Evidence Boundaries
What can be proven, what is inferred, and what is estimated
Most analytics tools present all numbers equally. A directly matched order and a modeled attribution receive the same visual weight. GravityAds introduces evidence boundaries: explicit labels on what can be proven with data and what requires inference.
Direct evidence
Shopify order exists. UTM parameters match a known campaign. Click ID can be traced. The profit attribution is a direct evidence chain from ad click to purchase to profit calculation. GravityAds marks these with high confidence.
Inferred evidence
The order exists but the UTM was lost (redirect chain, app browser, cross-device). GravityAds can infer the likely source through timing, landing page patterns, or session reconstruction, but labels this as inferred, not proven.
Estimated metrics
Some numbers require statistical modeling. LTV projections, cross-device attribution without deterministic matching, and platform-reported conversions that can't be verified against Shopify are labeled as estimates.
Why it matters
When your $50K/month ad budget decision depends on a metric, you need to know if that metric is fact or estimate. Evidence boundaries let you set different decision rules: use direct evidence for tactical changes, inferred evidence for strategic direction, and estimates for long-term planning only.