Signals arrive 13–20 days before dashboards. Here's how we measured that

A dashboard shows you the outcome, not the signal that came before it. We backtested a real client's history to measure exactly how much time that gap wastes.

Every growth team has dashboards, and every team believes it is watching the data. The problem isn't whether you look. It's that a dashboard only lights up after the outcome lands. By the time GMV dips, CPM climbs or returns spike, the behavior that caused it finished happening two or three weeks ago.

How the backtest ran

We took thirteen months of one real account's history and ran the detectors backwards: pretend the engine had been connected then, see which day it would have fired, and compare against the day the client actually noticed the same thing. The gap is the wasted time.

Why a range instead of a single number? Because the lead time depends on the signal. Stockout risk runs longest — inventory velocity is a slow variable with a clean trend. Creative-spread runs shortest: paid data was already daily, nobody was reading it for dispersion.

Thresholds have to come from your own history

A common mistake is to borrow industry benchmarks as thresholds. "Alert when CTR drops below 1%" is barely comparable between two brands. Our detectors calibrate to the brand's own historical distribution: a signal fires when this brand's distribution breaks, not when it crosses somebody else's benchmark. Where data is missing, the detector stays silent rather than inventing a number.

Lead time is only worth something if it becomes an action

Knowing twenty days earlier and acting twenty days earlier are different things. That's why execution and measurement live inside the same engine: an investigation ends in an action with an acceptance contract attached — metric, baseline window, measurement window, guard conditions, all fixed before anyone confirms, graded automatically when the window closes.