LIFT_PROTOCOL_01 // PROTOCOL_SPEC
Control Group Design
A defensible experimental design that isolates whether paid media changes outcomes—so lift claims are real, not attribution theatre.
Without a Control Group, You Can’t Claim Lift
If every account is exposed, you can’t know what would have happened without media. Control groups create the comparison needed to prove incremental impact on win-rate and revenue. Without them, you are merely correlating spend with organic success.
CRITICAL FAILURE MODES
- ✕ Attribution models show correlation, not causation
- ✕ Executives reject marketing impact claims as unverifiable 'black boxes'
- ✕ Teams optimize to clicks because revenue impact cannot be isolated
Incrementality Design Logic
EXPERIMENT INPUTS
Target account universe (Tier 1–3)
Randomization / matching method
Suppression rules (no exposure cohort)
Exposure window length
Outcome window (win/loss + stage movement)
→
CONTROL OUTPUT
Exposed vs Control Cohorts
Logic: [Randomized or Matched Cohorts] → [Exposure Window] → [Outcome Comparison]
Artifacts You Receive
- ✓Control Group Governance Policy
- ✓Randomization / Matching Method Spec
- ✓Suppression + Inclusion Schema
- ✓Outcome Window + Measurement Definitions
Implementation Steps
WK 1
Define universe and eligibility; select randomization or matched-pair methodology.
WK 2
Implement suppression rules and cohort flags in CRM and media platforms.
WK 3
Launch experiment and validate cohort integrity and exposure compliance.