Definition
Well Tested's internal graph of schema changes, table observations, and incident reports — linked together over time to form a memory of how systems fail. The Memory Graph enables recurrence risk detection: 'this table has broken 3 times in 90 days.' It is built by ingesting data from DiffRunResults, IncidentReports, and SchemaChange records, and updated automatically as new diffs run and new incidents are recorded.
Why it matters
Without a memory of past failures, every incident is a fresh discovery. The same column that caused a billing outage in January will cause another one in April if nobody remembers. The Memory Graph makes institutional knowledge durable: it remembers what broke before, so teams don't have to re-learn the same lesson.
How Well Tested handles it
The Memory Graph is populated by the backfill process (which ingests historical DiffRunResults) and updated continuously as new diffs run and incidents are recorded. At release time, the recurrence risk query traverses the graph to find failure patterns associated with the tables being changed. The graph is stored in Prisma as SchemaChange, TableObservation, IncidentReport, and SchemaChangeIncidentCorrelation models.
Related terms
No commitment. 20 minutes. We review your release workflow and show you what's possible.