Est. reading time: 5 minutes

Send timing is a real lever, and Mailchimp’s Send Time Optimization is a real tool for pulling it. But the post most brands read about it was written for 2020, and following that advice today means optimizing against a signal that’s been partly fabricated for years. Before STO can help you, you have to understand what happened to the data it runs on, because the brands celebrating open-rate lift from timing changes are, in many cases, celebrating phantom opens from Apple’s servers.

Here’s how the feature works, what broke underneath it, and how to use it in a way your revenue numbers will actually corroborate.

What Mailchimp Send Time Optimization does

STO analyzes your audience’s engagement history and recommends the send time when engagement is most likely, sparing you the folk wisdom about Tuesday at 10 a.m. Paired with Mailchimp’s timezone delivery, that timing lands in each recipient’s local clock, so a West Coast subscriber isn’t served breakfast email at 6 a.m. because your team works Eastern. The prediction is only as good as the engagement data feeding it, which is where the caveat lives, because the engagement signal email marketing spent two decades trusting has a corruption problem.

The open-rate problem you have to handle first

Open tracking works by loading an invisible pixel when an email is opened. Apple’s Mail Privacy Protection, enabled by default for Apple Mail users since 2021, preloads that pixel through Apple’s proxy servers when the email arrives, whether or not the human ever opens it. The consequences, per Mailchimp’s own documentation, are that emails to MPP users report as opened regardless of activity, open counts inflate, and open-based location, device, and timing data for those contacts becomes unreliable. Apple Mail’s share of many consumer lists is large enough that this isn’t an edge case. It’s the water you’re swimming in.

Three operational corrections follow before STO enters the picture. First, exclude Apple MPP and other bot activity from your reports, which Mailchimp supports with a toggle, so your baseline open metrics describe humans. Second, audit anything triggered by opens, because resend-to-non-openers, open-triggered journeys, and engagement segments defined by opens all misfire when half the list “opens” everything automatically, and Mailchimp itself recommends shifting those criteria to clicks and purchases. Third, demote opens in your own hierarchy of truth. Clicks and revenue are actions a human chose, and they’re the metrics every timing decision below gets judged against.

None of this makes STO useless. It still draws on non-Apple engagement and broader behavioral signal, and timing still moves results. It means you evaluate STO like an adult, by clicks and revenue per send, not by the open column it can inflate without helping you.

Feed it clean data

Whatever signal survives the privacy era, protect it. Prune chronically inactive contacts, defining inactive by absence of clicks and purchases rather than absence of opens, dedupe records, and capture timezone at signup wherever the form allows. Connect your store so purchase and browse events enrich each profile, giving Mailchimp engagement signal that MPP can’t touch. And segment before you blast, by lifecycle stage and purchase behavior, so the timing model isn’t reconciling a VIP repeat buyer and a two-year-dormant freebie signup inside the same send.

Then establish a baseline before flipping anything on. Document click rate, MPP-excluded open rate, and revenue per thousand sends by campaign type and segment, and UTM-tag everything so the results trace into GA4. Lift you can’t distinguish from variance isn’t lift, and the baseline is what makes the difference measurable.

Test the clock like you’d test anything else

Treat STO as your control and challenge it. Run tests where send time is the only variable, one arm on STO, the other on a fixed slot your own reporting suggests might rival it, sent simultaneously to randomized, comparable segments so nothing else contaminates the read. Decide the verdict criteria upfront, and make click rate and revenue per recipient the primary metrics, with MPP-excluded opens as a supporting signal at most, since an open-judged send-time test in 2026 is partly a test of how many Apple users you emailed. Require a minimum sample and a consistent result across two or three sends before declaring anything, because a single Tuesday proves nothing about Tuesdays.

Operationalize what wins. Document the best windows by segment and campaign type, bake them into templates and the automation workflows that run without anyone scheduling them, and keep a small ongoing holdout so you can see whether the advantage persists. Retest quarterly, since audiences drift, and a locked-in window is a hypothesis with a shelf life, not a law.

Measure it in money

Timing work justifies itself in one chain. Incremental engaged recipients times click-through rate times conversion rate times average order value. Track revenue per thousand sends by segment, alongside unsubscribes and spam complaints, so you’re confirming profitable growth rather than louder sending. A simple timing dashboard per key segment, send window, click lift, revenue per recipient, with major campaigns and seasonal events tagged, turns “is this working” from a shrug into a trendline.

Then scale it with restraint. Roll STO across recurring newsletters, promotions, and automations, but hold frequency caps in place even when the model keeps finding engagement, because fatigue costs compound quietly. And when a segment underperforms no matter what the clock says, believe it. The minute of delivery is the smallest variable in email, sitting downstream of the list, the offer, and the funnel it all feeds. Get the signal honest, let STO individualize what it can, judge everything in clicks and revenue, and timing becomes what it should have been all along, a quiet compounding edge rather than a vanity metric with a scheduling feature attached.