Summary
- The problem
- A characterisation report keeps the conclusion but loses its basis: the settings each run actually used, which materials and scales the results cover, and which observations fell outside the study criterion.
- Seal’s approach
- The study protocol is configured in Seal and each condition runs against it, keeping its protocol version, actual settings and results. Neil prepares the assessment and the follow-up protocol; scientists decide what to test and which criteria apply.
- What changes
- A low-recovery run stays in the assessment, and the study’s coverage (runs per condition, lots, quality results not yet measured) is visible before a conclusion moves on. Reviewed ranges connect to the operating protocol and PPQ requirements with the evidence behind them.
- Where to start
- One protocol and representative results from a study you need to run. Book a demo.
Run each condition against the configured protocol.
A characterisation report usually keeps the conclusion and loses its basis: the settings each run actually used, the materials and scale it covers, and the observations that missed the study criterion. In Seal the protocol is configured and every condition runs against it, so each execution keeps its protocol version, recorded settings and results.
Characterisation studies are traditionally designed in a statistics package, run from paper or notebook records and summarised in a report whose ranges are then retyped into the control strategy. In Seal the study, its runs and the proposed ranges are linked records, so when Neil drafts an assessment or a follow-up protocol, each claim points back to the runs and coverage behind it, and a range the scientists approve keeps that basis attached.
The example is a capture load study asking how much protein the column can take before recovery falls: four corner conditions of load and residence time, and three centre runs.
Load × residence time
Select a condition to inspect its execution.
Inspect the configured study
Study PC-021 uses a 1 mL column and one load lot and resin lot. The protocol requests 20 and 40 g/L resin at 2 and 4 minutes, with three centre runs at 30 g/L and 3 minutes. The illustrative recovery criterion is ≥80%. Planned and recorded factor settings match in these seven records. Product-quality results and cross-lot confirmation are not supplied.
- Recorded load
- 40 g/L resin
- Residence time
- 2 min
- Protein loaded
- 40 mg
- Protein recovered
- 27.2 mg
Below the study’s 80% criterion. Retained in the assessment.
Keep the low result in the comparison.
At 40 g/L, recovery depends on which residence time was tested. The run below the criterion stays in the assessment rather than being set aside.
Protein recovery
Check residence time with it.
At 40 g/L, the 2-minute run recovers 68%; the 4-minute run recovers 87%. The load effect differs between the two residence times.
I’d confirm the corner conditions and add quality results before selecting a range. These observations do not establish a proven range.
Inspect the source results
| Run | Load g/L | Time min | Recovery |
|---|---|---|---|
| PC-021-01 | 30 | 3 | 85% |
| PC-021-02 | 20 | 2 | 84% |
| PC-021-03 | 40 | 4 | 87% |
| PC-021-04 | 40 | 2 | 68% |
| PC-021-05 | 30 | 3 | 86% |
| PC-021-06 | 20 | 4 | 91% |
| PC-021-07 | 30 | 3 | 87% |
All seven runs are included. Centre replicates do not replace replication at the corners. The source values are assay-derived recovered protein mass divided by loaded protein mass.
Know what the study actually covers.
Four corners, three centre runs, one material and resin lot. See the coverage before carrying a conclusion into the next study or scale.
Does the load effect repeat?
The centre runs give repeat observations at one condition. They do not establish repeatability at the four corners. Preserve the observed recovery difference and agree the replication needed to test it.
Return to the seven run results →Does recovery tell us about product quality?
Not on its own. This example contains no purity, aggregate or impurity-clearance results. Link the relevant pool samples and analytical evidence before drawing conclusions about critical quality attributes or parameter criticality.
Will the conclusion apply to another lot or scale?
These runs use one load lot and one resin lot on a 1 mL column. Record the scale-down model rationale, relevant equipment differences and material variability. Decide which additional evidence is needed for the intended manufacturing context.
What supports a proposed control?
Retain the exact analysis population, deviations, exclusions, model assumptions and diagnostics. A control proposal should reference the supporting studies, quality evidence, uncertainty and remaining risks—not only the best-looking run.
Turn the finding into the next protocol.
Neil drafts the assessment and the follow-up study from the recorded runs and their coverage. The scientists decide what to test, which criteria apply and when the protocol is ready to run; until then there is no approved range.
Capture study follow-up
PC-021 · DraftConfirm the corners. Measure quality alongside recovery.
Scientific rationale
The higher-load runs differ by 19 percentage points in recovery. Retain both results and test whether the difference repeats.
PC-021-03 and PC-021-04 →Before the study can run
- Replication and run order
- To agree
- Quality responses and criteria
- To define
- Material and scale coverage
- To assess
Inspect the proposed follow-up protocol
- Design
- Confirm load × residence-time behaviour; agree replication and run order with the study statistician.
- Execution
- Record actual factor settings, material and resin lots, column history, deviations and linked pool samples.
- Responses
- Recovery plus the relevant quality attributes and reviewed analytical methods.
- Before execution
- Scientists agree the design, acceptance criteria and applicability questions. Test the configured workflow.
No approved range yet. The study’s recovery criterion is not a product-quality approval.
Keep the study connected to the control it supports.
The same pattern applies when PD and MSAT teams characterise culture, purification, formulation or other unit operations: the experimental work and the control rationale stay on one record, from the first risk question to the process in use.
Start with the uncertainty.
Connect prior knowledge and risk questions to the factors and quality attributes worth studying. Record where a scale-down model or platform assumption still needs confirmation.
Process development →Keep your analysis tools.
Use JMP, Minitab, R or Python alongside Seal. Retain the exact dataset, exclusions, model version and diagnostics behind the reviewed interpretation.
Connected systems →Carry decisions into execution.
Connect reviewed ranges and controls to the operating protocol and PPQ requirements. Preserve the evidence and unresolved risks behind each decision.
Process validation →Learn from the process in use.
Link later trends and deviations back to the characterisation study. Reassess a conclusion when new evidence changes its basis; keep the earlier version.
Continued process verification →AQuestions and answers
Can we configure and execute characterisation protocols in Seal?
Yes. Configure the study question, factor levels, run order, materials, sampling, calculations and review steps. Each run retains the protocol version, actual conditions, observations and source results. Test the workflow against its intended use before relying on the configured controls.
Does Seal replace our DoE or statistical software?
Specialist analysis can remain in JMP, Minitab, R, Python or another suitable tool. Connect the analysis dataset, exclusions, script or model version, diagnostics and outputs to the study records. The website chart is a comparison of fictional observations, not a fitted statistical model.
What is the difference between explored and approved ranges?
Explored conditions are those actually tested. A proposed operating range needs supporting performance evidence, uncertainty and applicability assessment. An approved range also has a reviewed rationale and authorised decision. A few passing observations do not establish a design space.
How are CPPs and CQAs handled?
Connect a candidate process parameter to the critical quality attributes it may affect and the studies supporting that relationship. Critical process parameter (CPP) classification requires the relevant quality evidence and risk assessment. Recovery alone is not a substitute for critical quality attribute (CQA) evidence.
Can small-scale or platform evidence be reused?
Record the source study, model suitability, materials, equipment, scale and relevant differences. Reuse depends on an applicability assessment and any necessary confirmation work. Similar operating settings do not demonstrate comparable process behaviour at another scale or site.
How does this connect to PPQ and continued verification?
Carry reviewed parameter controls, sampling requirements, residual risks and supporting evidence into process performance qualification (PPQ). Link later continued process verification (CPV) trends and deviations back to the original characterisation conclusion. New evidence can trigger reassessment without replacing the earlier record.