RFCs
We're bringing HeapDB to the systems you already run. Each RFC is a sidecar design. HeapDB reads a copy of your event stream, tells you which earlier events most likely caused what you're seeing, learns from every answer your team confirms, and writes its results back where you already work.
Designs we're building next. Each one ships with a runnable example on synthetic data, and HeapDB finds the planted cause. Your feedback decides what comes first. For what runs end to end today, see the use cases.
Outbreak source tracing
Which food that cases bought before falling ill most likely explains a cluster of foodborne illness?
Read the RFC: Outbreak source tracing →Common point of compromise
Given the first cards that report fraud, which merchant, terminal or website most likely leaked their data, and which other cards exposed there will be next?
Read the RFC: Common point of compromise →Delayed conversion credit for recommendations
Which earlier recommendation exposure most likely earned this purchase, when purchases arrive hours to weeks later?
Read the RFC: Delayed conversion credit for recommendations →Trip incident attribution
When a trip ends with a low rating or a complaint, which event in that trip's timeline most likely caused it?
Read the RFC: Trip incident attribution →Exposure attribution for creative work
When a monitoring service finds an unauthorized copy of a work, which earlier exposure most likely leaked it, and which kinds of sends are risky before the next release?
Read the RFC: Exposure attribution for creative work →Hospital transmission routes
When sequencing links patients' infections into a cluster, which shared unit, room, procedure, device or care team most likely carried it?
Read the RFC: Hospital transmission routes →Failures after maintenance
An asset failed soon after work was done on it. Which recent part, lot, crew, procedure or firmware most likely explains it, and which other assets got the same work?
Read the RFC: Failures after maintenance →Negative feedback and regret
Which recommended exposures most likely led to this unsubscribe, hide or return, and which surfaces and items should a guardrail demote?
Read the RFC: Negative feedback and regret →Security alert triage
When an account takeover is confirmed, which earlier event was most likely the entry point, and how risky are new events that look like it?
Read the RFC: Security alert triage →
The fit test
Here's how we decide whether a problem is HeapDB-shaped. Someone keeps asking “this happened to this entity, so which earlier events most likely explain it?” and they need a ranked, explained answer right then.
| Property | Question to ask | Strong fit | Weak fit |
|---|---|---|---|
| Effect | Is there a timestamped effect per entity? | A fraud report on a card | A quarterly KPI |
| Candidates | Are there many earlier events per entity, with features? | Every merchant a card visited | One event per entity |
| Lag | Does the delay from cause to effect have a shape? | Incubation days, conversion delay | Instant, or anything goes |
| Learning | Is there confirmation, rejection, or meaningful absence? | A lab confirms the food, a bank confirms the breach | No outcome is ever known |
| Per incident | Is the answer needed per case, on demand? | An investigator opening a case | Only a monthly report |
Fit map
We placed every use case in our catalog by how well it fits HeapDB's model and how directly the answer becomes a feature someone would pay for. These placements are our judgment calls.
Planned
We haven't written RFCs for these yet. If you want one sooner, tell us on the use-case board.
- Mobility
Driver churn and supply drop
Which pay changes, long pickups or app problems most likely preceded a driver going inactive?
- Mobility
Delivery lateness
Was this order late because of merchant prep, courier batching, routing or weather?
- Health
Drug safety signals
Which drug starts, dose changes or co-medications most likely explain this adverse event, given its time to onset?
- Health
Personal symptom triggers
Which meals, sleep or stress most likely triggered this migraine or flare? Covered by the personal health use case.
- Recommendations
Because-you explanations
Explanations for recommendations grounded in learned links instead of embedding similarity alone.
- Industry
Yield excursions
Which tool, chamber, recipe change or consumable lot most likely explains this run of low-yield wafers?
- Industry
Cloud cost spikes
Which deploy, config or autoscaling change most likely explains this jump in the bill?
Know a domain where this fits? Post the event stream and the question on the board. We write the next RFCs for the most-voted ideas.
Share your use case