Definition
SQAB measures the correlation between measurement signal quality (inverse of modeled conversion ratio) and AI-driven budget allocation changes across channels, revealing whether optimization algorithms favor measurable channels over effective ones.
The Measurement Gap
AI-driven budget optimization systems learn from feedback loops. When a channel provides clear, deterministic conversion signals, the optimizer can confidently attribute results and allocate accordingly. When a channel loses signal quality (higher modeled conversion ratios, noisier attribution, less deterministic feedback) the optimizer receives weaker training data and tends to shift budget away. This creates a systematic bias: the AI spends where it can see, not necessarily where marketing dollars generate the most value. Legacy metrics like channel ROAS and CPA compare reported performance across channels, but they cannot detect whether performance differences are real or artifacts of differential signal quality. A channel reporting high ROAS with 90% observed conversions carries fundamentally different information than a channel reporting lower ROAS with 80% modeled conversions, but legacy optimization treats both ROAS numbers as equally reliable inputs to budget decisions. The result is a feedback loop: channels with degraded signal lose budget, which reduces sample size, which further degrades signal quality, which further reduces budget. Meanwhile, channels with strong signal receive disproportionate investment not because they are more effective but because they are more measurable. This is not a theoretical risk: it is an emergent property of any optimization system that learns from noisy signals with differential noise levels across input channels.
Decision Supported:
Is the AI budget optimizer systematically allocating spend toward channels with better measurement signal quality rather than toward channels with better actual performance?
Formula
Plain language: Rank each channel by its signal quality (1 minus the modeled conversion ratio for that channel) and correlate those ranks with changes in budget share. A high positive correlation means the optimizer is systematically shifting money toward channels where it has better measurement signal.
Notation:
SQAB = Corr(Signal_quality_rank_c, Budget_share_change_c)
where Signal_quality_c = 1 − MCR_c
Variables:
| Variable | Definition | Unit |
|---|---|---|
| Signal_quality_rank_c | Rank of channel c by signal quality, where signal quality equals 1 minus that channel's modeled conversion ratio | Ordinal rank |
| Budget_share_change_c | Change in channel c's share of total marketing budget over the measurement period | Percentage points |
| MCR_c | Modeled Conversion Ratio for channel c, representing the proportion of that channel's conversions that are modeled rather than observed | Ratio (0–1) |
Proposed formula, pending validation pilots.