How do send-time optimization algorithms work?

Send-time optimization (STO) tries to deliver each email at the moment each individual subscriber is most likely to open it. Instead of sending a campaign to your whole list at 10am Tuesday, STO staggers delivery over a window and times each send based on when that specific subscriber has historically been most active.

What the algorithm is actually doing

Your ESP looks at each subscriber's open and click history: what time of day did they open previous emails, what day of the week, what device. From that history, it builds a predicted optimal send time for each subscriber. Some STO systems are simple (just find the hour they've opened most often and send then). Others use more sophisticated modeling that accounts for recency, open streaks, and day-of-week patterns.

After the campaign is triggered, delivery happens in batches over the send window. Subscribers with a predicted afternoon peak get their email mid-afternoon. Early-morning openers get it at 6am. The campaign might take 24 hours to fully deliver across your list.

The real-world caveats

STO requires data. If a subscriber has only received two emails from you, there's not enough history to predict their best send time. Most STO systems fall back to a default time for new or low-activity subscribers. This is worth knowing when you're evaluating the benefit for a small list.

Apple Mail Privacy Protection is a problem here too. If a significant share of your audience uses Apple Mail, their open timestamps in your ESP are when Apple's proxy fetched the pixel, not when the person actually read the email. An STO algorithm trained on that data is learning the wrong patterns for those subscribers. The optimization is only as good as the underlying engagement data.

When STO is worth using

STO tends to show the most impact for audiences with high geographic dispersion across time zones, for lists with strong engagement history, and for senders with enough volume that the send window doesn't create awkward timing gaps. For small lists (under a few thousand), A/B testing your own send times manually often tells you just as much.