The short answer
For most paid leads, two attempts. In our audit of AI calls to 993 people who had filled a travel company's Meta lead form, 19 to 22 Sep 2026, the first attempt reached 331 people for the first time, the second reached 90 more and the third 48. Fourth and later attempts took 544 calls to reach 19 new people. Each extra attempt costs more calls for every new person it reaches, and every repeat call to someone who is not answering is a chance for a complaint. Revvox stops after two attempts, stops retrying the moment someone picks up, and never re-dials a person who has already talked; that lead goes to a person on your team instead. If a reached lead is worth a lot to you, a third attempt may still pay for itself, so run the arithmetic in this post on your own numbers before you choose a cap.
How many new people did each attempt reach?
The table follows every person through their attempts. “Picked up” counts any pick-up on that attempt, including people who had already been reached before. “Reached for the first time” counts only people who had not picked up on any earlier attempt. The last column is our own division: calls on that attempt divided by the new people it reached.
| Attempt | Calls | Picked up | Reached for the first time | Calls per new person |
|---|---|---|---|---|
| 1 | 993 | 33% | 331 | 3.0 |
| 2 | 655 | 29% | 90 | 7.3 |
| 3 | 543 | 26% | 48 | 11.3 |
| 4 | 367 | 22% | 16 | 22.9 |
| 5 | 142 | 31% | 3 | 47.3 |
| 6 or more | 35 | 6% | 0 | none reached |
The first attempt reached a new person for every three calls. By the third it took about eleven calls, and from the fourth attempt on it took 544 calls to reach 19 people who had not picked up before.
Why does pick-up stay high while new reach falls?
Look at attempt 5. Pick-up was 31%, higher than attempt 4, yet it reached only 3 new people. The people answering late attempts were mostly people who had already answered earlier and were being called again. A pick-up rate by itself hides this. The number that tells you whether an attempt was worth making is how many new people it reached.
How many people were called three times or more?
542 of the 993 people were called three or more times. 233 were called four times, 107 five times and 25 six times. From the lead's side that is a lot of missed calls from one number in four days, and most of those later calls bought nothing.
There is a compliance side too. Every repeat call to someone who is not answering is another chance for a complaint, and complaints can get a sender's numbers flagged or blocked. We do not know how many of those repeat calls annoyed anyone. We do know they raised the risk and did not raise reach.
What happens when you re-dial someone who already talked?
The audit found a second, quieter problem. 222 people were called again by the qualifying agent after they had already had a conversation. A person who has talked is finished with qualification; calling them again with the same questions is noise. The closing calls suffered from the same crowding: 62% of them went unanswered, because they went to people who had already been called three or four times that day.
Why two attempts and not three?
The third attempt still reached 48 new people, so this is a trade-off, not a law. We set the default at two for three reasons from the same data. The third attempt cost about eleven calls per new person against three on the first. Repeat calls carry complaint risk that grows with each one. And an attempt spent on a person who has not answered twice is an attempt not spent calling a new lead within minutes of the form, which is where reach is cheapest.
What does Revvox do today?
- Two attempts at most per lead, and retries stop the moment someone picks up.
- Attempts only happen inside the hours and days you allow.
- A person who has already talked is not re-dialled by the qualifying agent; they go to a person on your team or a closing call.
- Retries are free. A lead uses one credit once it is reached or exhausted, however many attempts it took.
How was this measured, and what are the limits?
- One customer, a travel company buying Meta leads, and four days of pre-sales calls, 19 to 22 Sep 2026.
- Pick-up here was counted before we separated machines from people. Voicemail greetings and call screeners inflated pick-up by about a third overall; the audit notes that the shape by attempt holds after the correction.
- The pattern is from one lead source. Leads from a different ad, industry or form may behave differently, so measure your own.
How should your team set its own attempt cap?
- Count, for your last few hundred leads, how many new people each attempt reached, not just the pick-up rate.
- Divide calls by new people reached for each attempt. That is your cost per new conversation at each step.
- Compare it with what a reached lead is worth to you. Stop at the attempt where the cost overtakes the value.
- Spread attempts across the hours your leads actually answer instead of repeating the same hour.
- Never send someone who has already talked back into the qualifying queue. Hand them to a person.
- Keep every attempt inside a legal calling window and honour opt-outs at once.