Yield and measurement drift.
Catch a falling pass rate or a shifting measurement. Compare the affected runs with their own baseline, then open an investigation.
Yield drop · Measurement driftALERTS
Know when your test line changes.
Catch yield drops, drifting measurements, and reference checks that go wrong. Open the affected data and give your team a clear place to start.
FROM SIGNAL TO NEXT STEP
A change worth opening.
Explore three examples. See the measurement or run history, read the notification, and follow the next step.
EXAMPLE DATA
Measurement drift+61.3 mV
250 baseline readings → 100 incident readings
Interactive example. The charts explain the incident; actual firing also depends on the rule, history, and minimum sample counts.
WHAT TO WATCH
Three ways to watch the line.
Trust the test itself.
Check that known-good units pass and known-faulty units fail. An unexpected or incomplete result flags the reference check.
Golden sample failed · Failing sample passedSpot a pattern at the bench.
Find units tested repeatedly on the same procedure, or configure a rule for a streak of failed production runs.
Unit retest threshold · Run failedConfigure thresholds and sample requirements
Automatic detection
Detection follows recent samples and compares them with the process baseline. Minimum activity checks keep thin data from producing a verdict.
- Yield drop
- Info at 2σ · warning at 3σ · critical at 5σ
- Measurement drift
- Info at 3σ · warning at 4σ · critical at 5σ
Custom thresholds
Choose one alert type per rule and a threshold for each severity. For trend rules, compare a 1–90 day window with the preceding three times that window.
- Default comparison
- This week against the preceding three weeks
- Scope
- Procedures, stations, and parts
Use the right measurement
Mean drift measures the shift in process sigmas and works without specification limits. Cpk drift and Cpk floor require numeric limits.
- Yield thresholds
- Drop in percentage points
- Station or part scope
- Compares the selected slice with its own baseline
How much data is needed?
Automatic yield detection warms up with about 275 first runs over the past 90 days. Automatic drift needs 70 values and five distinct values per series, with the full severity range at 100. Severity also accounts for how close a measurement is to its specification limits.
Custom yield rules need at least 10 recent runs and 100 baseline runs. Custom mean drift needs 10 recent and 30 baseline samples; Cpk metrics need 30 on each side. These product minimums are not editable in the rule form.
The example charts above illustrate a completed incident. Their display window is not the automatic detector’s warm-up history.
KEEP THE CONTEXT
One incident. One place to follow it.
Keep the chart, notifications and investigation together as the incident develops.
Notify the right people.
Choose recipients and the severity that starts email delivery. Follow the same incident in the alert inbox.
Investigate the affected runs.
Open an AI investigation from a yield drop or drift alert. Its scope carries the affected runs and healthy baseline.
Explore AI investigationsKnow when it recovers.
Trends resolve on recovery; reference checks on the expected result. Retest and consecutive-run alerts resolve manually.
Explore next-step workflowsMake the next alert actionable.
Connect your test results, choose what to watch, and follow an incident all the way to the next check.
