ChatGPT said:

Great one — ATE and lift are closely related but not identical. Let’s pin down the difference.

🔹 Average Treatment Effect (ATE)

• Definition: The causal effect of treatment vs. control, expressed as a difference in outcomes.

ATE=Yˉ1 −Yˉ0

• Units: Same as the outcome (e.g. percentage points if outcome is conversion rate).
• Example:
  • Treatment conversion = 15%
  • Control conversion = 10%
  • ATE = 0.15 – 0.10 = 0.05 → 5 percentage points.

🔹 Lift (Relative Lift)

• Definition: The relative increase in the outcome due to treatment, compared to control.

Lift=(Y0 - Y1) / Yˉ0

• Units: Percentage (relative to baseline).
• Example:
  • Treatment conversion = 15%
  • Control conversion = 10%
  • Lift = (0.15 – 0.10) / 0.10 = 0.05 / 0.10 = 0.50 → 50% lift.

🔹 Key Difference

• ATE → absolute difference (how many points higher).
• Lift → relative difference (how much bigger relative to control).

✅ Interview one-liner:

“ATE is the absolute causal effect — the difference in outcomes between treatment and control. Lift is the relative effect — how much bigger the treatment outcome is compared to control, often expressed as a percentage.”

Would you like me to also map this to incrementality language (how marketers often call ATE = incremental conversions, and lift = % incremental gain) so you can sound aligned with Haus’s framing?