What Is Derived Context?
Derived context is state computed from raw events rather than stored directly: aggregates, velocity counts, running exposures, session summaries, ML features, embeddings. It is the form in which automated decisions actually consume the world — a fraud model reads transactions-per-minute, not individual transactions. Because derivation requires computation, derived context always trails the events it summarizes, and where that computation runs determines how far.
Updated
What is derived context?
The term is part of the context-gap vocabulary Tacnode uses to isolate the state that makes real-time decisioning hard. Raw facts are easy: a committed row is current the moment it commits, and a single stored record protected by a row lock needs no special infrastructure. The difficulty concentrates entirely in context that must be computed — summed, windowed, joined, embedded — before a decision can use it.
Derived context is what separates decisions that a plain database handles atomically from decisions with a structural freshness problem. Checking "does this exact record exist?" is a lookup. Checking "has this account moved more than $10k in the last ten minutes across all channels?" requires derivation over many events — and the derivation pipeline, not the database, becomes the bottleneck through which reality reaches the decision. The full taxonomy — which signals must be pre-computed, which can be computed on demand, and what each choice costs — is developed in the derived context post.
How derived context is produced
Three production modes, with different freshness properties:
- Batch derivation — scheduled jobs recompute features hourly or nightly. Cheap and simple; the derived value lags by up to the full schedule interval.
- Streaming derivation — a stream processor maintains the value continuously (event → broker → job → serving store). Lag drops to seconds, but each infrastructure hop adds delay and an independent failure surface, and lag stretches under burst via backpressure.
- In-engine derivation — the system holding the data maintains the value itself, via incrementally updated views that converge in sub-second time, or computes it on demand at query time against committed data. Fewer hops; the trade-off moves to engine capability and load.
A useful discipline is the derivation inventory: for each decision path, list every derived input, its production mode, and its measured event-to-readable lag — that lag is each signal's contribution to the decision's context gap.
Why derived context matters
Derived context is where context gaps concentrate. Systems rarely fail because a raw record was wrong; they fail because a summary of many records hadn't caught up with the newest few — the velocity counter missing the burst, the exposure total missing the concurrent draw-down. And because each derived signal is produced by its own pipeline at its own pace, decisions reading several of them inherit inconsistent reads between signals on top of each signal's lag.
The concept also draws the qualification line for real-time infrastructure generally: if a decision needs only a stored fact, ordinary transactional machinery suffices; the case for specialized context infrastructure begins precisely where decisions depend on derived state under freshness, concurrency, and scale pressure — the argument threaded through semantic context and context under concurrency.
FAQ
Related terms
A freshness budget is the maximum staleness a decision can tolerate in its inputs — and it equals the worst input, since one slow signal spends it all.
A context gap is the difference between the state of the world an automated decision acts on and the actual state at the moment the decision commits.
Event sourcing stores application state as an append-only log of immutable events; current state is derived by replaying them. Covers mechanics and trade-offs.
Change data capture (CDC) identifies row-level database changes and delivers them to downstream systems as ordered events. Learn how log-based CDC works.
A columnar database stores data by column rather than by row, speeding analytical scans and compression. Learn how columnar storage works and when to use it.
A validity window is the interval within which a decision's context remains an accurate basis for action — typically 10ms to 1s for automated decisions.