance, in a predictive model for customer churn. Is the model indicating that high bills are a significant indicator of the model? Next you will want to test the performance of the model. JMP includes the ability to perform these evaluations on your models to see how they are performing. You can perform evaluations to measure models in a variety of ways such as Lift, Gain or ROC Curves. These tests see how well a predicted value matches the actual value.


In the deployment phase, take your best models and present your findings to your business users. You will want to clearly explain how you achieved the results of your experiment. Starting from the basics ensure you list what data was included in the project. Likewise, if you excluded any data or made transformations to any columns show documents those as part of the project. While working in JMP any step in the process can be saved as a script and saved to a Journal. The scripts enable you or others reproduce the steps in your analysis. Business users will be more accepting of models if they understand the data inputs and the steps taken to produce the results. Even if they do not understand the algorithms, the closer you can get them to the data the better. Once reaching agreement with the business on the value of the model you move you model to production.

The CRISP-DM model can be applied to many use cases from basic data profiling to building advanced data mining solutions. Data mining professionals that utilize JMP can leverage it for any phase in the process. Having the ability to perform all of these phases in a single tool enables faster delivery of data projects.

Read Also

Need for an Enterprise-level Data Dictionary

Chad Benson, President, CEO & Board Member, CBE Companies

Translating Technology into Innovation

Shawn Shell, VP of Consulting, Hitachi Consulting

Evolution of IT in Healthcare

Mark Caron, CEO, Geneia

The Rubik's Cube of Business Intelligence

John Lin, Director of Business Intelligence, KVH Industries