Use Case: Regional Spot TV & Geo-Holdouts

Persona: Regional Marketing Director, Franchise Brand Lead, Econometrician
Objective: Measure localized Spot TV flights across specific states or Designated Market Areas (DMAs) without diluting the measurement signal across national web traffic, while validating zero false-positive lift in unexposed control regions.

---

The Challenge: The 10x National Signal Dilution Problem

When an advertiser runs regional Spot TV in selected test markets (e.g. California, Texas, and Florida):

1. Only viewers residing in those active markets are exposed to the broadcast.
2. If web traffic is evaluated at the national level (all 50 US states combined), 70–85% of measured traffic represents unexposed baseline noise from other states.
3. This dilutes the broadcast signal by 8x to 10x, sinking incremental visitor spikes below the statistical noise floor.
4. Furthermore, without unexposed holdout control markets, marketers cannot prove that observed visitor lifts were caused by TV rather than national promotions or organic seasonal trends.

SpotLift solves this with Regional Auto-Segmentation (region_col) and Holdout Control State Verification.

---

Step-by-Step Regional Workflow

Step 1: Pre-Flight Regional Alignment Check

Use spotlift validate to verify that both your airing schedule and your web telemetry contain matching geographic partitions:

spotlift validate \
  --spots us_regional_airings.csv \
  --web ga4_state_telemetry.parquet \
  --region-col state \
  --tz US/Eastern

The preflight auditor verifies that:

• Every market in the broadcast schedule exists in the telemetry stream.
• Unexposed control markets receive zero spend.

---

Step 2: Regional Attribution in Python

Pass region_col="state" to analyze_spots():

import spotlift as sl
import pandas as pd

# Load regional airings and state-level GA4 minute sessions
airings = sl.load_airings("us_regional_airings.csv", tz="US/Eastern")
telemetry = sl.load_telemetry("ga4_state_telemetry.parquet", region_col="state")

# Attribute with regional isolation (prevents national dilution)
results = sl.analyze_spots(
    airings=airings,
    web_telemetry=telemetry,
    region_col="state",
    window_minutes=8,
)

print(results.head())

---

Step 3: Test vs. Control State Scorecard

Group results by state to compare active broadcast markets against unexposed control states:

state_audit = (
    results.groupby("state")
    .agg(
        spots=("spot_id", "count"),
        spend=("cost", "sum"),
        lift=("incremental_lift", "sum"),
    )
    .assign(cpiv=lambda df: df["spend"] / df["lift"])
)

print(state_audit)

Benchmark Results from Synthetic Validation (generate_us_state_dataset):

State	Campaign Status	Spots	Media Spend	Attributed Session Lift	State CPIV	Econometric Note
Florida (FL)	Exposed (Test)	112	$179,160	+14,269.5	$12.56	High-efficiency test market
Texas (TX)	Exposed (Test)	114	$226,989	+17,261.7	$13.15	Reliable baseline lift
California (CA)	Exposed (Test)	188	$451,280	+33,556.4	$13.45	Strongest volume contribution
New York (NY)	Exposed (Test)	86	$176,119	+12,146.8	$14.50	Higher rate card market
Illinois (IL)	Control (Zero Spend)	0	$0.00	-26.6	N/A	Zero-mean symmetric noise ($\approx 0$)
Ohio (OH)	Control (Zero Spend)	0	$0.00	+32.2	N/A	Zero-mean symmetric noise ($\approx 0$)
Pennsylvania (PA)	Control (Zero Spend)	0	$0.00	-117.9	N/A	Zero-mean symmetric noise ($\approx 0$)

Key Methodological Takeaways

1. Zero False Positives in Control States: Because SpotLift preserves signed lift (clip_negative=False), unexposed states balance symmetrically around zero (-26.6, +32.2, -117.9). Tools using naive zero-rectification falsely report positive phantom lift in zero-spend control states.
2. Causal Incrementality Proof: The clear separation between active exposed states and zero-mean control states provides ironclad empirical proof to stakeholders that attributed web spikes were directly driven by local broadcast airings.