Methodology
One calibration. Every candidate model. One defensible answer.
WLTR evaluates every calibration curve the same way every time, from a bioanalytical run under ICH M10 to an initial calibration under EPA methods: every candidate model, scored against 15 acceptance criteria, ranked and documented. The method is patented (US 12,038,448 B2) and built on four decades of analytical chemistry practice.
The problem
Why r² is not enough.
Most calibration software reports a correlation coefficient and a pass or fail flag. A curve can post an r² above 0.998 and still recalculate its lowest standard with large error, right at the LLOQ or reporting limit where regulatory decisions sit. ICH M10, the bioanalytical method validation guideline FDA and EMA both apply, accepts a calibration curve on the back-calculated concentration of each standard: within ±15% of nominal, ±20% at the LLOQ, for at least 75% of the standards across at least six levels. TNI and the AAFS Standards Board likewise direct labs to judge calibrations by relative error and residuals rather than r² alone. The WLTR ICH M10 guide sets out these rules section by section, and Insights 13 applies them to a worked LC-MS/MS run.
How to read this: each dot is one calibration standard. The closer to the center line, the more accurately that standard is recovered. Pick a regression model:
Back-calculated accuracy by calibration level. The shaded band shows ICH M10 limits: ±15% of nominal, ±20% at the LLOQ. Illustrative data.
How WLTR evaluates a calibration
Four steps, applied the same way to every analyte.
Capture the source
WLTR reads the raw text export from your data system, records a SHA-256 hash of the original file, and keeps it as an immutable source record.
Fit every candidate model
Each analyte is fitted against the established calibration models your methods allow, including average response factor, linear and quadratic regression, unweighted or with 1/x or 1/x² weighting.
Score against 15 acceptance criteria
Every model is scored on the same 15 acceptance criteria, including percent error at each calibration level, sensitivity at your LLOQ or reporting limit, response factors and %RSD, minimum r or r², and verification against a second-source standard.
Rank and document
The candidate models are ranked, the simplest model that meets every criterion is recommended, and the full evaluation is stored for review and approval.
What the evaluation looks like
LC-MS/MS run · Analyte in human plasma, 1 to 1,000 ng/mL · Illustrative
| Candidate model | Fit | Accuracy at every level | Accuracy at the LLOQ | Second-source check | Criteria met | Outcome |
|---|---|---|---|---|---|---|
| Average response factor | ✕ | ✓ | ✕ | ✓ | 11 / 15 | Rejected |
| Linear, unweighted | ✓ | ✕ | ✕ | ✓ | 10 / 15 | Rejected |
| Linear, 1/x | ✓ | ✓ | ✕ | ✓ | 13 / 15 | Rejected |
| Linear, 1/x² | ✓ | ✓ | ✓ | ✓ | 15 / 15 | Recommended |
| Quadratic, unweighted | ✓ | ✕ | ✕ | ✓ | 12 / 15 | Rejected |
| Quadratic, 1/x² | ✓ | ✓ | ✓ | ✓ | 15 / 15 | Passes · less simple |
Two models meet all 15 criteria. WLTR recommends the simpler one, and your analyst makes the final decision.
Design principles
Built to give the same answer for the same data.
Same input, same result
Given the same calibration data and the same acceptance criteria, WLTR returns the same ranking and recommendation every time.
Every score is visible
Reviewers can see each criterion, each recalculated level and each model side by side. The recommendation is explained by the evidence, not hidden behind a single score.
Your criteria, enforced
Acceptance thresholds, roles and approval gates are configured to your methods. WLTR enforces your SOP rather than replacing it.
Parsimony by default
When more than one model passes, WLTR favors the simplest one, consistent with ICH M10, which calls for a simple regression model that adequately describes the concentration-response relationship, and with SW-846 Method 8000D and ANSI/ASB Standard 036.
Where it comes from
Four decades at the bench, written into software.

The WLTR methodology was developed by Walter “Chip” Hinchman, our Chief Science Officer, from more than 40 years in analytical chemistry and laboratory operations, including work with the NELAP and TNI programs labs are assessed against today. It is protected by U.S. Patent 12,038,448 B2.
Questions
Frequently asked questions
Is WLTR a statistical black box?
No. WLTR evaluates established calibration models with documented criteria. Every score and every recalculated level is visible to the analyst and reviewer.
Does WLTR change our calibration or our results?
No. WLTR evaluates the calibration data your data system already produced. Your chromatography data system keeps running as it does today, and your lab decides which model it reports.
What are the 15 acceptance criteria?
They follow the evaluation parameters described in the WLTR patent, including percent error (back-calculated accuracy) at each calibration level, sensitivity at your LLOQ or reporting limit, the minimum number of calibration points for each model, %RSE, response factors and %RSD, minimum correlation (r or r²), and verification against a second-source standard such as an ICV. Your lab sets the acceptance value for each one, so a bioanalytical lab can apply ICH M10 limits and an environmental lab its method limits. We walk through the full list and how each is configured in a demo.
Does WLTR work for bioanalytical runs under ICH M10?
Yes. WLTR is built for quantitative methods that report against a calibration curve, including the LC-MS/MS, HPLC and UPLC assays used in bioanalysis. Your lab sets each acceptance value, so the back-calculated accuracy limits can follow ICH M10 (±15%, ±20% at the LLOQ), and WLTR shows how every candidate model performs at every level, including the LLOQ and ULOQ. Your validated method and your SOPs still decide which model you report.
Is the methodology patented?
Yes. The WLTR methodology is covered by U.S. Patent 12,038,448 B2.
Next step
See the methodology run on your own calibration.
Bring a recent calibration export, from an LC-MS/MS run or an ICAL. We will show you every candidate model, every criterion and the recommendation for each analyte.