Audio Opener vs Text Opener — Mentor Mitch
Did the voice-note opener improve conversion? · 11,437 conversations analyzed
Verdict — yes, the audio opener is a clear win on bookings
Opening with a voice note books ~3× more calls per conversation (6.1% vs 2.2%) and converts replying leads ~4× better (12% vs 3%). It trades some top-of-funnel reply volume (51% reply rate vs 75% for text) for far hotter leads — net strongly positive. Recommendation: keep it as the default; monitor inbound volume since the lower reply rate makes the funnel top slightly thinner.
Reply rate = lead
sent ≥1 message after the opener. Booked = call_booked
custom property. Agent-handled convos only (agent_conversation=true).
Observational comparison (not a controlled A/B) — the like-for-like and same-month
April cuts control for time/seasonality.
Head-to-head
| Cut | Opener | Convos | Reply rate | Booked / convo | Booked / replied |
|---|---|---|---|---|---|
| All-time | AUDIO | 3,220 | 50.7% | 6.1% | 12.0% |
| TEXT | 8,217 | 74.8% | 2.2% | 3.0% | |
| Like-for-like since first audio (2025-02-18) |
AUDIO | 3,220 | 50.7% | 6.1% | 12.0% |
| TEXT | 7,940 | 73.9% | 2.1% | 2.8% | |
| April 2026 same month, both at scale |
AUDIO | 1,330 | — | 6.2% | — |
| TEXT | 2,797 | — | 3.5% | — |
Monthly trend
Volume + booking rate by opener, 2026
| Month | Audio convos | Audio booked | Audio rate | Text convos | Text booked | Text rate |
|---|---|---|---|---|---|---|
| 2026-01 | 0 | 0 | 0.0% | 169 | 9 | 5.3% |
| 2026-02 | 5 | 0 | 0.0% | 338 | 7 | 2.1% |
| 2026-03 | 44 | 0 | 0.0% | 4,027 | 9 | 0.2% |
| 2026-04 | 1,330 | 82 | 6.2% | 2,797 | 99 | 3.5% |
| 2026-05 | 1,830 | 111 | 6.1% | 278 | 17 | 6.1% |
Method & caveats
How this was measured and how to read it
What this is
An observational before/after-style cut, not a randomized A/B. Leads weren't randomly assigned audio vs text — so treat the magnitude as directional, not a precise lift estimate.
How the confound is controlled
Two cuts isolate the opener from time effects: (1) like-for-like — only conversations since the first audio opener; (2) April 2026 — both openers ran at high volume in the same month. Audio wins clearly in both, so the effect isn't seasonality.
Caveats
The March 2026 text row is a volume/attribution artifact and is excluded from the headline. Tiny 2025 audio samples (n=1–3) are noise and not used for conclusions. Reply rate being lower for audio is real and reported honestly — it's a quality filter, not a flaw.
Cross-check
Agus may have a parallel analysis — worth aligning definitions (opener = first agent message type; booked = call_booked) before circulating externally.