- The core metrics are pass rate, real-vs-safe ratio, and both of those broken down by segment.
- A healthy pattern is stable and settles over time; watch trends, not single-day points.
- Sudden shifts in the ratio usually mean something changed — a rule, a traffic source, or the mix of devices.
- Break every number down by device, geo and source; a blended number hides the segment that is actually broken.
The core metrics
Two numbers carry most of the signal. Pass rate is the share of visitors routed to the real page; the real-vs-safe ratio is how your traffic splits across the two surfaces over a period. Read together, they tell you whether the setup is separating traffic the way you intend. Everything else on a dashboard is context for these two. The meaning of pass rate specifically, and how too-strict or too-loose settings show up in it, is covered in tuning pass rate.
What a healthy pattern looks like
Health is about stability over time, not any single value. A healthy setup shows a pass rate and ratio that hold steady day to day and settle rather than jump around. That is why you read trends across a window, not one day's number — a single day can be skewed by a traffic spike or an odd source. On a newly launched domain the numbers are expected to move as it settles; that is the warmup period, and reading it correctly is part of warming a new domain.
Break every number down by segment
A blended number hides the segment that is actually broken. Always break pass rate and ratio down by device, geo and traffic source. The classic case: an overall pass rate that looks fine, but a mobile-only view that has collapsed — real in-app users routed to the safe page by a desktop-tuned rule. You would never see it in the blended number. The mobile-specific reasons this happens are in mobile traffic routing, and confirming which surface a segment reaches is in how to check a page.
Which anomalies mean something changed
A sudden shift in the ratio is the dashboard telling you something changed — and the useful reflex is to ask what. The three usual causes are a rule you edited, a traffic source that changed its mix, or a shift in the device split (for example a campaign that suddenly skews mobile). A sharp swing right after you changed a rule points at the rule; a swing with no change on your side points at the traffic. Either way, investigate before you re-tune, so you fix the cause and not a symptom. The disciplined tuning loop that follows is in tuning pass rate.
What not to read into the numbers
Not every wiggle is a signal. Small day-to-day movement, especially on low volume, is ordinary noise, and overreacting to it — retuning after every minor dip — is how a stable setup gets destabilized. The judgment is to distinguish a trend from a blip: a sustained move across several periods is worth acting on; a single-day flutter usually is not. Keeping this discipline is easier when you know the boundary you are operating within — the red-line checklist and the two-surface fundamentals in safe page vs real page keep the goal in focus.