Facing a data-related challenge? Explore some of the most common issues and discover how we can help you tackle them, gain deeper insights from your data, and find practical solutions.
Turn semiconductor production data into engineering evidence
Q-Sensei connects manufacturing and test data with analytics and AI to help semiconductor engineers investigate yield loss, pinpoint failures, identify production patterns, and uncover potential contributing factors.Learn more
What is causing semiconductor yield loss?
The challenge
When yield drops, I need to understand what changed and whether the issue is associated with particular lots, wafers, test conditions, equipment, or device characteristics. Finding the relevant evidence can mean repeatedly filtering and comparing data across different sources.
How Q-Sensei helps
Q-Sensei helps you compare affected and unaffected production, examine failure distributions, and filter results by relevant production and test attributes. You can narrow your investigation progressively, exploring the data in greater detail without repeatedly rebuilding your analysis.
Analytics and AI-assisted investigation help surface patterns and potential contributing factors worth examining. Use this evidence to prioritize follow-up analysis and decide which explanations to validate.
Compare affected and unaffected production
Affected production
Unaffected production
Comparable lots, wafers and test conditions
Compare lots with the same production and test conditions. Sample data.
Data typically used in this investigation
Wafer sort and final test results
Electrical test results collected during wafer sort and final testing show which devices pass or fail specifications. They help quantify yield loss and identify affected products, lots, wafers, and electrical parameters.
MES and lot genealogy records
Manufacturing Execution System (MES) records capture production steps, material history, and lot genealogy. They help trace yield changes back to specific production stages, routes, and material batches.
Equipment logs
Equipment logs document machine operating states, events, alarms, and process-related information. They help identify equipment-related patterns that coincide with yield deterioration.
Process history
Process history captures the manufacturing steps, recipes, and conditions applied to a production lot. It helps compare affected and unaffected production and investigate process-related differences.
Inline metrology
Inline metrology provides measurements of physical and material properties during manufacturing. It helps reveal process deviations that may be associated with subsequent yield loss.
Defect-inspection data
Defect-inspection data records detected defects and their locations on wafers. It helps identify defect patterns and assess whether specific defect types or regions correlate with yield loss.
Where are wafer and die failures concentrated?
The challenge
An increase in failures tells me that something may be wrong, but not where to focus. I need to determine whether the issue affects production broadly or is concentrated in particular lots, wafers, test sites, or failure categories. Overall counts can hide important patterns.
How Q-Sensei helps
Q-Sensei helps you move from overall failure trends to detailed breakdowns by relevant production and test attributes. Filter the affected population, compare failure frequencies, and examine distributions to identify where a problem is concentrated.
This helps establish the scope of the issue and determine which lots, wafers, or test conditions to investigate next. Where wafer-map data and visualization are configured, you can also explore spatial failure patterns.
Locate a concentration on the wafer
A spatial pattern to investigate
Failed dies cluster near the upper-right edge, with scattered failures elsewhere.
PassFail
Compare the pattern with other wafers and relevant test conditions.
Check whether the cluster appears on other wafers from the same lot. Sample data.
Data typically used in this investigation
Die-level pass/fail and binning results
These results record the test outcome and assigned bin for individual dies on a wafer. They help pinpoint failure concentrations and distinguish between different failure categories.
Wafer maps
Wafer maps visualize the spatial distribution of dies and their test results across a wafer. They help reveal localized failure patterns, such as edge effects, clusters, or recurring spatial anomalies.
Failure classifications
Failure classifications group failed devices by error type, test parameter, or failure category. They help distinguish dominant failure modes and identify which types contribute most to the overall failure rate.
Lot and wafer identifiers
These identifiers connect test results to their corresponding production lots and individual wafers. They help determine whether failures are isolated to specific production groups or occur more broadly.
Test-site records
Test-site records identify the physical test locations used during device testing. They help establish whether failures are concentrated at particular test sites rather than across the entire production population.
Defect-inspection maps
Defect-inspection maps show the location and distribution of detected defects on a wafer. They help compare physical defect patterns with electrical failure distributions.
When did semiconductor production performance change?
The challenge
Production behavior can change gradually or shift abruptly. I need to distinguish meaningful changes from normal variation, identify when a shift began, and establish which production or test populations were affected.
How Q-Sensei helps
Q-Sensei helps you analyze production trends over time, detect anomalies, and investigate change points. Compare different time periods and narrow the results by relevant lots, testers, sites, or test conditions.
Explore the underlying distributions and production details around a change rather than relying on a single alert or summary metric. This helps determine when and where to focus your investigation and identify the evidence that needs closer review.
Examine the evidence around a change
Earlier productionChange to examine
Time periods, lots, testers and conditions
Compare time periods, then inspect the production and test populations involved. Sample data.
Data typically used in this investigation
Historical yield and test results
Historical test datasets capture yield, pass/fail rates, and electrical performance over time. They help establish baselines, identify trends, and determine when production behavior started to change.
SPC charts and process-control measurements
Statistical Process Control (SPC) data tracks process measurements against statistical control limits. It helps distinguish normal process variation from unusual shifts or trends that may require investigation.
Equipment sensor traces
Equipment sensor data records operating parameters such as temperature, pressure, flow, and vibration over time. It helps identify equipment behavior that coincides with changes in production performance.
Alarm and event logs
Alarm and event logs record equipment warnings, faults, and operational events with timestamps. They help correlate production changes with specific machine events or interruptions.
Lot processing timestamps
Processing timestamps document when individual lots enter, leave, or complete manufacturing steps. They help align production events with performance changes and identify affected time windows.
Manufacturing history
Manufacturing history provides a chronological record of production activities and process changes. It helps reconstruct what happened around a performance shift and determine which production groups were affected.
Are test equipment or conditions affecting semiconductor test results?
The challenge
A change in failure rate may reflect device or manufacturing behavior—or differences in how devices are tested. I need to compare test equipment, sites, programs, and conditions while accounting for differences in the production being measured.
How Q-Sensei helps
Q-Sensei enables comparisons across testers, test sites, test programs, and test conditions. Examine how failure patterns vary across lots and other relevant dimensions, using targeted filters to investigate specific questions.
These comparisons help assess whether an issue appears associated with the test environment, a particular production subset, or another factor. You can focus on the most relevant evidence without assuming that a correlation proves the cause.
Compare the same production across test setups
Matched production, shown on the same scale
Test setup A
Reference result pattern
Test setup B
A different pattern in the marked subset
Examine tester, site, program and condition records behind a difference in the results. Sample data.
Data typically used in this investigation
ATE results
Automated Test Equipment (ATE) results contain electrical measurements and pass/fail outcomes from semiconductor testing. They help compare test performance across equipment, sites, and production groups.
Test-program versions
Test-program records document the software versions and test sequences used to evaluate devices. They help identify whether changes in test logic or program versions coincide with altered test results.
Tester and test-site records
These records identify the testers and physical test locations used during production testing. They help reveal systematic differences in results between machines or test sites.
Probe-card and load-board records
These records document the hardware interfaces used to connect devices to test equipment. They help investigate whether interface hardware differences are associated with inconsistent test results.
Test conditions
Test-condition records capture parameters such as voltage, temperature, timing, and frequency during testing. They help determine whether differences in measurement conditions influence observed failure rates.
Retest history
Retest records capture repeated test attempts and their respective outcomes. They help identify intermittent failures, inconsistent results, and potential test-related effects.
Lot and wafer identifiers
These identifiers link test measurements to their corresponding production populations. They help ensure that tester comparisons account for differences in the devices and lots being measured.
Which electrical parameters show significant deviations?
The challenge
Electrical measurements vary across devices, wafers, and lots. Averages and pass/fail summaries may conceal distribution shifts, unusual values, or subgroups approaching specification limits. I need to understand how measurements behave—not just see a headline statistic.
How Q-Sensei helps
Q-Sensei lets you explore parameter ranges, statistical summaries, histograms, and outliers. Focus on a particular measurement, compare its distribution across production groups, and investigate unusual values or measurements near relevant limits.
This helps distinguish broad distribution shifts from isolated outliers and identify which parameters deserve closer attention. You can then prioritize the measurements and production subsets most relevant to your investigation.
Examine distribution shape, unusual values and relevant specification limits. Sample data.
Data typically used in this investigation
Wafer acceptance test (WAT) and PCM measurements
Wafer Acceptance Test (WAT) and Process Control Monitor (PCM) data capture electrical measurements of process-monitor structures on wafers. They help identify process-related electrical deviations before or alongside downstream device testing.
Wafer sort and final-test parametric results
Parametric test results contain measured electrical values for individual devices and test parameters. They help reveal distribution shifts, unusual values, and differences between production groups.
Electrical specification limits
Electrical specifications define the acceptable measurement ranges and performance thresholds for device parameters. They help identify measurements approaching or exceeding critical limits and assess their engineering significance.
Lot, wafer, and die identifiers
These identifiers connect electrical measurements to their respective production groups and individual devices. They help localize parameter deviations and compare distributions across lots, wafers, and dies.
Which manufacturing and equipment factors influence yield?
The challenge
A yield or reliability issue may be associated with several factors, from manufacturing history and equipment conditions to device characteristics and test results. Relevant information may be spread across structured datasets, reports, and documents, making cross-domain investigation difficult and time-consuming.
How Q-Sensei helps
Q-Sensei prepares structured and unstructured data for analytics and AI according to your environment. Once relevant sources and attributes are configured, you can explore performance alongside manufacturing, equipment, device, and test dimensions to investigate potential relationships.
AI-assisted analysis can help identify relevant patterns and suggest hypotheses or next steps. Examine those suggestions, check whether patterns persist across comparable groups, and determine what the evidence supports. The goal is to help investigate plausible explanations—not to replace engineering judgment.
Bring different evidence into the same view
Shared window to investigate
Manufacturing events
Equipment signal
Electrical results
Earlier recordsLater records
A recorded event, a signal change and different test results fall within the marked time window.
Check which lots and test steps coincide with the signal change. Sample data.
Data typically used in this investigation
MES process and routing history
Manufacturing Execution System (MES) records document the manufacturing steps, process routes, and production history of each lot. They help connect yield behavior with specific process stages and production paths.
Equipment sensor data and FDC records
Equipment sensors and Fault Detection and Classification (FDC) systems capture machine parameters and detect abnormal equipment behavior. They help investigate equipment conditions and process deviations potentially associated with yield or reliability issues.
Inline metrology
Inline metrology provides physical and material measurements collected throughout manufacturing. It helps identify process characteristics that may correlate with electrical performance or yield changes.
Defect-inspection data
Defect-inspection systems record detected physical defects and their characteristics across wafers. They help connect defect types and spatial patterns with observed electrical or yield abnormalities.
Electrical test and yield results
Electrical test and yield datasets capture device performance and production quality outcomes. They provide the performance indicators needed to examine relationships with manufacturing and equipment factors.
Engineering reports
Engineering reports document investigations, process observations, failure analyses, and technical findings. They provide contextual knowledge that helps interpret patterns in quantitative production data.
Maintenance records
Maintenance records document equipment servicing, repairs, component replacements, and interventions. They help investigate whether equipment maintenance events coincide with changes in process behavior or yield.
Did process changes improve semiconductor yield?
The challenge
After a process or test adjustment, I need to determine whether the expected improvement is visible in production data. A simple before-and-after comparison can be misleading if lot composition, test conditions, or other factors have also changed.
How Q-Sensei helps
Q-Sensei helps you compare relevant periods or production groups using yield, failure rates, and parametric distributions. Break down the results by lot, wafer, tester, site, or test condition to assess whether the observed change is consistent across comparable subsets.
This provides a systematic way to evaluate the evidence after an intervention, identify where improvement appears strongest, and flag areas that need further investigation. Assess whether the results support the intended outcome without treating a change in metrics as proof of causation.
Compare yield before and after a process change
20 devices per group in each period
Before process change
Matched group A
Matched group B
After process change
Matched group A
Matched group B
PassFail
Match production groups and account for condition changes
Failures fell from four to one in group A and from three to two in group B. Sample data.
Data typically used in this investigation
Before-and-after yield and test results
These datasets capture yield, failure rates, and electrical performance before and after an intervention. They help quantify performance changes and determine whether improvements are consistent across comparable production groups.
Process recipes and engineering change records
These records document manufacturing recipe settings and approved engineering changes. They help establish exactly what was modified and when the change was introduced.
Equipment maintenance and intervention logs
These logs document equipment repairs, maintenance activities, and technical interventions. They help connect equipment-related actions with subsequent production performance.
Lot genealogy
Lot genealogy traces production lots through materials, process steps, and manufacturing history. It helps identify which production groups were exposed to a particular intervention and supports comparable group analysis.
Test-condition history
Test-condition history captures the testing parameters applied to devices over time. It helps distinguish actual process improvements from changes caused by different test conditions.
SPC trends
Statistical Process Control (SPC) trends track process measurements and statistical behavior over time. They help assess whether an intervention resulted in sustained process stability or a temporary improvement.
Parametric distributions
Parametric distributions show how measured electrical values are spread across devices and production groups. They help determine whether an intervention improved overall device performance or shifted only selected measurements.
