Houston AI

Friction Drivers · Feb 28–Mar 30, 2026 · ET

How this works

Highlights drivers of negative reactions and quality issues.

  • Uses template-level frustration rates.
  • Buckets are aggregated over the date range.
Limits
Drivers are heuristic, not causal proof
What to do next
Focus on high-exposure drivers first.

Where negative reactions are coming from: templates, timing, sequence, and response delay.

Negative reactions are explicit negative inbound SMS replies within the reply window. A template can have low negative reaction even if many replies are logistics/clarification.

Metrics are computed per lead exposure (unique parent_id) to avoid chatty leads skewing results.

Use this to improve messaging and timing. Not for eligibility decisions.

  • Use this to decide whether to rewrite wording or adjust when / when not to send.
  • Start with the highest exposure + highest negative rate segments to get the biggest impact.
  • Use Evidence to validate patterns before changing templates globally.

Reply rate
Percent of exposed leads who replied within the reply window.
Negative Reaction Rate (of leads)
Percent of exposed leads whose reply included complaints or opt-out signals.
Silence
Percent of exposed leads with no reply in the window.