Case study · Outbound telecalling
Two in five calls never connect. JoyzAI needed to know what happened on the other three.
JoyzAI runs an eleven-person outbound telecalling team. In six months on Sahay they made 6,542 calls, had 97% of the answered ones captured and analysed, and report conversion up 10% — mostly by finding out which parts of the script were not working.
The problem: a call floor with no memory
Outbound telecalling is a volume business, and volume businesses run on discipline. JoyzAI had the volume. What they did not have was a reliable record of what happened.
Three things were missing at once. There was no dependable lead stage tracking, so nobody could say with confidence where a given prospect actually stood. There was no rep performance tracking, so coaching was based on impressions rather than evidence. And there was no call analysis — the conversations happened, and then they were gone.
On top of that sat the tax every telecalling team pays: manual data entry. Reps typing notes after calls they had already moved on from, badly, if at all.
What six months of calls looks like
Between 16 February and 6 August 2026, eleven users made 6,542 calls across 2,540 leads, with 1,592 follow-ups scheduled from what was actually said on those calls.
| Outcome | Calls | Share |
|---|---|---|
| Answered and recorded | 3,652 | 56% |
| Unanswered outbound | 1,908 | 29% |
| Missed inbound | 655 | 10% |
| Rejected inbound | 201 | 3% |
| Answered, not recorded | 126 | 2% |
Two numbers matter here. The first is that 2,764 calls — 42% — never connected at all. That is normal for outbound, and it is exactly why the connected ones have to count.
The second is capture rate. Of 3,778 answered calls, 3,652 were recorded and analysed — 97%. Not a sample. Not the ones a rep remembered to log. Effectively all of them, captured from the reps’ own phones on ordinary SIM calls.
What changed
Once every conversation was on the record, three things became possible that were not before.
The script got fixed. With a corpus of thousands of calls rather than a handful of anecdotes, the patterns in what prospects actually asked — and where conversations died — became visible. That is a different exercise from a manager sitting in on a few calls a week.
Manual entry stopped. Lead records, stages and follow-ups now populate from the calls themselves. 1,592 follow-ups were scheduled without anyone typing them.
Requirement patterns emerged. Across two and a half thousand leads, what buyers repeatedly asked for stopped being folklore and became something you could read off the data.
JoyzAI report conversion roughly 10% higher since adopting Sahay. That figure is theirs — conversion is measured on their side, not ours.
The part worth stealing
The 97% capture rate is the quiet enabler behind all of it. Call analysis is only as good as its coverage: a tool that captures a third of your calls tells you about a third of your business, and the third it misses is not random. Fixing a script from a biased sample is worse than not fixing it.