Table of contents
At a glance
How do you validate JMP or JMP Pro for analyses that affect quality?
Define the JMP edition, version, source data, analysis platform and intended use of results. Test row states, model roles, scripts and the relationship between the current data and report. Supplier reference tests can support numerical evidence but do not assess your analytical workflow. Statistical model validation and computer system validation address different questions. A reproducible script still needs appropriate data, justified settings and controlled records.
- Hide and exclude are different row states; their significance additionally depends on the analytical pathway.
- JMP Pro adds functionalities but does not replace the professional evaluation of the method.
- The data table, script, and visible report must correspond to the same relevant evaluation stage.
- Installation traceability, numerical reference calculation, and proof of the customer process have different purposes.
1. Define the specific intended use of the result
First, describe the decision that JMP supports. During process development, an interactive investigation, a predefined trend report, and a reviewed quality investigation may require different data pathways and controls. Exploration is not automatically outside the scope of validation: what remains critical is whether and how their data or results influence a GMP-relevant activity or decision.
This guide addresses human medicinal product GMP. The current EudraLex index identifies Annex 11 and Chapter 4 as relevant guidelines. The scope and depth of validation evidence are aligned with actual usage. Requirements from clinical studies, other product areas, or additional regulatory jurisdictions cannot be inferred from this alone. The software designation JMP alone does not establish this distinction.
A useful intended-use statement identifies the source, unit of observation, statistical question and recipient of the analysis. Define permitted interactive changes and who assesses their significance. If a one-off investigation becomes a routine report, further define the approved workflow and its limits.
2. JMP, JMP Pro, and related products differ
JMP Statistical Discovery LLC offers JMP and JMP Pro as distinct versions. JMP Pro expands the feature set with additional modeling methods and options. JMP Live for sharing analyses and JMP Clinical for clinical data evaluation are separate offerings. These should only be included in your system description if they are actually in use; similar product names do not imply shared functionality.
The manufacturer documentation index identifies JMP 19.2 as current as of October 6, 2026. This serves as a reference point for this guide, not as a statement regarding your installation. Record the actually deployed build, the operating system, and the JSL scripts and add-on components in use. Dates on individual help pages are not to be interpreted as release dates or evidence of an installed patch.
Supplier documentation may reference a different version. The numerical whitepaper evaluates JMP and JMP Pro 19 with a build dated 19 August 2025. This distinction remains visible upon transfer. Evidence of a specific version must not be considered conclusive proof of a later functional status solely due to a similarly sounding product name.
3. Define data roles and analysis platform jointly
An evaluation is not solely determined by its data table. Data type, selected variables, row states, and model roles influence which observations are actually utilized. Capture these characteristics in the evaluation request. This enables a reviewer to assess whether the results address the intended question, rather than merely comparing the visible metric.
The following internal workflow matrix illustrates why identical inputs do not always produce the same outcome. The validation examples in the validation column are derived from various help sections. Here, validation refers to the statistical evaluation of a model using appropriately segmented data. This represents a different purpose than the validation of a computerized system.
| Control point | Documented behavior | Definition for the procedure flow | Appropriate test validation |
|---|---|---|---|
| Edition | Certain validation column functions are described for JMP Pro | Actual edition and required option | License/Version reference and implemented functions |
| Standard Least Squares in JMP Pro | Four or more different values: validation column is ignored | Permissible values and expected data roles | Roll alignment and report on actual usage |
| Neural with validation column | More than three different values: K-Fold Validation | Selected procedure and intended division | Evidence of appropriate platform rules |
| Missing values | Standard Least Squares excludes affected rows by default; options may vary. | Missing Rules and Used Options | Input data and actually used observations |
| Report | Automatic Recalc is not supported by every platform. | Procedure following changes to the table | Current report with appropriate data/script status |
4. Do not confuse visibility with computational exclusion
JMP distinguishes between hiding and excluding. Hidden data points are not automatically removed from statistical calculations solely because they are hidden. Conversely, excluded observations are not automatically omitted from diagrams. Therefore, a cleaned-up graph does not prove which data points are included in the associated metric. Explicitly verify the data role, not merely the visual impression.
Additionally, the analysis platform may carry particular significance for excluded data. In Neural, the setting "Excluded Rows Holdback" uses non-excluded rows for training and excluded rows for validation. The statement "excluded means used everywhere" would therefore be incorrect. In the intended workflow, the platform, option, and role must collectively be traceable and logically consistent.
A striking observation does not justify automatic exclusion. Investigate its origin and significance using the appropriate investigation procedure. A measurement error, a differently defined population, and an actually unusual process result are distinct circumstances. Document relevant decisions and changes without classifying an exceptional value as fraud, out-of-specification (OOS), or invalid observation.
5. Example Calculation: Test Case with Independent Mean Calculation
The following example illustrates a simple mean report, not a predictive model. Five observations, A through E, have values of 10, 11, 12, 13, and 34 in the same unit. All initially belong to the predefined data set. The high value E is intentionally included; no claim is made that it is factually incorrect.
The variants are evaluated in a test environment starting from a documented initial state. Exclusion of E serves exclusively for the predefined functional verification. This does not constitute authorization to remove undesirable results from actual investigations. The computational basis, row state, and actually generated report are each recorded together.
| State | Underlying data | Independent specified value | Decision on testing |
|---|---|---|---|
| A–E inclusive and visible | 5 values; total 80 | 80 / 5 = 16 | Identify the source report unambiguously |
| E only hidden | Calculate additional 5 values; display without E | Furthermore 16 | Hiding must not be documented as a computational exclusion. |
| E excluded for simple report generation | 4 used values; total 46 | 46 / 4 = 11.5 | Demonstrate data ownership, test specifications, and new calculation methodology |
| Report not yet synchronized | Table updated, report may appear outdated. | 16 and 11.5 do not confuse | Check the recalculation behavior or warning functionality of the specific platform. |
| E is again visible and enclosed | Again 5 values; total 80 | Again 16 | Return to initial state with newly generated report documented |
6. Put the script and report on the same level
JMP can save the path to an analysis report as a JSL script, for example at the data table or in a workflow, thereby supporting reproducibility. However, a reference saved in the Journal requires the referenced data table. Therefore, an existing link or a script without corresponding data is still not a credible reproducibility justification.
For supported platforms, Automatic Recalc can account for changes such as added or excluded rows in the report. When the feature is disabled and the table is modified, the help text describes a warning regarding the loss of synchronization. Certain platforms do not natively support Automatic Recalc, including DOE and Make Validation Column. Do not generalize the behavior of a single window.
Therefore, verify the approved workflow from the correct data version up to the newly generated report. Document the script version, relevant options, and the expected computational basis. Upon repetition, compare more than one heading or file name: the data volume, variables used, model role, and key performance indicators must align with the original decision. For stochastic methods, additional technical validation of reproducibility must be established.
7. Get an understandable evaluation file
For a later verifiable decision, there must be the relevant linkage between source, processing, and result. Define which source data, selection rules, scripts, parameters, outputs, and evaluations are involved. Chapter 4 requires a traceable documentation relationship; it does not prescribe a universal JMP file package for this purpose. The specific retention purpose determines the required extent.
A practical record can include the relevant data table, the controlled script version, the associated report, and a brief justification for key data selections. This serves as a work aid, not a mandatory template for every run. When supplementary documentation is maintained outside of JMP, the reference to such documents, their availability, and the required controls must likewise be clearly traceable.
The ability to save a script or session log does not constitute a complete protected change history. Identify and evaluate the actual access and change controls in the operational environment. An additional review is not automatically sufficient to compensate for missing technical functionalities. Prior to a usage decision, it must be demonstrable that the overall design meets the specific audit objectives.
8. Correctly apply installation accuracy and numerical reference tests
JMP provides two distinct tools: an installation validation tool for verifying the integrity of installed files and a NIST-STRD-based test framework for numerical testing. The installation tool is available on the quality side for Windows. A corresponding macOS availability cannot be inferred from this information. Registration is required for both tools.
JMP also publishes a precision report for JMP/JMP Pro 19. It includes build information, operating systems, and tested procedures. The results constitute a limited supplier validation, not a direct assessment of a customer's installation, and do not represent complete verification of all functions in JMP 19.2.
Review the accepted evidence against the required functions. Installation integrity can provide a technical basis; reference data can support specific numerical calculations. However, acceptance of the evidence does not automatically confirm that the data are appropriately selected, that the method is suitable, or that a script accurately represents the intended analysis. These open points determine the additional required assessments.
9. Assigning responsibilities and managing changes within SMEs in a concrete organizational structure
Begin with a representative quality-related report and its complete development pathway. The department explains the observations and decisions. A statistically qualified responsible party justifies the method and data roles. IT or the system responsible party records edition, build, scripts, accesses, and storage. QA supports in aligning requirements and evidence with the quality system.
Develop a limited inspection plan with independently defined expectations. Record actual results and deviations, and determine the allowable usage scope. Effort arises particularly due to multiple analysis platforms, unstructured JSL customizations, additional components, changing data sources, and unclear historical reports. A repeatable routine workflow requires a different planning approach than an open tool suite for continuously evolving inquiry needs.
When changes are made, it is assessed which of these relationships are affected. A different build, modified script, or new platform selection may require different comparison cases rather than merely a redactional update to the report. A full revalidation of each change is no more justified than an unchanged release of all functions. Periodic evaluation also takes into account relevant incidents, changes, and existing proof-of-concept limits.
10. Checklist for the approved JMP analysis pathway
This dedicated checklist enables a comprehensive evaluation of a specific process. Open items are assessed by significance and linked to responsible parties along with corrective actions. A general note "JMP validated" does not replace this level of distinction.
- Intended use, quality release decision, and limitations of the evaluation pathway are described.
- Edition, Build, operating system, and actually used add-on components are recorded.
- Traceable observation unit, data origin, identifiers, and missing data rules are understandable.
- Hide, Exclude, Model Roles and their platform-specific significance were reviewed.
- Method and key settings are technically justified.
- Independent deviation investigations identify relevant state changes and incorrect usages.
- Data, scripts, and reports are part of the critical evaluation baseline.
- Supplier documentation was restricted to build, platform, and actual coverage.
- Access, modification, retention, and recovery are aligned with the intended use.
- Scope, responsibilities, and evaluation of future changes are defined.
Project experience and collaboration
Move from exploratory analysis to a controlled evaluation
With the responsible specialists, we follow your JMP analysis from data preparation to the conclusion used. For an existing analysis, the script version, settings and a known dataset provide a starting point. For repeated use, we develop a controlled workflow and test cases with your team. We establish when exploratory results begin to support a relevant decision and what evidence is needed for that use.
Frequently asked questions about JMP / JMP Pro
Are JMP and JMP Pro considered the same system for validation purposes?
They belong to the same product family but have different functional scopes. The key factor is which edition and which specific analysis platforms are used in your process. A documented routine report in JMP does not automatically cover the use of a newly introduced feature. JMP Live and JMP Clinical must be considered separately. Document the actual combination used, rather than granting release based solely on the brand name.
Must exploratory analyses also be included within the GMP scope?
Exploratory analyses are not automatically excluded from GMP scope. Once results, settings, or conclusions support a relevant decision, the transition must be documented and controlled. An open investigation pathway may require different forms of evidence compared to a fixed routine report. This assessment applies to the context of human medicinal product GMP; it does not constitute a general statement applicable to all research or clinical applications.
Has a hidden point been removed from the calculation?
Not solely due to hiding. The JMP Help distinguishes between the visibility of a point and its computational exclusion. In the mean example presented in this article, the scenario checks whether only the hidden value 34 remains; the expected mean value remains 16. This is because of the documented distinction between visibility and exclusion; the example does not represent a JMP error and is not a performed product test. Review the data role and the analysis used, in addition to the graphic output. For a true exclusion, a professionally justified rule and a traceable handling of the original dataset are required.
Are excluded rows in JMP always completely ignored?
No. In addition to the standard meaning of exclusion, platform-specific usages exist. Neural can use excluded rows for validation and non-excluded rows for training in an Excluded Rows Holdback scenario. Therefore, document the specific option and data role. A table with only exclude markings is insufficient to fully explain the impact of the corresponding observations on the overall model evaluation.
Why is the same validation column not universally applicable?
The rules depend on the analysis platform. In JMP Pro, the Standard Least Squares method ignores a validation column containing four or more distinct values. In contrast, Neural employs k-fold validation when more than three distinct values are present. Therefore, a copied column name does not constitute valid partitioning. Define the intended roles and verify how the selected platform actually interprets them.
Is statistical model validation already computerized system validation?
No. Evaluation of a statistical model on training, validation, or test data assesses specific characteristics of the model. Computer system validation evaluates the suitability of the defined system and analysis process for its intended use. This may include data transfer, configuration, scripts, permissions, and records. Both objectives can be interrelated; however, a good modeling result does not substitute for evaluation of the remaining relevant process steps.
Does each JMP report update automatically after a data modification?
No. Automatic Recalc is a setting available on supported platforms; some methods do not support it. When automatic recalculation is disabled, the help documentation indicates a warning is issued if data and reports are no longer synchronized. Verify the actually used report path and document its current status. An open window or unchanged filename is insufficient evidence that the current data base has been accounted for.
Sources: [6]
Does a stored JSL script constitute sufficient evidence of completeness?
A script can support repeatability, but it requires the correct data foundation and a traceable version. Appropriate retention must also be maintained for selection rules, relevant parameters, and professional judgment. A journal reference, for example, only functions when linked to the targeted data table. The ability to store scripts does not guarantee complete change history or overall regulatory release. Review the full required documentation, not merely the existence of a file.
What does the JMP Installation Tool and NIST Test Framework demonstrate?
The publicly described installation tool verifies installed files against manufacturer specifications and is offered for Windows. The NIST-based framework supports selected numerical validations. Prior to adoption, version and scope must be clarified. The appropriate customer data selection, statistical method, and correct process application remain separate validation issues.
How can we limit effort and testing scope for a update?
Start with the actually released analysis pathway: Edition, Build, Data sources, Platforms, Scripts, and Reports. Assess which changes impact this scope, and select appropriate comparison and error cases accordingly. For many individual workflows, the effort increases proportionally. A fixed test volume or full revalidation cannot be inferred solely from a version change; a justified evaluation must remain traceable.
Sources and review date
As of October 6, 2026. Supplier reference: current JMP Documentation Index 19.2 and named Help chapters; separate numerical whitepaper for JMP/JMP Pro 19, Build 19. August 2025. Registry-dependent validation tools were not referenced or performed; the installation tool is publicly described for Windows. Regulatory focus: Human medicinal product-GMP. The tables are internal work aids and do not represent completed product testing.
- Buy JMP — Select your solution
- JMP documentation
- Our commitment to quality — Quality Statement
- Assessing the Numerical Accuracy of JMP 19
- Exclude and Hide Markers
- Save Script Menus in JMP Reports
- Validation in Standard Least Squares
- Validation Methods for Neural
- European Commission: EudraLex Volume 4
- EU GMP Annex 11: Computerised Systems, January 2011 revision
- EU GMP Chapter 4: Documentation, January 2011
Sources checked: .
Review the JMP analysis pathway purposefully.
We assist you in aligning data roles, scripts, and report approvals with specific quality processes and planning appropriate checks in coordination with functional, QA, and IT.
- Computer system validation →
Bring your GxP systems into validated operation, from laboratory software, MES and eQMS to ERP. We define the scope, plan risk-based tests using GAMP 5 and hand the evidence to QA for release.

