Definition
MCR measures the percentage of reported conversions that are modeled estimates rather than directly observed events, tracked by channel and over time.
The Measurement Gap
When a marketing dashboard reports "1,247 conversions," the number appears authoritative. But in a post-cookie, post-ATT environment, a growing share of those conversions were never directly observed: they were filled in by platform statistical models, aggregated through SKAdNetwork postbacks, or inferred by Google's Attribution Estimation Model. Legacy metrics like conversion rate, cost per acquisition, and return on ad spend treat every reported conversion as equal regardless of whether it was deterministically tracked or probabilistically modeled. These metrics have no mechanism to distinguish a directly observed purchase from a statistically imputed one. The result is that marketers make budget allocation decisions, channel optimization choices, and performance comparisons based on numbers whose error bars are unknown and growing. A channel reporting 500 conversions where 400 are modeled carries fundamentally different decision risk than a channel reporting 500 where 50 are modeled, yet legacy ROAS and CPA metrics present both identically. Without decomposing reported conversions into observed and modeled components, the confidence level of every downstream metric built on conversion counts remains invisible.
Decision Supported:
When the dashboard reports a conversion number, what percentage was actually observed versus filled in by a model, and therefore how much confidence should we place in metrics derived from that number?
Formula
Plain language: Divide the number of modeled (estimated) conversions by the total reported conversions and multiply by 100. A companion metric adjusts the total by weighting modeled conversions by the model's historical accuracy.
Notation:
MCR = (Conversions_modeled / Conversions_total_reported) × 100
Companion: Confidence-Adjusted Conversion Count:
CACC = Conversions_observed + (Conversions_modeled × Model_accuracy_estimate)
Variables:
| Variable | Definition | Unit |
|---|---|---|
| Conversions_total_reported | All conversions reported by the platform or analytics system, regardless of observation method | Count |
| Conversions_modeled | Conversions derived from statistical modeling, attribution estimation, or aggregated postback mechanisms | Count |
| Conversions_observed | Conversions confirmed through deterministic, directly observed signals | Count |
| Model_accuracy_estimate | Historical accuracy rate of the platform's conversion model, measured against known-truth holdback tests | Ratio (0–1) |
Proposed formula, pending validation pilots.