Every morning, you face the same challenge: dozens of metrics across multiple platforms, and limited time to figure out what actually needs your attention. Crux solves this by automatically analysing your metrics and telling you exactly which ones are performing unusually well or badly.
The problem we solve
“Did CR drop 15% yesterday because we launched 5 new ads… or was it because it was a Monday?”
How it works
The health system answers a simple question for every metric: “Is today’s value normal for this day of the week?”
Mondays perform differently to Saturdays. Black Friday looks nothing like a typical Tuesday. Rather than comparing today against a simple average, we compare each day only against the same day of week from the past 12 weeks.
For example, if today is Wednesday, your conversion rate is compared against the previous 12 Wednesdays—not against all days combined.
The lookback window
We analyse 12 weeks of historical data to establish what’s “normal” for each metric on each day of the week. This provides enough data points (12 per weekday) to calculate reliable baselines while remaining responsive to recent trends.
A minimum of 4 data points is required for health indicators to show. For new metrics or those with sparse data, you’ll see a neutral status until sufficient history accumulates.
Percentile-based detection
Rather than using arbitrary thresholds, we calculate where today’s value sits within the historical distribution. If your revenue today is higher than 85% of previous same-weekday values, that’s noteworthy. If it’s lower than 95% of previous values, that’s concerning.
This percentile approach automatically adapts to each metric’s natural variance, no manual threshold configuration required.
What you get
