This is INTERNAL USE ONLY. Please do not share any part of this document outside the members of the division. This is still subject to change as we get closer to the JMP 13 launch.
JMP® 13 and JMP® Pro 13 Marketing Features and Business Pains Addressed
Theme of JMP 13: Version 12 did pictures, Version 13 does text.
JMP
1. Text Explorer
2. Virtual Join
3. Dashboard Builder
4. Compare Designs
5. JMP Query Builder
6. Projects Defer to 14
7. HTML5 Improvements
8. Graph Builder general improvements
9. EZDOE Defer to 14
10. Export Options
11. Multi-dimensional Scaling (MDS)
12. Latent Class Analysis LCA
13. MaxDiff (Best/Worst) + MaxDiff Design Creation
14. Support for No Choice
15. Non-normal capability analysis
16. Multiple system reliability growth
17. Cumulative Damage
18. DOE Simulate Responses Refresh
19. JSL Additions
20. Recurrence analysis new features?
21. Process Screening
JMP PRO
- Text Explorer Analytics
- Generalized Regression improvements
- Repairable Systems Simulation (RSS)
- Gaussian Process fitting of larger problems
- Mixed Models, additional covariance structures
- General simulation functionality
- Association Analysis
- Stochastic Gradient Boosting
- Hierarchical Bayes for choice models.
- Formula Depot + Output SQL/SAS/C Scoring Code for JMP Pro Models
iPad Graph Builder
———————————————————————————————————————
JMP:
I. Text Explorer (Training, SE, Web) — In good shape, still active development
A. Description: A data mining technique where NLP (Natural Language Processing) algorithms can search a Corpus (collection of documents) to identify facts and relationships from a mass of “big data”. Tool for organizing, triaging and dealing with text data.
B. Audience: Market Researchers, Warranty Engineers, Medical Professionals, Anyone who may be subject to an audit. Engineers and scientists who have free text data e.g. engineering changes, repair notes, production notes.
C. Key Business Pains Addressed: In previous versions of JMP, very little if anything could be done with text data. You could do simple word-counts in categorical, or recode them into categories, but JMP lacked the ability to triage text data in a meaningful way and understand the latent information available in free text data that many of our Scientists and Engineers have access to. Provide methods for basic key word extraction.
D. Customer(s) asking for it: P&G, Adsurgo, Eli Lilly, Micron
————
II. Virtual Joins (Xan, SS?) — Good shape
A. Description: Reference other data tables without having to actually join it. Link ID is the column that encapsulate the data set. Link reference is a foreign key reference. Enrich your data tables with metadata without burdening it.
B. Audience: General
C. Key Business Pains Addressed: Data tables have information that are linked normally would require you to join tables. When data sets are long joining to wide (especially at different time sampling scales) there is a lot of duplication in the long data table potentially running into burdensome data storage issues.
D. Customer(s) asking for it: CAB, Tom Traynor (Zymergen), Jon W
————
III. Dashboard Builder (Xan; Training, Webcast) — Good, almost done.
A. Description: File -> New Dashboard is a single click to aggregate JMP reports into a presentation-ready dashboard. Also surface selection filters in a single click to use a Graph to filter another graph. New front page that lets you see templates and starting points for building an application. Click Drop reports to dashboard drop zones like Graph Builder. Bigger, more visible drop zones for building dashboards. At runtime, also reorganize reports in the dashboard, also create tabs. Save as interactive HTML preserves layout. If there is an embedded query it refreshes the database. Brings the Query in, and then recreates the dashboard.
B. Audience: Scientists and Engineers as creators. C-level and others as consumers.
C. Key Business Pains Addressed: Dashboards are difficult in previous version of JMP, requiring too many clicks. Application Builder (full) is too complex, too scary and has too many options. New JMP users are intimated by things like vlistboxes and other internal JMP layout items. Previous versions of Application builder created a lot of windows, creating a “wake” of windows behind it. Preview mode takes care of this, allowing you to go from view to edit without creating more windows.
D. Customer(s) asking for it: Unilever, GE, John Salmon
E. Resources:
1. https://sas.webex.com/sas/lsr.php?RCID=1b789984e7a8406f938210519aa53981
————
IV. Design Comparisons (DOE) — Launch of compare is tricky (needs multiple data table, C++ is done.
A. Description: Compare 2 or more designs. Produce a number of diagnostics that compare designs based on predictive capability. Prediction variance, amount of aliasing, lots of dimensions to compare. Make a design using one design optimality d- vs. I-, DOE compare tool will look at the two designs and gives you comparisons. 6-factor, 12-run design i-optimal vs. alias-optimal
B. Audience: DOE practitioners.
C. Key Business Pains Addressed: In previous versions of JMP it was difficult to compare designs easily especially with more than 2 designs are considers; it was a manual process requiring copy/paste or looking at multiple evaluate designs, now everything is on the same report.
D. Customer(s) asking for it: CAB
————
V. JMP Query Builder (Brian, SS, Webcast) — Good shape
A. Description: Multi-table Query/Join capability. Same interface as SQL Query Builder and SAS Query Builder. Up to 60 table Join. (Internal) Uses open-source SQLite; virtual table to SQLite. Use similar interface to what we have. Incorporate SQL as a query language (Base SQL) in JSL; Later converge Query Builder with upcoming (14) multi-table join.
B. Audience: JMP users
C. Key Business Pains Addressed: Sometimes you don’t want to create a database just to join data that is already in JMP. Joining more than 2 tables in JMP was a chore.
D. Customer(s) asking for it: CAB
————
VI. Projects (Xan; Training Needed, Webcast) — UI in good shape, buggy. Still on unstable ground conceptually, but closer now after 11/16/2015/ DEFER TO 14
A. Description: A project is a way of organizing analyses, windows, scripts, files and other artifacts using JMP in JMP. Projects can include zero or more files (from anywhere on disc) and zero or more live folders (also from anywhere on disc). Including a folder in a project will include all current and future files from that folder in the project similar to how the (single) project folder works today. The project file itself (i.e., the .jmpprj file) can live in one of the project folders or outside it. File->New Project will immediately open an unsaved project, rather than the new project dialog it shows today; this makes it easier to create temporary projects and for us to implement the “create project from all open windows” feature.
B. Audience: New JMP users, JMP users who missed the master window. JMP users that have window management pains.
C. Key Business Pains Addressed: The project serves many of the same roles we used to have with the master window on Windows, providing a container rather than having all the views scattered around. Shared menu bar. When analyses are inside the project window, there is already a fully formed menu bar, without the need to have a menu bar for each data table and analysis. The menu bar is always visible and available and not duplicated. Tolerance to Maximized windows - for users that routinely use Maximized Windows, it doesn’t hurt anymore in allowing you to see multiple views, as long as the user shows multiple tabs simultaneously. Border list of files - the list of available files is always available without having to go back to the home window or an operating system window. Saved State - a principal benefit is the natural ability to bring back a session to a saved state for that project, with the ability to have many session saved states for various projects. Presentation Geometry - rather than overlapping windows, projects disciplines the reports to be in either a tiled space, or in tabs. Default save folder - when a project is open, it is easier to save files because there is a clear default folder to save it to. Not all these benefits are maintained when we make projects different from the files in a folder.
D. Customer(s) asking for it: Starbucks, Roche, Abbott, Abbott Vascular, Gore, Intel, P&G, Micro, G3, Apple, Schott AG, Atotech, EM Microelectronic, Telefonica, Micron, Microsoft.
E. Resources:
1. http://sww.sas.com/jmpmediawiki/Project_Concepts
2. http://sww.sas.com/jmpmediawiki/Project_Concepts:_Overlapping_Features
3. http://sww.sas.com/jmpmediawiki/JMP_Projects_JMP13
4. http://sww.sas.com/jmpmediawiki/JMP_13_Projects_Issues_and_Features_List
5. http://sww.sas.com/jmpmediawiki/JMP_Projects_Future
————
VII. HTML 5 Improvements (Brian, SS, Webcast)
A. Description: Support for most of the segments in Graph Builder done state. Support for Dashboard layout from Application Builder tabs.
B. Audience: All
C. Key Business Pains Addressed: Graph Builder is the primary data visualization tool for communicating resulting in JMP; it previously was not interactive when saving as interactive HTML. Dashboards lost their layout when saving as HTML 5.
D. Customer(s) asking for it: General; CAB
————
VIII. EZDOE (DOE without tears) Defer to 14
A. Description: UI to design and work through a DOE workflow. Knowing what to do and how to use the modeling platforms are not easy for a novice. Beginning to end tabbed dialog from define factors -> create design -> table in the report -> analyze the report -> diagnostics etc.
B. Audience: Basic JMP users. Six Sigma Green Belts.
C. Key Business Pains Addressed: JMP is too difficult to learn for a novice who might just want to get into designing an experiment without having to deal with multiple tables.
D. Customer(s) asking for it: CAB, Bill Kappele
————
IX. General Improvements to Graph Builder (Xan, SS, Webcast)
A. Description: Graph Builder is the gold standard for creating graphs in JMP; Addition of Tree Map and Parallel Plot updates to GB; the goal is consolidation of all graphing platforms into GB; Connect through missing from Line; New transforms like Cum Sum; Customizable alpha level (maybe); Replace 3 buttons on top with min tool bar (like redo, filter; maybe); Additional dodging options for points (maybe); Addition of waterfall plots (maybe); Legends inside of plot (maybe); Multiple summary stats (maybe); More fitting in regression (exponential); Width of bars and box plots; customizations; Colored box plots (maybe); Center step lines (maybe); More robust undo (maybe); Shading on smoothing; Less likely: Tabulate-like output, drill-down support (tree-map), generally shortcut into Local Data Filter, customizations of Violin plot; Means Diamonds
B. Audience: All JMP users
C. Key Business Pains Addressed: Graph Builder has 700+ Open Defects. Every release we need to chip away at these defects while triaging new defects that come in. The listing above is that effort for JMP 13.
D. Customer(s) asking for it: N/A
————
X. Export Options – Excel Output and Generate Web Report (Brian, SS, Webcast)
A. Description: Take a number of JMP tables and output to a single Excel Workbook. Take a number of interactive HTML reports and aggregate into a single Web page.
B. Audience: Everyone trying to share results outside of JMP.
C. Key Business Pains Addresses: It’s difficult to put together a summary website of results and some times the single reports are out of context. One-click to generate a website. Sometimes you need to generate a workbook of JMP tables into Excel. Now with one-click you can do that.
D. Customer(s) asking for it: Abbott, Jon W.
————
XI. MDS (Chris, Training)
A. Description: Multidimensional Scaling or MDS is a popular multivariate statistical technique that is primarily visual. There is little formal inference, such as hypothesis testing, associated with MDS. The purpose of MDS is to create a visual representation of the pattern of proximities (i.e., similarities or distances) among a set of objects. For example, a matrix of perceived similarities between various brands of beer, MDS plots the beers on a map where those beers perceived to be similar are close together on the map, and those beers perceived to be dissimilar are far apart on the map. Nonlinear least squares. Shepard diagram – line will mean good, cloud means the story is not that simple. Beyond 2-3 d.
B. Audience: Sensory Analysts, Market Researchers, Sociologist, Biologist
C. Key Business Pains Addressed: Provide a method for customers to compare products and make judgments regarding the product similarities without the need to provide a list of product attributes to compare against such as methods used in conjoint, factor analysis etc.
D. Customer(s) asking for it: George Washington Univ., P&G, Kenan-Flagler, Univ. of Wisconsin, Stockholm Univ., Dokkyo Univ.
————
XII. LCA (Training)
A. Description: Latent Class Analysis (LCA) is a statistical method for identifying unmeasured class membership among subjects using categorical and/or continuous observed variables. For example, you may wish to categorize people based on their drinking behaviors (observations) into different types of drinkers (latent classes).
B. Audience: Market Research, Sociology, Psychology, Healthcare
C. Key Business Pains Addressed: Cluster analysis for categorical variables. Previously we only provided Factor Analysis for continuous.
D. Customer(s) asking for it: Walter Paczkowski, Rutgers University
————
XIII.Maximum Difference Analysis (Training)
A. Description: MaxDiff (sometimes known as “Best-Worst”) is a technique used to analyze customer preference. Respondents will make a choice, or trade-off. Similar to choice analysis, it is often considered easier to execute and more informative for certain studies. Respondents simply select the best and worst choice from a set of fixed set of products.
B. Audience: Market Researchers, Product Mangers, Advertising Managers
C. Key Business Pains Addressed: Frequently requested option offered by competitors such as Sawtooth. Max Diff is used a lot in industry and was missing from our Consumer Research offering.
D. Customer(s) asking for it: Walter Paczkowski, Data Analytics Corp., Isometric Solutions, P&G
————
XIV. No Choice (Training)
A. Description: Choice design and analysis will now offer a “No Choice” option. Market researchers can now offer respondents the option since often they may not like either choice offered in the choice set. No preference is the way to think about this. Show how you can bias in your analysis with previous choice models where no choice wasn’t considered.
B. Audience: Market Researchers
C. Key Business Pains Addressed: Commonly offered by competitors and frequently employed by market research professionals. If you are going to try and predict how popular a product will be, if you don't have a no choice option, your predictions could be wrong.
D. Customer(s) asking for it: Walter Paczkowski, Data Analytics, P&G
————
XV. Non-Normal Capability Analysis (SS)
A. Description: Capability Platform now offers support for non-normal distribution fitting and non-parametric estimation for capability indices.
B. Audience: Quality assurance analysts, Quality Engineers, Supplier Quality, Six Sigma professionals.
C. Key Business Pains Addressed: Multiple platforms required to perform capability in JMP, depending on the distribution type. Now all can be performed in one platform.
D. Customer(s) asking for it: Medtronic, Phil Ramsey, Tom Little
————
XVI. Reliability Growth – Multiple Systems (Training, Web?)
A. Description: Added two new tabs to support multiple systems growth analysis. 1) Concurrent Systems: When test are performed on many like systems and a failure occurs in one system, the design correction is applied to all of the systems under study. 2) Parallel Systems: Study and compare reliability growth on independent multiple systems by system and/or Phase. **Note: this may change by release date.
B. Audience: Reliability Engineers, test engineers, design engineers focused on repairable systems.
C. Key Business Pains Addressed: Users and prospective customers selecting competitor software to perform the required analysis.
D. Customer(s) asking for it: Chrysler, GE, Abbott
————
XVII. Cumulative Damage
A. Description: Analyze Time to Event data under non-constant stress situations. Perform accelerated life test under increasing stress and analyze field data under arbitrary stress scenarios. Four model options are available: Step Stress, Ramp Stress, Sinusoid and Piecewise Ramp Stress).
B. Audience: Design and manufacturing and test Engineers
C. Key Business Pains Addressed: Provides and alternative to ALT when engineers are limited in number of test units and test time.
D. Customer(s) asking for it: Dow Chemical, Keurig, Biotronik, Tessera NA, OLEDWorks llc, numerous requests during RAMS and Explorer events.
————
XVIII. Custom Design — Simulation (In good shape)
A. Description: Simulate responses from DOE custom design. DOE Simulate script in the data table to adjust simulate after the design has been created. Normal, Binomial and Poisson. N Trials for Binomial to simulate. Normal is usual, Binomial and Poisson for counts. Counts come in integers — when number of counts are small GLM is more appropriate for Poisson models. Way of getting at a power calculation for a complicated problem.
B. Audience: DoD: System tests and have a lot of system test where results of test are pass/fail. Always better if you come up with a continuous response, but if you can’t it would be nice to say: given this test, how big of the difference between the predicted value at the low vs high needs to be. T&E engineers.
C. Key Business Pains Addressed: In the past, simulate responses assumed a normal distribution of the Y. And the resulting simulate into the Y column was not a formula so it could not be adjusted after the effect. Very complicated for binomial models with lots of underlying effects. Power and sample size tools assume that you only have 1 x you are interested it, this is not true for real problems as the variance is non-linear. Figure out just how many extra tests you need to do to detect a small shift in probability e.g. difference between .95 and .99: needs lots of trials.
D. Customer(s) asking for it: DoD.
————
XIX. JSL Improvements
A. Description: Every release we make efforts to improve the experience for novice and advanced scripters to use JSL. Some new features are scripting-specific and some are architectural changes/additions that make life easier for both the programmer and general JMP user.
B. Audience: Scripters
C. Key Business Pains Addressed: Many times when JMP crashes you would lose work in a script or other JMP files because JMP does not have any auto-saving mechanisms. Now it does. JSON was not easily importable, now it is. It was difficult to compare difference in two scripts, now it is easy with File Compare. Toplists were very slow, they have optimized and now they are fast. Non-JSL languages were not syntax color-coded making them difficult to read, now there is styles for color coding the text editor for various languages. UsingNamespace (); function helps with the pain of wanting to run a block of code in a specific namespace. JSL did not have class, now it does: functions, methods (virtual); Support Inheritance; Polymorphism is not supported, unless done by argument counts; This and super are supports; Send messages to a class; Namespace code blocks.
D. Customers asking for it: General
———————————————————————————————————————
JMP Pro :
XX. Text Explorer Analytics
A. Description: Multivariate analysis (like PCA, LCA) for dealing with text data cleaned up and triaged by the Text Analysis platform. Deal with survey questions, engineer reports.
B. Audience: Anyone with text data looking to build predictive models from those data, perform multivariate analysis with text data.
C. Key Business Pains Addressed: Turns the problem of dealing with text data into a standard data analysis problem.
D. Customer(s) asking for it: Adsurgo, CAB, Applied Materials, FDA, Micron
————
XXI. Generalized Regression Improvements
A. Description: Additional features being added to GenReg. Ability to handle censoring; kind of models you would fit with parametric survival, you could fit with Gen Reg, with variable selection. Time to event model. ERIC — a validation techniques that will choose sparser models for most fitting problems, since it was tailor-made for the adaptive lasso unlike AICc and BIC. Cox regression model into Gen Reg framework. Interactive ROC curve adjustment. Change the model and look at the ROC training/validation.
B. Audience: Biostatisticians (Cox regression). Reliability engineers (Censored models)
C. Key Business Pains Addressed: Cox regression platform was old and needed updating; unable to fit censored data in Gen. Reg.; Models built with standard validation techniques are often too big, because AICc and BIC assume maximum likelihood.
D. Customer(s) asking for it:
————
XXII. Repairable Systems Simulation (RSS)
A. Description: Internal simulation engine that can use discrete event simulation. Primarily concerned with repairable. Things big enough to consider repair.
B. Audience: Anyone looking at simulating the repair of expensive stuff (turbines, aircraft engines).
C. Key Business Pains Addressed: Simulation to get results of how long something is going to go down, which is expensive. What should they fix while they are in there. Existing RBD only did time-to-event (single failure). Simulation solves a harder problem. Need to come up with a budget (forecast). How many repairable events should I expect in some time period.
D. Customer(s) asking for it: GE Energy, P&G, Rolls Royce, Fleet maintenance (school buses e.g.). Applied Materials.
————
XXIII. Gaussian Process fitting of Large Problems using Fit Mixed Framework
A. Description: Analyze much larger data sets easily. Handles blocking code. Ability to analyze categorical variables as input. A.K.A. surrogate models. Set up a block size; Allowing for Fast GASP.
B. Audience: People running computer simulation experiments.
C. Key Business Pains Addressed: It was slow to fit Gaussian process problems when they got bigger than, approx. 1000 records. If your simulation included categorical variables as inputs you could not analyze the problem.
D. Customer(s) asking for it: Gov’t, Semiconductor, Sandia, P&G, Pratt & Witney, GE Energy
————
XXIV. Mixed Models Additional Models
A. Description: Added a number of correlation functions to the repeated structure, Exchangeable, Interdependent Equal/Unequal, Unequal Variances,
B. Audience: People performing mixed model fitting especially in Ag and human studies.
C. Key Business Pains Addressed: Close the gaps vs. competitors who fit Mixed Models.
D. Customer(s) asking for it:
————
XXV. General Simulation Functionality
A. Description: Switch out any input column with the random number generator column. Power analysis for terms in the model. General Monte Carlo simulation. Also perform parametric bootstrapping. Save simulation formula for any distributions in GenReg. Run an experiment 100 times and see what the variability is across the experiments. For normal examples — the power and sample size cals in Custom Design tell you what you need to know; but for Binomial and Poission experiments, you need to simulate.
B. Audience: General; statisticians.
C. Key Business Pains Addressed: Need to perform Power Calculations, perhaps you have counts data. How well is your experiment is going to perform under certain assumptions?
D. Customer(s) asking for it: DoD, Army, Gov’t general, Researchers submitting grants.
————
XXVI. Association Analysis
A. Description: Market basic analysis, analyze which products occur together in a transaction and make recommendations. Generate from structured terms or items and want to see which ones occur together. Create associations rules with Market Basket platform. Generate table for association from the Text Explorer platform. Possible use for data mining of failure data; which failures occur together.
B. Audience: Market researchers. Marketers.
C. Key Business Pains Addressed: Checklist item for Market Research programs.
D. Customer(s) asking for it: Matt Flynn (Aetna), Rob Reul, P&G, Curt's professors who want to teach Market Research with JMP.
————
XXVII. Stochastic Gradient Boosting
A. Description: Performance improvements for boosted trees; combination of Random Forest and Gradient Boosting.
B. Audience: Predictive modelers and data miners.
C. Key Business Pains Addressed:
D. Customer(s) asking for it: Russ, Mia.
————
XXVIII. Hierarchical Bayes
A. Description: Random-effects model for choice analysis. Segmentation analysis of choice models. Different way of modeling subject effects and you can validate statistically how much of the heterogeneity is due to observable difference between subjects and how much is random variation.
B. Audience: Market researchers. You can also fit HB in Max Diff if you have the Pro version of JMP.
C. Key Business Pains Addressed: In previous version of JMP, we were unable to properly model heterogeneity in choice models resulting in imperfect predictions and overall lower quality models. Hierarchical Bayes is an industry standard techniques and is available in competitive software including Sawtooth, R and others.
D. Customer(s) asking for it: P&G, P&G Europe, Walt P (Industry thought leader). Anyone that is doing choice models will eventually want this.
————
XXIV. Formula Depot; Output SQL/SAS/C Scoring Code for JMP Pro Models.
A. Description: A hub for organizing and deploying models built in JMP Pro without burdening the data table with columns and columns of prediction formulas; Much more efficient way of storing JMP Pro models.
B. Audience: Predictive models; those needing to deploy models built in JMP Pro into other systems.
C. Key Business Pains Addressed: In previous versions of JMP, when building multiple models, the data table would be burdened with many extra columns of prediction formulas, which had to have been output from fitting platforms in order to make the Model Comparison platform work. Also all of the columns we named by the predicted response, which made it challenging to find out what model source the prediction formula came from. With large tables, this could lead to a large footprint due to these extra columns. Deployment was also difficult and JMP Pro lacked a single hub for organizing prediction formulas and selectively deploying them to code for other systems like SQL or C. Now the model center fixes all this.
D. Customer(s) asking for it: P&G, GE (GE Power), Starbucks, IMFT, Micron, Intel, Rob Reul.
———————————————————————————————————————
iPad Graph Builder App – now lets you import excel and text files. Also adds the ability to go into Box, Dropbox and iCloud storage (And other cloud-providers). Beta in Feb. 7.0 or 8.0 minimum iOS.
Misc JMP new features —
• New JMP Starter (Including a JMP Pro-specific starter) (SS)
• HDF5 import (SS): Raytheon, Johns Hopkins, General Dynamics, Navy (http://sww.sas.com/ds/DefectsSearch/S0780/S0780852.html)
