1. State the operational question and quantity
“Is reference response shifting?” suggests a reference parameter. “Is precision worsening?” suggests a variability metric. “Did a control-lot change move the assay?” requires a lot-aware comparison. Give each chart one quantity, unit, transformation, direction, and interpretation.
Define whether the display monitors process stability, method behavior, or a control material. Do not use statistical control limits as unnamed substitutes for assay acceptance or product specifications.
2. Build the ordered record set
Select runs using explicit method, status, control, lot, instrument, site, and date rules. Retain stable run identities, source links, exact values, method and software versions, suitability, exclusions, and report links behind each point.
A plotted value is not enough. The worked point below shows the context needed to reproduce and investigate one signal without reopening an undocumented spreadsheet.
| Field | Worked value |
|---|---|
| Run | RUN-009 |
| Executed | 2026-08-12T14:05:00Z |
| Method | RP-4PL-v3 |
| Metric | Reference recovery |
| Unrounded value | 109.0% |
| Baseline | LJ-RECOVERY-v1 |
| Control lot | CTRL-24-017 |
| Instrument | READER-02 |
| Suitability | PASS |
| Source | Analysis AN-8821 and report RP-8821.pdf |
3. Calculate and approve the baseline and rules
In the linked Levey–Jennings example, baseline version LJ-RECOVERY-v1 has center 100%, SD 4 percentage points, and lines at 96/104%, 92/108%, and 88/112%. A predefined warning rule treats one point beyond ±2 SD as a signal.
Retain baseline membership, formulas, calculated values, rule logic, effective date, rationale, and approval. A rebaseline creates a new version and effective point; it never rewrites the limits that governed RUN-009.
Worked lines: 100% ± 1×4%, ± 2×4%, and ± 3×4%4. Plot results with changes on the same timeline
Show the result, center, control limits, and any separately labelled acceptance limits. Annotate reagent, reference, cell, and control lots; instruments and maintenance; analysts; method and software changes; and known deviations at their effective run.
Use non-color encodings for points, limits, and signals. A reviewer should be able to open the source analysis from a point and inspect the exact unrounded value and context.
5. Reproduce and triage each signal
RUN-009 is 109.0%, above the +2 SD line of 108% and below +3 SD at 112%. Under the worked warning rule, the chart creates one warning linked to RUN-009; under a rule set that omits the pattern, the same point does not acquire a new interpretation after it is observed.
Store the rule version, observed value, limits, calculation time, active filters, and notification. Triage scientific impact using run suitability, neighboring results, and known changes without altering the source result.
6. Retain investigation, action, and baseline decisions
Link the evidence, causal conclusion, affected records, disposition, correction, action, approval, follow-up, and any decision to bridge or rebaseline. Preserve unresolved signals and conclusions with no confirmed cause.
Review chart usefulness and false-signal burden under change control. Exported views identify their record set, baseline, rule, filters, and generation time so a static image remains interpretable.
Limits and current product support
Drift, shift, isolated extremes, and changing variability require explicit definitions; no visual pattern is self-interpreting. The SPC guide addresses method choice, while the Levey–Jennings and Westgard pages address specific conventions.
Current Provenarium trending is bounded to method, date, and metric views with one explainable rule signal. Customer-configurable chart families, arbitrary rule libraries, and automated quality dispositions are not current standard capabilities.
Frequently asked questions
What is drift?
Drift is a gradual change in a monitored quantity over time. Its practical definition and detection rule should be specified for the method and distinguished from random fluctuation.
Should each lot have separate limits?
Not automatically. Use bridging evidence and the monitoring objective to decide whether to annotate, stratify, adjust, or establish a new baseline.