Desrochers Validation, Félix Desrochers Back to the offer

What I don't do, and where to check everything I claim

This page documents the boundary of the service, the sources used, public resources for doing the work yourself and alternatives that may be a better fit. It is reasonable to be cautious before sending an internal workbook to a new supplier.

1. What the service establishes

Which file was examined. The record identifies the workbook by filename, hash, worksheet inventory and formula inventory.
Which calculations were tested. Supported calculations are compared with expected results derived independently from their requirements and inputs.
What the tests found. Differences, review items and unevaluated calculations remain separate instead of being collapsed into a pass.
What remains with the laboratory. The record maps all 47 NIST GLP 15 checks and leaves laboratory-only answers and authorization to your technical manager.

The result is file-specific validation evidence prepared for laboratory review. It is not a certificate and does not establish that every calculation in the workbook passed.

2. What the service does not establish

Read this column on the right rather than the one on the left. A list of things I refuse is only useful if it tells you what you are still on the hook for.

I do not	Which means
Decide which contributors your measurement needs	ILAC P14 5.4 calls for relevant short-term contributions, contributions reasonably attributable to the customer's device and applicable contributions used in the CMC evaluation. It does not prescribe one universal list. I report when the customer-device basis is not identifiable from the workbook, but deciding which contributors apply to your measurement remains your laboratory's metrology judgment.
Answer 11 of the 47 checks	Your record arrives with blanks in it: backups, password handling, macro behaviour, whether the correct procedure was chosen. It is not a finished document until you complete those. I leave them visible rather than filling them in, because no examination of a file can answer them.
Test macros, buttons, or printed output	GLP 15 section E leaves those to the laboratory and so do I. I detect that macros are present; I do not run them or judge whether they work.
Know what your workbook was meant to do	Section 3 of the record is a requirements specification I reverse-derived from the file when you do not supply one. If it misstates your intent, every test below it is measuring against the wrong specification. Reading that section is the one step you cannot delegate to me, and it is the single most important thing on this page.
Modify the legacy `.xls` file you send	I make a separate `.xlsx` working copy with Microsoft Excel or LibreOffice because the analysis library does not read `.xls` directly. Your original stays unchanged. The record identifies both files by hash, names the conversion tool and states that the converted copy was the file examined. Macro and button behaviour remains assigned to the laboratory.
Validate your LIMS	Waters, LabWare and LabLynx do that and they are good at it. I work on the side workbooks that feed it, which is a different problem.
Work under 21 CFR Part 11 or pharma CSV	Different regulator, different liability, much heavier documentation. If that is your world, see section 5.
Certify, attest, approve, or accredit anything	Desrochers Validation is not accredited, and this record is not presented as accreditation, certification or approval. Your technical manager reviews the evidence, confirms the intended requirements and decides whether to authorize the workbook for use.
Predict what your assessor will conclude	I do not control your assessor and cannot guarantee an assessment outcome.
Give an opinion on results you already issued	I validate software and do not issue a conformity conclusion about released results. If a finding may affect previous results, I report it confidentially to your laboratory contact. Your laboratory determines the investigation, correction and notification required by its management system.
Have a track record	No record I have produced has been through an accreditation assessment yet. I started in 2026. Being able to complete a NIST form is not the same as your assessor accepting it, and I have no evidence about the second thing. That is the honest reason the first file is free.

3. What happens to your file

Send a carefully redacted working copy. Remove customer names, sample identifiers and unrelated observation or result rows. Retain every value used by a formula, uncertainty contributors, calibration points, formulas, labels, units, named ranges and workbook structure. Removing calculation inputs can prevent independent testing or change the record.
Kept only for the engagement and the retention period agreed in writing, then deleted, with the deletion confirmed to you in writing.
Analysis stays local. I do not send workbook content to AI tools, cloud analysis services or outside contractors. Your email provider and mine necessarily transport an attachment sent by email.
Software improvements use synthetic cases. If your file reveals a new formula, error or layout that the tool should handle, I reproduce the technical issue in a new workbook without your data, labels or workbook content. Your file is not retained as a test fixture or used for model training.
Your NDA can be reviewed first. Send it before the workbook if you want confidentiality terms in place before file transfer.
If I find a problem that may affect previously issued results, I contact your technical manager before the document goes anywhere. I contact no one else: not your customers, not your accreditation body, not a regulator. It is not my call and not my role. I flag what I found and how far it reaches; your laboratory then follows its own investigation and notification requirements.

Potentially consequential findings are handled confidentially with the laboratory contact. The laboratory decides the investigation, correction and notification steps required by its own management system.

4. Standards, sources and self-service resources

Where every claim comes from

The table distinguishes primary requirements and guidance from secondary background material so you can check the basis of the method.

Claim	Source	Kind
Uncertainty must be evaluated for each measurement function	ISO/IEC 17025:2017 clause 7.6. The standard is not free to redistribute, so I describe what it requires in my own words rather than quoting its text. The method I actually test against is the GUM, below, which is free and public.	Paid standard
How an uncertainty budget is built and combined	JCGM 100:2008, the GUM, free from BIPM. It supplies the uncertainty framework used by the supported calculations; workbook implementations vary.	Primary, free
Expanded uncertainty to at most two significant figures; coverage factor and coverage probability both stated	ILAC P14, free. A2LA's R205 defers to it.	Primary, free
NIST OWM publishes a per-file software quality-assurance form	NIST GLP 15, and NIST IR 8250. GLP 15 is an Office of Weights and Measures procedure. I use its 47-item form as the record structure; I do not claim every ISO/IEC 17025 assessor requires that particular form.	Primary, free
Your accreditation scope and the range in my first email	Your own certificate in the A2LA public directory, with the date I retrieved it recorded.	Primary, free

Public resources for doing much of this work yourself

Several core references are free. The ISO standard requires licensed access, and NIST training may have a fee. If you have the time and appropriate technical judgment, here is a practical starting path.

Download the form. NIST GLP 15. Appendix A is the checklist used as the record's organizing structure.
Download the GUM. JCGM 100:2008. Sections 4.2 to 4.3 cover Type A and Type B evaluation, 5.1.2 the combination rule, 6.2 coverage factors. Those four sections are most of what matters.
Read ILAC P14 section 5 for the calibration-certificate uncertainty reporting rules, including the limit of two significant digits for expanded uncertainty.
Consider the NIST training. NIST OWM runs the paid course Software Verification and Validation, Parts I and II, focused on Excel software quality assurance in calibration laboratories.
Use NIST's own templates and NIST Technical Note 1297 for worked uncertainty guidance.
Review A2LA's public requirements and guidance at portal.a2la.org/documents for the documents A2LA makes available to laboratories and assessors.

The work requires time to document requirements, derive independent expected results, execute tests and record what remains unresolved. The service is an option for laboratories that do not want to build and maintain that process internally.

5. The complete NIST GLP 15 checklist

The record is built around the NIST Office of Weights and Measures Software Verification and Validation form, published as GLP 15 (2019), Appendix A.

Of its 47 checks, 30 are designed to use evidence from the file, 6 gather file evidence and leave the judgment to you, and 11 only you can answer. A check is answered from the file only when that file establishes the whole claim. Every check appears in the record, including those the service cannot answer.

Show all 47 checks

A. Software Inspection

Check	What it asks	Who answers it
A1	Spreadsheet is clear and makes sense	I gather the evidence, you judge it
A2	There are instructions for use	I assess it from the file when the required evidence is present
A3	Instructions and data input appear on the visible portion of the first worksheet	I assess it from the file when the required evidence is present
A4	Data-entry fields are labeled and color coded (avoiding red and green)	I assess it from the file when the required evidence is present
A5	The Standard Operating Procedure used is clearly specified	I assess it from the file when the required evidence is present
A6	The number of digits to be rounded to is specified	I assess it from the file when the required evidence is present
A7	The user is warned whenever data-entry fields are left blank	I assess it from the file when the required evidence is present
A8	Data-entry fields are blank when opened, so old data is not reused	I assess it from the file when the required evidence is present
A9	The software opens at the right location within the file	I assess it from the file when the required evidence is present
A10	Unused fields/cells are locked	I assess it from the file when the required evidence is present
A11	Rows/columns the operator need not see are hidden	I gather the evidence, you judge it
A12	Unused sheets are removed	I assess it from the file when the required evidence is present
A13	Worksheets are named appropriately	I assess it from the file when the required evidence is present

B. Mathematical Specification

Check	What it asks	Who answers it
B1	The correct SOP is used	Only you can answer it
B2	The formulae and methods chosen from that SOP are specified	I assess it from the file when the required evidence is present
B3	Sources and references for formulae are specified	I assess it from the file when the required evidence is present
B4	The SOP, its methods and formulae are appropriate to the level of precision	Only you can answer it

C. Code Review

Check	What it asks	Who answers it
C1	The formulae in the fields exactly match the procedure	Only you can answer it
C2	Repeated calculations appropriately reference the correct cells	I assess it from the file when the required evidence is present
C3	Calculations, tested with reference data, show appropriate accuracy	I assess it from the file when the required evidence is present
C4	Rounding is done at the appropriate locations in the file	I assess it from the file when the required evidence is present

D. Numerical Stability

Check	What it asks	Who answers it
D1	Calculations are stable with large numbers and small differences	I assess it from the file when the required evidence is present
D2	Fields are categorized as Number and not General when appropriate	I assess it from the file when the required evidence is present
D3	Number cells are locked to an appropriate number of decimal places	I assess it from the file when the required evidence is present

E. Component Testing

Check	What it asks	Who answers it
E1	Conditional logic cells handle negative values properly	I assess it from the file when the required evidence is present
E2	Conditional logic cells withstand a boundary value test	I assess it from the file when the required evidence is present
E3	Each macro used is functional	Only you can answer it
E4	Each command/button is functional	Only you can answer it
E5	Combinations of interdependent macros are functional	Only you can answer it
E6	Plotted graphs are accurate	I gather the evidence, you judge it
E7	Plotted graphs and their axes are properly labeled	I assess it from the file when the required evidence is present
E8	Worksheets/reports print properly, if needed to	I assess it from the file when the required evidence is present
E9	Conditional (color and non-color) formatting is functional	I gather the evidence, you judge it

F. Numerical Reference Results

Check	What it asks	Who answers it
F1	Look-up tables and lists match the latest calibration report	Only you can answer it
F2	Uncertainties match the latest Scope	I assess it from the file when the required evidence is present
F3	Values that reference another workbook are dated	I assess it from the file when the required evidence is present
F4	Behaviour is as intended when a master list's date is updated	I gather the evidence, you judge it

G. Embedded Data Evaluation

Check	What it asks	Who answers it
G1	Embedded data (conversion factors, reference values) is correct	I gather the evidence, you judge it
G2	The evaluation of the embedded data is dated and documented	I gather the file evidence; you confirm the evaluation record

H. Back-to-Back Testing

Check	What it asks	Who answers it
H1	Newer and older spreadsheets agree down to intermediate calculations, dated	I assess it from the file when the required evidence is present

I. Analysis Without Computer Assistance

Check	What it asks	Who answers it
I1	Hand calculations agree with the spreadsheet, or differ insignificantly	I assess it from the file when the required evidence is present

J. Security

Check	What it asks	Who answers it
J1	Equation and calculation cells are protected against inadvertent editing	I assess it from the file when the required evidence is present
J2	Cells are locked in place and cannot be moved or dragged	I assess it from the file when the required evidence is present
J3	Confidentiality of passwords is appropriate	Only you can answer it
J4	Files are backed up automatically	Only you can answer it
J5	Additional back-up is available at alternate facilities	Only you can answer it
J6	Files on network drives cannot be accidentally deleted	Only you can answer it

6. Alternatives and when to use them

ISOBudgets

Go to them, not me, if you do not have an uncertainty budget yet. They build budgets; I document and verify budgets you already have. Different product, same reader. isobudgets.com.

The 17025 Store, now trading as Uncera

Sells an Equipment Uncertainty Calculation Excel template with a guide. It is a tool for building a budget, not a validation of an existing workbook. Check its current product page before purchasing. uncera.com.

Pharma and 21 CFR Part 11 specialists

If you are FDA-regulated, I am the wrong supplier and you should not hire me. Ofni Systems sells ExcelSafe, which adds audit trails and electronic signatures to Excel, and does validation packages around it. That world requires different controls and documentation because the regulator is different. ofnisystems.com.

A general ISO 17025 consultant

A general consultant offers a broader engagement and may make the metrology judgments I explicitly refuse to make, write your quality manual and support an assessment. If what you actually need is someone to decide what belongs in your budget, hire one of them.

LIMS validation vendors

Central-system validation is a different engagement. If you need validation of a LIMS rather than evidence for individual Excel calculation files, use a supplier that specializes in that system.

Doing nothing

Sometimes the right answer. This service is most useful when important Excel calculation files lack current, file-specific evidence. The number of scope lines does not by itself determine the number of workbooks or budgets.