Analyze
Visual investigation first. Linked scatter, histogram, wafer, and spatial views stay in sync, and PAT or DPAT starts directly from the tests you are already reviewing.
Wafer Sort Review
From failing lot to signed-off decision.
Stratum keeps the wafer sort lot, the investigation, PAT or DPAT screening, the exact limit review, and the signed-off artifact in one workflow. It is built for mixed-signal and analog teams still doing wafer sort analysis in Excel or JMP, then rebuilding the decision, the audit trail, and the access control somewhere else.
Prefer docs first? Read the quickstart.
How Analysis Works
Three modes, one context.
Most wafer sort work starts in the visual surfaces. When the question gets more custom, the same lot, selection, and limits can continue into SQL or Python notebooks.
SQL
Query directly against the same lot, snapshot, selection, and active limit context when you need a custom table or grouped cut.
Python
Run notebook analysis against the same data and limits for custom screening logic, debugging, or cross-lot work without rebuilding context again.
Move the limits. Watch the review change.
This is the part many wafer sort teams still do in Excel or JMP, then rebuild in slides and spreadsheets for review. Here the charts and the review move together, and the same scoped lot can continue into SQL and Python notebooks.
Drag the limit lines. Whether the proposal came from manual what-if, PAT, or DPAT, the review stays attached to the same evidence.
Example review
Lot 18A7 ยท SKEW_CLK
Scatter One failing cloud, one moving review window
Histogram The distribution stays fixed while the review window moves
Baseline window Proposed LSL Proposed USL In window Outside window
Workflow
One wafer sort workflow from ingest to sign-off.
The product is opinionated about one thing: the review should stay attached to the same lot, the same screening logic, and the same evidence all the way through the decision.
Ingest
Upload STDF, review parsed metadata before ingest, and keep each lot as a clear snapshot instead of a pile of files.
Investigate
Use linked scatter, histogram, wafer, and spatial views to isolate the population that matters.
Screen
Generate PAT and DPAT screens or stage manual limit changes without leaving the same notebook context.
Decide
Resolve exact die-level impact, move the proposal through review, and keep the audit trail attached to the same evidence.
Export
Send operations a CSV die list and management a review artifact from the same proposal.
What Ships
Built for wafer sort review, not just exploration.
Stratum is strongest when the analysis has to persist past one engineer and one session.
PAT and DPAT screening
Generate screening limits from the current lot or from reference lots, then carry that screening provenance through the rest of the review.
Exact die-level review
What-if proposals resolve to exact outcomes with rescued dies, new fails, top impacted tests, and top impacted wafers.
Audit trail and sign-off
The proposal, rationale, review status, and exported artifact stay in versioned notebook documents instead of screenshots and side-channel threads.
RBAC and controlled workflows
Role-based spaces, controlled notebooks, and approval boundaries keep the workflow usable for both the engineer building the proposal and the manager approving it.
Ingest review and alerting
Review parsed lot metadata before ingest, keep snapshot lineage intact, and surface ingest and excursion alerting without bolting on a separate monitoring tool.
Linked lot context
Analyze, SQL, Python, and Decide stay attached to the same lot, selection, snapshot, and active limits instead of rebuilding context at every handoff.
Who It Is For
Built for the engineer making the proposal and the manager signing it.
Yield and test engineers
Build the proposal, inspect the exact impact, and stop rebuilding the review in Excel, JMP, and side-channel spreadsheets.
Engineering managers and quality leads
Review one artifact instead of reconciling a spreadsheet, a slide deck, and an email thread that all drift apart.
Need quick local inspection instead of a shared review workflow? Use stdf.io.