whystrohm-voice-scorer
Measure voice drift between your website and social content. Find where the brand is leaking.
Your website sounds different from your LinkedIn. Find out by how much. Drift is measurable; fix what's actually broken.
The Problem
Brand voice degrades on social. The agency posts. The intern posts. The founder posts at 11pm. Three weeks later your LinkedIn doesn't sound like your homepage.
What It Does
- 01Compares two voice profiles and outputs a drift score
- 02Identifies specific divergence points (vocabulary, cadence, structure)
- 03Recommends which channel needs voice retraining
What it measures
Drift between two voice profiles is computed as:
- Vocabulary overlap (Jaccard similarity)
- Cadence deviation (sentence length distribution)
- Structural alignment (hook/proof/CTA ordering)
- Tone delta (authority + emotional temperature axes)
- Forbidden word violations
Score is 0-100 (100 = identical voice). Below 70 = real drift. Below 50 = the channel is a different brand.
Why drift matters
Voice drift is a quiet killer. Each post in isolation looks fine. Across 30 posts the brand loses its center. Customers start to read your social as "a marketer wrote this" instead of "the founder wrote this."
The fix isn't more guidelines. The fix is measurement. Measure the drift weekly, retrain whichever channel is leaking. That's the whole loop.
Install
git clone https://github.com/whystrohm/whystrohm-voice-scorer.git ~/.claude/skills/whystrohm-voice-scorer
Full docs on GitHub →
How It Composes
Runs after voice-extract. Compares two extracted profiles (website vs LinkedIn, or current vs 6-months-ago). Pairs with audit for full diagnostic.
Related Skills