What Is a Validity Window?
A validity window is the interval within which the context an automated decision reads remains an accurate basis for that action. If the state a decision depends on can change materially within, say, 500 milliseconds, then context older than 500 milliseconds is not merely stale — it is invalid for that decision, regardless of how fresh it is by pipeline standards. For high-velocity automated decisions the window typically spans roughly 10 milliseconds to 1 second.
Updated
What is a validity window?
The term is part of the context-gap vocabulary Tacnode uses to make freshness a per-decision requirement rather than a system-wide average. "Real-time" is an ambiguous claim — sub-second for one team, five minutes for another. A validity window replaces the adjective with a budget: for this decision, on this entity, how old can the input context be before the action taken on it stops being correct?
The window's width is set by two things: state velocity (how fast the relevant state changes) and consequence (what a wrong action costs). A product-catalog lookup has a wide window — descriptions change rarely and errors are cheap. A margin call, card authorization, or agent tool-call against a shared limit has a narrow one — the state changes concurrently with the decision, and a wrong action commits money or an irreversible step. The interplay between window, velocity, and concurrency is formalized in the decision-time system model.
How a validity window works
The window turns freshness into a pass/fail comparison per decision path:
Decision validity window context age (p99) verdict
card authorization ~200 ms 2–3 s (velocity) invalid
margin liquidation check ~50 ms 300 ms (positions) invalid
loyalty tier on checkout ~1 min 5 s valid
weekly churn score ~1 day 1 h validThe comparison must use the tail age of context, not the median, and must account for bursts — the moments when both context age and state velocity spike together. A decision is protected only if its context's worst realistic age fits inside the window; a system can pass on averages and fail on every incident that matters.
Why the validity window matters
The concept forces the right engineering conversation. "We need fresher data" is unactionable; "this authorization path has a 200 ms validity window and its velocity feature has a 3-second p99 age" is a specific, falsifiable gap — a context gap — with a measurable size. It also explains why some staleness is fine: most data in most systems never needs sub-second freshness, and chasing it everywhere is waste. The window localizes the investment to the decisions where staleness converts into loss.
It also sharpens what "solved" means. Meeting a validity window is not achieved by faster hardware alone — it requires every input's pipeline lag and cross-system skew to fit inside the window simultaneously, under burst conditions, for every concurrent decision. That composite requirement is what separates decision-grade context infrastructure from generically fast analytics, as argued in why real-time decisions fail.
FAQ
Related terms
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.
State velocity is the rate at which the state a decision depends on changes. High velocity plus concurrency is what makes cached context structurally invalid.
Decision coherence is the property that concurrent automated decisions act on the same version of shared state, instead of contradicting each other.
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.
Cache invalidation is the process of removing or updating cached data when the underlying data changes. Covers TTL, write-through, and event-driven strategies.
A race condition occurs when a system's outcome depends on the timing of concurrent operations, letting interleaved reads and writes corrupt shared state.