Email Send-Time Optimisation in 2026: What AI Changes and What Research Actually Shows

Email send-time optimisation (STO) — sending email at the specific time each individual subscriber is most likely to engage, rather than a single fixed time for the entire list — is one of the most marketed features in the ESP industry. Klaviyo, Salesforce Marketing Cloud, Mailchimp, and most major ESPs offer AI-powered STO that predicts each subscriber's optimal engagement window based on their historical email behaviour. The marketing around these features suggests dramatic engagement improvements. The actual research evidence is more nuanced — STO helps meaningfully in specific contexts and has minimal impact in others. In 2026, with Apple MPP inflating open data and Gemini AI changing how Gmail processes email, the timing question has become more complex than the ESP marketing implies.

Send-Time Optimisation: What It Is and How It Works

Traditional email campaigns are sent at a single scheduled time — the entire list receives the email at 9:00 AM Tuesday, for example. Send-time optimisation replaces this single-time approach with individualised delivery — each subscriber receives the email at the specific time the algorithm predicts they are most likely to engage based on their past email engagement history. Subscriber A might receive the email at 7:30 AM on Tuesday. Subscriber B receives the same email at 2:00 PM on Wednesday. Subscriber C receives it at 9:00 PM on Thursday. Every subscriber receives the same campaign email, but each receives it at their individually predicted optimal engagement time.

The algorithm: STO tools build a per-subscriber engagement profile by analysing historical open and click timestamps for each subscriber. If a subscriber consistently opens email at 8 AM on weekday mornings, the algorithm predicts that future sends should be delivered in that same window. The prediction is based on historical behavioural patterns — the assumption being that a subscriber's email engagement behaviour is consistent enough over time to predict optimal future delivery windows.

The technical implementation: when a campaign is "sent" with STO enabled, the ESP does not immediately deliver to all subscribers. Instead, it queues delivery, releasing batches to each subscriber as their predicted optimal window approaches. A campaign "sent" on Tuesday with STO enabled may have some subscribers receiving it on Tuesday morning, others on Tuesday afternoon, others on Wednesday, and a tail of subscribers receiving it over the following 2-3 days. The campaign has a "send date" but individual delivery is spread across a multi-day window.

The Research Evidence: Does Timing Actually Matter?

The research on email send-time impact on engagement is consistent in its finding that timing matters — but inconsistent in the magnitude of its effect and which "optimal" times are identified. Multiple studies across different industries show:

The modest average effect: Sendgrid, Mailchimp, and Campaign Monitor have all published analyses of their sending data showing that optimal timing improves open rates by 5-15% on average relative to worst-case timing. This is a real but relatively modest effect — the difference between sending at the worst time (e.g., 3 AM Saturday for a B2B audience) and the best time (8 AM Tuesday for the same audience) is real, but the difference between two reasonably good times (8 AM Tuesday vs 10 AM Wednesday) is often minimal.

The individual variation problem: Aggregate analyses identify industry-level "optimal" send times (e.g., "B2B email performs best on Tuesdays 10 AM-12 PM") that mask enormous individual variation. A subscriber who works night shifts has a completely different optimal send time than the aggregate audience model assumes. A subscriber in a different time zone is misserved by a UTC-normalised optimal time. Per-subscriber STO is a genuine improvement over aggregate recommendations precisely because it accounts for individual variation.

The content dependency: Send-time impact varies with content type and audience engagement level. Time-sensitive offers (flash sales, event registration deadlines) benefit more from precise timing than evergreen content (newsletters, educational series). Highly engaged subscribers who check email frequently are less sensitive to send time — they will see the email regardless. Lapsed subscribers who check email infrequently may never see the email regardless of when it's sent.

AI-Powered STO: How Klaviyo, Salesforce, and Mailchimp Do It

Klaviyo Smart Send Time: Klaviyo's STO analyses each subscriber's email open history to identify a recurring engagement window. Klaviyo's algorithm defaults to the same day of the week and time of day the subscriber has most frequently opened past emails. Minimum data requirement: approximately 3 months of send history for the subscriber. Subscribers without sufficient history are delivered at a default time.

Salesforce Marketing Cloud Einstein STO: Salesforce's Einstein Send Time Optimisation uses a machine learning model trained on engagement data across the broader Salesforce platform (not just the specific programme's data), providing predictions for subscribers with limited history. Einstein provides per-subscriber optimal day-and-time predictions with a confidence score — lower-confidence predictions default to programme-wide optimal times.

Mailchimp Send Time Optimisation: Mailchimp's STO analyses each subscriber's individual engagement patterns and delivers within a 24-hour window of the campaign's scheduled send time. Mailchimp also offers a "Campaign Optimiser" that recommends the best scheduled send time for the campaign based on the audience's aggregate historical engagement patterns — a simpler tool than per-subscriber STO.

The data quality problem across all platforms: All AI-powered STO tools rely on open event timestamps as the primary engagement signal. In 2026, open timestamps are polluted by Apple MPP pre-fetching — which fires the open tracking pixel at email delivery time, not at human engagement time. A subscriber who receives email at 9 AM (when Apple's proxy pre-fetches it) but who actually reads the email at 7 PM (when they check their inbox) has an open timestamp of 9 AM — which the STO algorithm records as a "9 AM engagement" and uses to predict future optimal send times. The algorithm is optimising for the time the proxy loads the email, not the time the human engages with it.

How Gemini AI and MPP Changed the Timing Calculus

Apple MPP and Gmail's Gemini AI have both changed the timing calculus for email send-time optimisation in ways that the ESP marketing for STO features has not yet fully incorporated:

MPP's impact on STO data quality: As documented, MPP fires the open tracking pixel at delivery time rather than human engagement time. For programmes with significant Apple Mail audiences (40-56% of consumer sends), the open timestamp data that STO algorithms use is a blend of proxy-generated delivery timestamps and genuine human engagement timestamps. STO algorithms trained on this data are partially optimising for delivery time rather than human engagement time. The algorithm "learns" that subscriber X engages at 8 AM — but X is an Apple Mail user whose email is pre-fetched at 8 AM while X actually reads email at lunch. The STO algorithm is delivering email at 8 AM to minimise delay between send and "open" — but the human engagement that actually matters commercially (clicks, purchases) occurs at noon regardless of when the email was delivered.

Gemini AI's inbox ranking: Gmail's Gemini AI ranks email within the inbox based on engagement history and content quality, not primarily on send time. An email from a high-engagement sender receives prominent placement in the Gmail inbox whether it arrives at 7 AM or 7 PM — Gemini surfaces it when the user opens Gmail regardless of when it was delivered. For Gmail recipients, send time's impact on visibility is reduced by Gemini's active inbox ranking. The email that arrives overnight is not buried at the bottom of a chronologically sorted inbox — it is ranked by Gemini's relevance score and may appear prominently when the user opens Gmail in the morning.

Click-based STO as the 2026 correction: The solution to both the MPP data pollution and the Gemini ranking change is to shift STO from open-based optimisation to click-based optimisation. Click timestamps are not inflated by MPP or Gemini processing — they require deliberate human action. STO based on per-subscriber click history would use genuinely human engagement signals to predict optimal timing. This data is available in ESP click logs but is not yet the primary input for most commercial STO implementations. ESPs that update their STO models to weight click data more heavily than open data will provide more accurate timing predictions for the 2026 email environment.

Send-Time Benchmarks by Industry and Email Type

Aggregate industry send-time benchmarks based on 2025-2026 campaign data across major ESPs (using click rate as the primary metric to avoid MPP open rate distortion):

The benchmarks above should be used as starting points for testing, not as fixed rules. Audience composition, geography, and programme-specific subscriber behaviour all produce deviations from aggregate industry benchmarks. B2B audiences in East Asia have completely different optimal send times than US-based B2B audiences. Consumer newsletter subscribers who work non-standard hours have optimal times outside the benchmark window. Use the benchmarks to frame the first A/B test; use the A/B test results to calibrate the programme-specific optimal time.

When STO Significantly Helps Performance

STO provides the most significant performance improvement in specific scenarios:

Geographically diverse audiences: If the subscriber list spans multiple time zones (international audiences, US programmes with coast-to-coast coverage), STO provides clear value by ensuring each subscriber receives the email at their local optimal time rather than being disadvantaged by a single global send time. A 9 AM EST send time is ideal for East Coast US but is 6 AM in California and 2 PM in the UK — all of which are suboptimal for significant fractions of the audience. Time-zone-aware STO eliminates this geographic send-time penalty.

Programmes with high day-of-week variation: Some audiences have strong day-of-week preferences — B2B audiences that check email primarily during the work week, consumer audiences that check email most heavily on weekend mornings, or specific subscriber cohorts with recognisable patterns. Programmes with high day-of-week engagement variation see more STO benefit than programmes with relatively consistent day-of-week engagement.

Long-tail promotional windows: For promotional campaigns with 7+ day validity (week-long sales, month-long programme promotions), STO can stagger delivery across the promotional window in a way that feels personalised rather than broadcast — each subscriber receives the promotion when they are most likely to act, rather than all receiving it simultaneously.

When STO Hurts Deliverability

STO creates specific deliverability risks that are rarely mentioned in ESP feature marketing:

Irregular volume patterns: STO-enabled campaigns deliver email in a drip pattern spread over 24-72 hours rather than a single volume spike. For ISPs that monitor per-hour volume patterns as sending behaviour signals, an unusually consistent drip pattern (sending to 500 subscribers per hour for 72 hours) may look unusual compared to the programme's historical sending pattern (which has been batch-based). Monitor SNDS and Postmaster Tools data during the first few STO campaigns to verify the drip pattern does not trigger unusual filtering.

Time-sensitive content stale by delivery: If the campaign is time-sensitive (a flash sale that expires Tuesday night, a webinar that happens Wednesday) and STO delivers to some subscribers on Thursday, those subscribers receive stale content — the sale has expired, the webinar is over. For time-sensitive campaigns, disable STO and send at a single fixed time before the deadline. STO is appropriate for evergreen or long-validity campaigns, not for tight deadline promotions.

MPP-contaminated STO creating false patterns: As documented above, STO optimising primarily on open data for Apple Mail-heavy audiences may deliver email at times that minimise the Apple-proxy open delay rather than maximising genuine human engagement. Monitor click rate (MPP-resistant) rather than open rate to evaluate whether STO is actually improving genuine engagement or just improving the proxy-generated open timestamp correlation.

Send-Time Optimisation Implementation Guide

The 2026 implementation framework for STO that accounts for MPP, Gemini AI, and genuine engagement optimisation:

(1) Baseline measurement: Before enabling STO, measure click-to-delivered rate (CTD) at the programme's current fixed send time. This is the baseline against which STO impact will be measured. Do not use open rate as the baseline metric for STO evaluation — MPP contamination makes open rate unreliable for this comparison.

(2) A/B test STO vs fixed time: Split the active subscriber list (50/50) — one group receives the campaign with STO enabled, the other at the programme's traditional fixed send time. Run this test for 3-5 campaigns to accumulate statistically meaningful sample sizes. Compare CTD rates between the two groups. If STO produces >10% CTD improvement, the investment in STO is justified. If CTD improvement is less than 10%, the STO benefit for this programme is marginal.

(3) Click-data-weighted STO: When configuring STO in the ESP, look for options to weight click engagement more heavily than open engagement. Not all ESPs provide this configuration, but it is the direction the best STO implementations are moving. If the ESP does not support click-weighted STO, use STO with awareness that open-data contamination from MPP may reduce its accuracy for Apple Mail-heavy audiences.

(4) Disable STO for time-sensitive campaigns: Establish a campaign tagging system that marks time-sensitive campaigns (those with deadlines within 48 hours of intended send) for fixed-time delivery rather than STO-enabled delivery. All other campaigns can use STO if the A/B test has confirmed its value for the programme.

Email send-time optimisation, correctly implemented and evaluated against click data rather than open data, provides genuine performance improvement for programmes with geographically diverse audiences and high day-of-week variation. The 5-15% click rate lift that well-implemented STO provides represents real commercial value that compounds across a full year of campaigns. The key is implementing it correctly — using the right signal (clicks), testing before committing, and disabling it for time-sensitive campaigns where the drip delivery pattern introduces more risk than the timing optimisation provides benefit.