Skip to main content
Every signal is evaluated over a specific time window. Some fire the instant a field changes between two crawls. Others only fire once a pattern holds across weeks or months. Knowing the window tells you how fresh a signal is and how much weight to give it, a single-snapshot change is immediate but noisy, while a 90-day pattern is slower but far more reliable. CUFinder evaluates signals across five practical windows.

Week (7 days)

Short-fuse patterns that need to be caught fast, like a cluster of crisis posts.

1 month (30 days)

Rolling-activity windows: posting cadence, engagement shifts, and the nightly momentum score.

3 months (60 to 90 days)

Quarter-scale patterns: vacancies, consolidations, churn spikes, and most composite events.

6 months (180 days)

The longest patterns, like the pre-IPO buildout that unfolds over two quarters.
Two more windows sit outside the week-to-six-month scale: an instant snapshot diff (the change is detected the moment two crawls differ) and a 12-month lookback used by the first-in-function and first-in-geography signals.

​
Snapshot (instant)

These fire the moment a tracked field differs between two consecutive crawls. There is no waiting window, the change is the signal. They are the most immediate but also the most sensitive, so filter them by magnitude. This is the largest group. 71 signals use this window:

​
Crawl-over-crawl baselines

These compare the current crawl against a short rolling baseline (the previous 4 or 6 crawls) to catch anomalies like spikes and surges. They fire only when current activity breaks sharply from the recent norm. 9 signals use this window:

​
Week (7 days)

Short-fuse patterns evaluated over a rolling 7-day window. Fast-moving by design, so they are caught and acted on quickly. 1 signal use this window:

​
1 month (30 days)

Rolling 30-day windows, mostly social-activity rhythms compared against the preceding 30-day periods, plus the nightly momentum score that decays over 30-day steps. 4 signals use this window:

​
3 months (60 to 90 days)

Quarter-scale patterns. A vacancy that stays open 60 days, three closures or departures inside 90 days, or a composite event whose parts must land within a 60-to-90-day window. Slower to fire, but high-confidence. 9 signals use this window:

​
6 months (180 days)

The longest patterns in the catalog. These need two quarters of evidence before they fire, which makes them rare and strong. 2 signals use this window:

​
12-month lookback

These check whether something is happening for the first time in a year, a first job posting in a function or geography. The window is a backward lookback rather than a waiting period. 3 signals use this window:

​
Using timeframes in practice

Pair the window with magnitude. Act fastest on short-window, high-magnitude events (a funding round, a crisis cluster), and treat long-window composites as durable, high-trust signals worth a more considered approach. The longer the window a signal clears, the longer your outreach stays relevant.

See how detection works

The full pipeline behind snapshots, baselines, and windowed triggers.