Incremental Sign-ups From a Super Bowl QR Ad (48h)

CoinFactory ran a 60-second Super Bowl TV spot on 2025-02-09 with a QR code to a signup page. Successful sign-ups receive a $15 coupon. Estimate incremental sign-ups attributable to the ad in the first 48 hours and quantify uncertainty.

Provide a concrete plan that includes:

1. Identification

• Propose at least two distinct causal identification strategies (e.g., high-frequency time-series counterfactual with synthetic control, geo-lift/DiD with control DMAs, calibrated MMM over a short horizon).
• State the identifying assumptions explicitly for each method.

2. Data You Would Use

• Minute-level traffic/sign-ups for the prior 8 comparable Sundays.
• QR UTM-tagged sessions and device/IP de-duplication rules.
• Coupon issuance/redemption logs (coupon_id, user_id, issued_at, redeemed_at).
• TV air-times/GRPs by DMA.
• Press mentions timestamps.
• Bot-filtering heuristics.
• App store ranking changes.
• Site latency/error logs.

3. De-duplication and Leakage

• Handle multi-device scans, dark social reshares of the QR URL, bots, and post-game press coverage spillover.
• Explain how you’ll separate organic baseline from paid lift and how you’ll attribute delayed sign-ups within the 48-hour window.

4. Back-of-the-Envelope (Compute) Assume:

• 12,000,000 QR scans in logs.
• 40% remain after de-duplication.
• Landing→signup conversion = 22%.
• Baseline = 50,000 sign-ups/day (absent the ad).
• Press coverage added an 8% lift to baseline for the first 24 hours.
• A competitor ran a similar QR ad in 8 DMAs that represent 12% of our reach and cannibalized 25% of our QR traffic there.

Estimate incremental sign-ups and provide a 90% CI using a reasonable variance model; show each adjustment step (baseline subtraction, cannibalization, spillover).

5. Validation

• Cross-check with coupon redemptions (assume 20% redeem within 7 days) and with geo heterogeneity.
• Describe how to reconcile differences across methods and decide on the final estimate.