Every team optimising a messaging funnel starts in the same place: the message copy. Then the offer. Then the segments. Almost nobody starts with the clock, which is the one number that gates all three.
What "first response time" actually means
It is the gap between a person sending their first message and a human or an automation sending something useful back. Two words there do a lot of work.
First means the very first inbound, not the first one an agent happened to open. Useful excludes the auto-reply that says "thanks, we will get back to you", which is a receipt, not a response. A funnel that counts receipts as replies reports a median of four seconds and converts like a funnel with a median of four hours.
The measurement trap
If your dashboard shows a first response time under ten seconds, check what it is counting. An acknowledgement template is almost always the thing being measured, and it moves the number without moving a single deal.
Three fixes, in the order they pay off
From slowest to fastest
Fix one
Answer the four questions everyone asks
Pull the last 500 inbound conversations and cluster the opening messages. In almost every business, four to six intents cover 70% of them: price, availability, location, timing. Those do not need a person. They need a correct answer in under two seconds.
Fix two
Route the rest by intent, not by round robin
Round robin optimises for fairness between agents. Intent routing optimises for the customer. A pricing question that lands with someone who cannot quote gets a slow answer no matter how fast they typed it.
Fix three
Make the handoff carry context
When an automation hands a conversation to a person, everything the customer already said has to travel with it. Asking a customer to repeat themselves resets the clock to zero, and they feel it as a slower response than if nobody had answered at all.
What to measure instead of averages
An average first response time is dominated by the tail. One conversation answered on Monday for a message sent on Friday moves the mean by more than fifty fast replies move it back.
Track the median and the 90th percentile, split by hour of day. The gap between them is where the fix lives.
| Metric | What it tells you | What to do with it |
|---|---|---|
| Median FRT | The typical experience | Target under 60 seconds for automated intents |
| P90 FRT | The bad experience | This is the number that generates complaints |
| FRT by hour | Where your coverage breaks | Automate the hours you cannot staff |
| Handoff rate | How often automation gives up | Rising rate means intents have drifted |
Common questions
Only when it pretends to be a person. An answer that is correct, specific and obviously automated outperforms a vague human reply sent an hour later, in every test we have seen.
Under 60 seconds median for the intents you have automated, and under 15 minutes at P90 during staffed hours. Chasing lower than that usually means optimising the acknowledgement, not the answer.
Related reading
See it on your own numbers
Connect a channel and the response-time report fills itself in.
