Opening Line Forensics V2

2,933 creatives across NRSC, RNC, NRCC — Feature-first methodology | Feb 2026

2,933
Creatives Analyzed
17
Features Extracted
1
Universal Predictor
r=0.041
Strongest Universal r
V2 Key Finding: Almost no linguistic feature in the opening line strongly predicts performance across all clients. The only universally positive feature is personalization tokens ([firstname]) — and even that is weak (r=0.041). V1's "curiosity gap" finding was an NRSC artifact that doesn't replicate.

Feature Correlations with Performance (by Client)

Feature Correlation Table

FeatureOverall rNRSCRNCNRCCUniversal?
has_personalization+0.041+0.023+0.087+0.089✓ YES
curiosity_gap+0.119+0.158-0.006-0.020NRSC only
has_name_ref+0.035+0.051+0.033-0.060Mixed
exclamation_count-0.037-0.039-0.051-0.013~Negative
is_exclamatory-0.032-0.034-0.050-0.008~Negative
is_question-0.026-0.028-0.032-0.010~Negative
char_count+0.017-0.002+0.077+0.070Mixed
cap_ratio-0.007+0.014-0.054-0.074Mixed
has_urgency-0.011-0.018-0.006+0.024NOT a factor
personal_intro-0.003+0.005-0.033-0.030Doesn't help

Performance by Character Count

Winners vs Losers (Feature Averages)

Personalization Effect by Client

Client Baselines

ClientnMean PollsMedianStdev
NRSC2,0263.601.014.42
RNC4371.761.02.89
NRCC4702.511.04.65

🏆 Top Openers (Cross-Client)

NRSC · Cr-1243 · z=23.5
There's something we haven't told you yet, [firstname|Friend].
342 polls
NRSC · Cr-0218CPA · z=19.7
LIVE POLL: BIDEN-HARRIS APPROVAL.
287 polls
RNC · Cr-992SMS · z=12.9
I'm choosing [firstname|you] as the newest member of my Vice Presidential Cabinet.
39 polls
NRCC · Cr-416SMS · z=9.8
This silence will cost us, [firstname|Friend]
48 polls
NRSC · Cr-1267 · z=10.5
Two things, [firstname|Friend]:
155 polls
RNC · Cr-1021SMS · z=7.0
Elon Musk just issued a dire warning to the swamp.
22 polls

💀 Bottom Openers (What Losers Look Like)

NRCC · Cr-19
BREAKING!
1 poll
NRCC · Cr-22
Your Membership has EXPIRED!
1 poll
NRCC · Cr-13
Hey, [firstname|Friend].
1 poll
NRCC · Cr-24SMS
It's Mike Pompeo.
1 poll
NRCC · Cr-12SMS
You've won the Golden Gavel Award for [zipcode]!
1 poll
NRCC · Cr-16
[firstname|Supporter], I am all in for President Trump.
1 poll

V1 Claims vs V2 Reality

V1 ClaimV2 VerdictEvidence
"Curiosity gap = 4.1x lift"NRSC artifactRNC r=-0.006, NRCC r=-0.020
"Personal intro is highest category"Doesn't replicateRNC r=-0.033, NRCC r=-0.030
"Name-drop then speak = 2.4x"MixedNRSC +0.051, RNC +0.033, NRCC -0.060
"Urgency words work"NOT a differentiatorTop and bottom use urgency equally
"Personalization helps"✓ Confirmed (weakly)Positive in all 3 clients

📋 Actionable Recommendations (Evidence-Backed)

1. Use [firstname] tokens. Only universal positive predictor. Small effect but free to implement. r=+0.041 overall, positive in all 3 clients.
2. Keep openers 50-150 characters. Under 30 chars underperforms (z=-0.054). Over 150 chars crashes (z=-0.178). Sweet spot: 51-70 chars (z=+0.065).
3. Don't over-punctuate. Fewer exclamation marks correlate with better performance across all clients. "BREAKING!" is a loser pattern.
4. Skip questions in openers. Negative correlation in all 3 clients. Statements outperform questions.
5. Focus on body copy, not opener tricks. Opener features explain <1% of performance variance. The remaining 99%+ is body copy, targeting, timing, and landing page.