“Send-time optimization” is one of those features that every SMS platform claims and almost none of them do well. Most are glorified time-zone splitters: they look at where the recipient lives and pick something between 10am and 2pm. That's not optimization; it's a default.
The interesting question isn't “when do people generally reply to texts?” It's “when does this person reply to your texts?”
What we actually do
Treply's send-time model is per-recipient and per-sender. For each contact, we look at the last 12 months of inbound responses to your organization — replies, link clicks, opt-ins, conversions — and compute a posterior probability distribution over hour-of-day. We then pick a 6-minute window inside that distribution where we're 80% confident the recipient is actively using their phone.
For new contacts (no response history), we fall back to a cohort model — same age range, same geography, same source channel — until we have enough first-party signal to switch.
Why 6 minutes
Carrier throughput. If you tell Treply “send 200,000 messages, each at the recipient's perfect minute,” you'll exceed per-second sending caps and the carrier will throttle you, scattering your perfect-minute sends across a much bigger window. We pre-bucket into 6-minute windows so we can plan throughput against your assigned codes and stay inside carrier rate limits.
Does it work?
On A/B tests against time-zone-only baselines, send-time optimization adds 11-14% to reply rates and 6-9% to attributed revenue. The effect is biggest for promotional messages and smallest for transactional alerts (which people respond to whenever they hit them).
What it doesn't do
Send-time optimization is not a substitute for sending less or sending better. If your message reads as a mass-blast, the perfect minute won't save it. Order of impact: cohort > copy > send time. We just thought you should know that we're doing the third one properly.