What changes in the inbox,
counted on client accounts.
No number on this page comes from a survey or a client's opinion. Each one is computed from the message history on their account and compared against the period before Setor AI was first used.
Accounts are described by niche, without names and without screenshots, so that neither the clients nor the people who messaged them are exposed. The niches come from their agent configuration, not from guesswork.
See the report on 1.58 million messagesAlmost half of all people got no answer
Read the full caseAn account that looked fast until somebody measured the tail
Read the full caseFrom two hours to one minute, three in the morning included
Read the full caseFive times the conversations, and almost everyone gets an answer
Read the full case7,742
calls booked with an actual scheduled time, across 48 client accounts, in the last 90 days. The spread is enormous and there is no point hiding it: the median is 16.3 calls per account per month, while the largest account books 421.7. It depends on reach, on the offer, and on how many people write in at all.
The method is here because without it these numbers are worth exactly as much as anybody else's claims.
- The cutoff is the first trace of Setor on the account, meaning the earliest contact in our database, not the date the official API was connected. Four of the five accounts studied were using our extension months before the API connection, so a window measured from the API date would compare a tool against the same tool.
- The before window is the year preceding that point, the after window is the last 90 days. Months separate the windows, so volume is confounded by account growth and is not comparable. Time based and percentage metrics are robust to this, and only those are placed side by side.
- One denominator for everything: conversations where the lead wrote something in text. Story reactions, bare emoji and empty threads drop out, identically on both sides of the comparison. The first version of this page mixed three different denominators, which made follow-up look flat on one account and volume look twentyfold on another. Both were artefacts, recomputed 20.08.2026.
- Follow-up is defined from messages alone: a message from the business preceded by another message from the business that the lead never answered, separated by a gap of at least thirty minutes.
- Pairs longer than 24 hours are dropped from response time. A reply after 26 hours is not a reply any more, it is a follow-up, and it is counted as one.
- Booked calls from before installation do not exist in any system, because bookings are only recorded once an account is connected. We established them by reading conversation content with a language model, using the same measure on both sides of the comparison.
Checking the classifier
Before computing anything we ran the classifier over 500 conversations from the period where the outcome is recorded in the system, and compared its verdicts against reality. Precision 98.3 percent, recall 92.8 percent, accuracy 95.6 percent. It errs toward undercounting, meaning it will miss a booking more often than invent one.
Want to see these numbers on your own account?
On the call we go through your inbox and show you how many messages currently go unanswered.