Send time, day, and channel optimization

MessageGears provides three ML models that each answer a distinct optimization question for your recipients:

Model	Question it answers	Output field	Value
Send Time Optimization (STO)	What hour of the day should I send?	ML-OptHour	0–23
Optimal Day	What day of the week should I send?	ML-OptDay	0–6
Best Channel	Which channel should I use?	ML-OptChan	email, push, sms

Each model trains independently, scores independently, and produces its own output. You can activate any combination of them based on your use case.

Shared inputs and training

All three models train on the same historical engagement data: messages sent, opens, clicks, and push interactions. Recipients are re-scored at a cadence that accounts for new engagement events and demand, so predictions stay current as customer behavior shifts.

The training process accounts for industry-specific signals such as Apple's Mail Privacy Protection, filtering out automated behaviors and keying on actual customer interactions where possible.

Requirements

• Email and channel engagement data, such as clicks, opens, sends, and push interactions
  • Data Input Schemas for Transaction/conversion data

---

Send Time Optimization (STO)

Predicts the best hour of day to send a message to each recipient.

How it works

The model analyzes when each recipient has historically engaged (opened, clicked) with messages and produces a single optimal hour (0–23) per recipient. That hour is then used to populate channel-specific send time timestamps (EmailSendTime, PushSendTime, SmsSendTime) that control when the message is actually released for delivery.

tip

For details on how send time timestamps control message delivery, see Send Time Designation.

Output fields

Field	Type	Range	Description
ML-OptHour	INT	0–23	Optimal hour of day to send.
EmailSendTime	TIMESTAMP	e.g., 2019-06-20 16:00:00-04:00	Scheduled release time for email, derived from ML-OptHour.
PushSendTime	TIMESTAMP	e.g., 2019-06-20 16:00:00-04:00	Scheduled release time for push, derived from ML-OptHour.
SmsSendTime	TIMESTAMP	e.g., 2019-06-20 16:00:00-04:00	Scheduled release time for SMS, derived from ML-OptHour.
ML-OptHourGroup	CHAR(1)	C, R, O	C – Control group, R – Random/Training, O – Optimized

---

Optimal Day

Predicts the best day of week to send a message to each recipient.

How it works

The model identifies which day of the week (Sunday through Saturday) each recipient is most likely to engage. This can be used to build day-specific audience segments or schedule campaigns on per-recipient optimal days.

Output fields

Field	Type	Range	Description
MLOptDay	INT	0–6	Optimal day of week to send (0 = Sunday, 6 = Saturday).
MLOptDayGroup	CHAR(1)	C, R, O	C – Control group, R – Random/Training, O – Optimized

---

Best Channel

Predicts the best channel to reach each recipient. This is also referred to as "next best channel" (NBC).

How it works

The model compares each recipient's engagement across email, push, and SMS to determine which channel they are most likely to interact with. Use this to route recipients to their preferred channel or to prioritize channel selection in cross-channel campaigns.

Output fields

Field	Type	Range	Description
ML-OptChan	VARCHAR	email, push, sms	The channel the recipient is most likely to engage with.

---

Full schema reference

All fields are keyed on RecipientId and can be joined back to your customer tables.

Field	Type	Range	Model	Description
RecipientId	VARCHAR		All	The unique customer identifier.
MLOptHour	INT	0–23	STO	Optimal hour of day to send.
EmailSendTime	TIMESTAMP		STO	Scheduled release time for email.
PushSendTime	TIMESTAMP		STO	Scheduled release time for push.
SmsSendTime	TIMESTAMP		STO	Scheduled release time for SMS.
MLOptHourGroup	CHAR(1)	C, R, O	STO	Experiment group assignment.
MLOptDay	INT	0–6	Optimal Day	Optimal day of week to send.
MLOptDayGroup	CHAR(1)	C, R, O	Optimal Day	Experiment group assignment.
MLOptChan	VARCHAR	email, push, sms	Best Channel	Optimal channel to send.