Sentinel Query Set Builder
Build your query set with statistically grounded sizing for ACF and SOV-AI measurement.
Step 1: Define Your Measurement Context
More competitors = more SOV-AI noise sources. Industry practice: 3-5 competitors.
Step 2: Understand Your Metrics
You'll measure two metrics. They behave very differently:
% of queries where you're cited
Your weighted share vs competitors
Select Query Count
More queries = detect smaller changes
Detection thresholds vary by your current performance — see tables below for specifics.
🎯 ACF Detection Thresholds
Varies by your current performance level
| Current ACF | Detectable Change | Relative Impact |
|---|
⚖️ SOV-AI Detection Thresholds (estimates)
Higher due to position variability + competitor noise
| Current SOV-AI | Est. Detection | Reliability |
|---|
Use SOV-AI for quarterly competitive context — it's too noisy for monthly decisions.
Why Variance Matters: The p(1-p) Formula
ACF is a proportion (e.g., "you were cited in 30% of queries"). The statistical variance of a proportion follows the formula p × (1-p), where p is your current rate. This determines how much random noise exists in your measurement.
Maximized at p = 0.50 (variance = 0.25)
Practical meaning: If your ACF is around 50%, you're in the "maximum uncertainty zone" — random sampling noise is at its peak. At 5% or 95%, the same sample size gives you more precise measurements.
Why SOV-AI Is Significantly Noisier
SOV-AI is a position-weighted ratio, not a simple proportion. With 4 competitors, it has 11 sources of variance:
Same citation, different position = 75% weighted swing (1st vs 4th)
Whether you're cited at all (binary)
4 competitors × 2 dimensions = 8 variance sources
If you're 1st, competitors can't be — creates covariance
Real-World Scenarios: How Position Shifts Mislead
Step 3: Build Your Query Set
Distribute 50 queries across five pillars