Sales Talk Track for Statistics, Predictive Modeling, and Data Mining

anne_milley

Statistics, Data Mining, Predictive Analytics

TALK TRACK

Statistics, Data Mining, Predictive Analytics with JMP® 15

 

 

AUDIENCE:Anyone who needs to explore and analyze data to gain deeper insights, to deal with data more effectively, or to make more informed, data-driven decisions.

CRITICAL BUSINESS ISSUE:  Be efficient and effective in the face of constant change: —predicting outcomes of interest, testing hypotheses, and better understanding relationships to help you make better decisions with real-world data—missing values, outliers, correlated inputs, etc.

GOAL:Creating value, and facilitating innovation.  Using data to better understand your organization, better anticipate the future, and be ready to respond in ways that minimizes waste, maximizes returns, and mitigates risk to be efficient and effective. 

 

1. Defining the Problem

Customer Issue Description

·   Generate value. Find ways to create more value through data analysis, to ferret out waste, and to identify good opportunities. ·   Manage uncertainty. Make better decisions in the face of constant change. ·   Manage risk. Understand how to make changes that can positively affect the outcomes we care about. Learn more about what we don’t know; today’s noise may contain tomorrow’s signal.  ·   Continuously improve and learn. Keep learning and making progress in the face of uncertainty. Expedite cycles of learning and determine ways to tease out more signal from noise. (Not an exact science, but improvements and discovery are intimately linked; both require a synthesis of discipline, domain knowledge, good software, etc.) ·   Reveal opportunity. With increasingly complex data, finding the best opportunities is more challenging. More efficient strategies are needed to search for new opportunities. More complex data also means that more strategies and compromises must be considered. It becomes imperative to home in on the best opportunities.

 

Business Impacts of Not Addressing the Issue

·   Inefficient use of resources, resulting in unnecessary waste and increased costs. ·   Suboptimal decisions, resulting in missed opportunities, lower profits or higher costs. ·   Lost market share to competitors who are better at analyzing their data and making more informed decisions. ·   An increased likelihood that things won’t turn out the way you expect, which may put you out of business. ·   Unnecessary risks, false alarms and missed opportunities, missing important driver of a response. By realizing there is a better way, you could avoid wasting time and resources. ·   Lack of confidence that, as a practitioner, you’ve controlled modeling errors in the best way possible; greater risk of being blamed if things don’t go well.

 

External Validation of the Business Impact

“Today it’s no longer possible to work with only one test.” Instead, the company must correlate data for multivariate experimentation. “This kind of analytics requires advanced statistics and methods,” she says – capabilities available in JMP Pro. Corrine Berges, PhD, NXP Semiconductors  “I continue to see steady progress in advancing JMP’s multivariate platform. The percent contribution plot in the principal component platform is a great addition.” Cy Wegman, President and Founder, SY64 LLC

 

 

Challenges/Barriers to Addressing the Issue

·   Data availability. Organizations need to determine what’s worth looking at and what can be ignored. They can then decide if they are measuring what really matters.  Progress is unlikely if you don’t have the right data.  ·   Perceived difficulty/lack of knowledge. Some organizations may need to hire more analytically savvy staff, invest in training, or bring in a consultant. ·   Culture of the organization. Some organizations don’t invest in being data-driven; their decisions aren’t evidence-based. ·   Unsure how to show value. We may have to help them see the art of the possible.

 

1. JMP’s Solution to the Problem

Our Approach and Value Proposition

·   Cost of ownership. JMP provides a comprehensive repertoire of tools for quickly and easily getting value from your data. JMP supports in-depth analysis in many areas, yet at a low cost of ownership. ·   Easy to use, yet powerful. JMP provides interactive visual data exploration to more advanced – but still easy-to-use – analysis, including experimental design, trees, neural nets, generalized regression, PLS, etc.  ·   Dynamic visualization. Competitors don’t have the graphical capabilities that JMP has, which allow you to interactively explore your data so you can find what’s important faster. ·   Statistical discovery designed for humans. The world is full of software. Software by its nature is flexible, but using software designed for specific purposes will make you more productive. JMP starts with the user to make you be as good as you can be, from visual exploration to modeling. JMP is crafted to help humans, not replace them. Algorithms are commodities; with code and time, anything is possible. But JMP provides the power of algorithms in a form more people can use more easily.  ·   JMP is complimentary and synergistic. JMP connects with other popular tools (SAS, R, Matlab, Python, Excel) and provides most of what you need in a visual way. JMP has helped many people become more successful; we can do the same for you.

 

Differentiating Capabilities

·   Dynamic visualization. Our competitors are mostly static. JMP’s dynamic visualization is coupled with statistics, making your discovery process better and giving you greater confidence that your modeling is well-grounded. ·   Cost of ownership. JMP is less expensive than most of our competitors’ products. We can support the most-advanced to least-advanced user.  ·   Easy to use, yet powerful. JMP is very easy to use to study and display information, but we also have advanced capabilities, such as experimental designs, partitioning, trade-off analysis, etc.  ·   Comprehensive. JMP has a wide variety of analysis methods; using multiple methods is powerful. As a practitioner, we allow you to have more confidence that you've struck a balance with modeling errors. Within any one modeling technique, JMP provides the right way to make use of the data while balancing the risks. ·   Design. —Everything that’s done in JMP can be done with pen and paper.  Software is important.  JMP is as much about the way we do things as what we do—it’s the interplay between user and software that generates value. JMP is designed to make the user more productive.  ·   Extensibility. You can use JMP as it is and be productive, but you can also make a custom JMP application and tailor it to users’ needs. We provide a greater level of customization and automation compared to any other software (even though they may appear equivalent on the surface). JMP is integrated with SAS, R, Matlab, Python, and Excel, so you can enjoy visual discovery in JMP along with other tools already in your analytic workbench.  In addition, JMP now offers JMP Live allowing easier results sharing and collaboration.

 

Case Study

NXP Semiconductors

Challenge:  Design, test and manufacture new semiconductor solutions that can power a rapidly changing automotive industry.

Solution: Use advanced analytics in JMP Pro to transform chip testing and manufacturing processes.

Results: JMP Pro tangibly improves solution design and manufacturing, Six Sigma training and other process-improvement efforts that help keep NXP at the forefront of the market.

 

JMP® 15

·   Explore patterns utility to detect suspicious patterns in data. ·   Bootstrap is available for Partition, Bootstrap Forest, and Boosted trees. ·   Process screening enhancements. ·   New time series forecasting platform.

 

JMP® Pro 15

·   New Support Vector machine algorithm, another machine-learning method for classification. ·   Structural Equation Modeling to analyse structural relationships and assess often unobservable latent constructs. ·   K Nearest Neighbor for nonparametric classification and regression. ·   Functional Data Explorer enhancements. ·   Generalized regression enhancements.

 

JMP Live

If they need to share information in an interactive way with others that scales (not in Powerpoint, Excel, emails, etc.), make JMP reports available to non-JMP users, publish reports that are created automatically, have a repository for project data and reports, deploy dashboards, or create a shared space to foster collaboration, JMP Live may be an option.

 

1. Sales Cycle Guidance

Prospecting Criteria
Value creation:
·   Organizations that seek efficiencies in their processes and in their products and services. ·   Organizations that seek to learn more quickly from data to reveal opportunities. Stay (or become) more innovative: ·   Organizations that want to be more competitive/innovative with the products and services they offer and how they do it. ·   Individuals doing research who want to gain better and faster insights.

 

Conversation Starters

·   What functional group are you in and how does your group fit in your organization? (Determine the relevance of analytics to the organization and to the prospect as an individual.) ·   What is your role? How do you fit in? What’s your background? (Get prospects to talk about their products or services even if you are already familiar with their product portfolio or their role in the organization.) ·   What kinds of things are you trying to understand better? ·   When the rubber hits the road, what’s really important? What are people in your organization paying attention to and why?  ·   Tell me a little about the kind of data you have. ·   How do you identify trends and patterns? What outcomes are you trying to predict? ·   How important is it to quantify uncertainty? If you’re faced with a decision between doing two things (for example, either investing in a building or hiring someone else), how do you determine what to do?  ·   How do you know you’re staying ahead of the competition? How do you assess how well you’re doing?  ·   What tools do you use? Who are the primary users of the current tools?

 

Sales Cycle Tips

·   The prospect must have some affinity for numbers. All depends on data and domain expertise. We can’t be experts in everything, but we can represent the potential and work in partnership to see what it might add up to.  ·   Get a picture of how they fit into their functional group and how their group fits into the business. How do they approach problem solving? Are they in a group that does real work or in a support group that is perhaps undervalued? If the latter, help them realize that the way they communicate results (dynamically/visually) can have a greater impact. We can help them with visualization so that others can see that they’re doing a great job. ·   If they are looking at a number (e.g., inflation rate, score for a class), the number or single point estimate is bound to be wrong. Uncertainty is intrinsic to this idea. A more sensible answer is a range. Uncertainty is everywhere, such as a competitor’s reaction to your new program, so there will always be unknowns.    ·   If the goal is data-driven decisions, the prospect has to think about whole picture. If they write code themselves, how long does it take? If their time is free, then they’re not valued. If they invest in people, software and training, they will get more back than they put in.  ·   Fat data (more columns) means greater complexity, which makes it harder to reveal opportunities. More robust tools and dynamic visualization can help. ·   The tone should be along the lines of “you need to be thinking about this,” nothing too assertive. Quickly screen out people who are not interested and get them out of the way or move on to another opportunity. ·   Justify JMP Pro with specific reasons that are contextualized to their business.

 

 

1. Capabilities Map

Expanded Discussion of JMP^®Capabilities and the Value Created/Vision

Capability Description	Diagnostic Questions	So What? Value/Impact
General, as pertains to predictive modeling methods (regression, trees, neural, SVM, PLS, etc)	·   Do you need a variety of methods to see what works best with your data? ·   Could you benefit from more modern, automated methods in modeling? ·   Can you easily see which models perform best? ·   Do you need to see which variables are most important to outcomes?	·   JMP has implemented powerful methods that are suited for users at all levels.  Pro adds more advanced analytics. ·   JMP provides options to automate parts of the modeling process: variable selection, missing values, automatic stopping criteria, etc. ·   JMP Pro lets you easily compare competing models to assess interpretability and overall best fit. ·   Validation for honest assessment of model fits in JMP Pro.
Regression	·   Do you have data that requires different kinds of regression methods?	·   JMP provides a comprehensive set of manual and automated modeling methods.  ·   JMP Pro provides mixed models, penalized and quantile regression.
Categorical data	·   Do you work with response data?	·   JMP provides tables, summaries and tests for such data.
Trees	·   Do you need models that are easy to understand and explain?	·   JMP provides the ability to build straightforward decision trees automatically either with a single pass through the data or interactively. ·   JMP Pro provides more (such as bootstrap forests and boosted trees) and incorporates missing values.
Neural networks and Support Vector Machines	·   Do you need flexible modeling tools that predict well and reveal any nonlinear relationships?	·   JMP’s neural nets can have one hidden layer.  ·   JMP Pro automatically handles missing values; it also incorporates early stopping and boosting, has multiple activation functions and allows for up to two hidden layers. ·   JMP Pro SVMs allow another means of doing classification and regresssion
Structural Equation Modeling	·   Do you need to examine relationships between measured variables and latent constructs?	·   JMP Pro SEM prvides a flexible modeling framework that allows you to specifiy a variety of statistical models, including models where variables can be inputs and outputs simultaneously.
Multivariate interdependence techniques	·   Do you need ways to identify correlations, outliers or missing values? ·   Do you have many variables? Do you need to reduce dimensionality?	·   JMP provides principal components analysis, factor analysis, clustering, normal mixtures and self-organizing maps, MCA so you can see issues with the data and determine your next steps.
Multivariate dependence techniques	·   Do you need to determine relationships between responses and effects?	·   JMP provides partial least squares, which is handy as a general-purpose predictive modeling method – especially if you have more columns than rows. JMP Pro has advanced PLS, PLS-DA and imputes missing values and supports categorical response. ·   JMP also provides discriminant analysis to help you see which combinations of inputs can help determine group membership in the outcome of interest.
Funtional Data Explorer	·   Do you need to analyze more data streaming from sensors?	·   JMP Pro simplifies the clean-up so you can gain insights faster from sensor streams and lets you output features to build predictive models.
Time series / Forecasting	·   Do you need a better understanding of your time series data? ·   Do you need to forecast sales, units...? ·   Do you need to decompose time series to see trend, cyclical and irregular components?	·   JMP allows you to visualize, model and forecast time series so you can see patterns in the data and better anticipate the future.