Overview
This article provides tips, best practices, and common pitfalls to help you configure and optimize Optimal Send Time (OST) predictions. It also highlights common pitfalls to avoid when using both the built-in OST options (in scenarios or campaigns) and the OST prediction template.
Tips & best practices
1. Data range selection
The data range you use depends on the method:
Built-in OST (scenarios or campaigns): Uses a fixed 90-day customer event history.
OST prediction template: You can select your own timeframe. In general, a more extended historical range captures more behavioral data and improves prediction accuracy—but you should test what works best with your own dataset and campaign frequency.
2. Event selection
Optimal Send Time (OST) optimizes for clicks by default.
The previous "optimize for open vs. click" option has been removed because open event data can be unreliable.
OST uses opens only as a fallback when there is insufficient click data.
Ensure your engagement tracking—especially click events—is properly implemented, as predictions rely primarily on this data.
3. Default time as a fallback
Fallback behavior differs depending on where you apply OST:
Built-in OST (scenarios or campaigns): The fallback is the scheduled send time of your campaign.
OST Prediction template: You define a default hour (for example, 9 = 9:00 UTC) for users without enough data to calculate a personalized time.
Avoid setting the default hour to 0 (00:00 UTC) unless you are sure it aligns with your audience's active hours.
4. Time zone awareness
OST predictions and default send times are always calculated in UTC (UTC-0). The timezone defined in user settings doesn't affect calculations.
For global audiences, consider creating regional segments or using Silent hours in scenarios to avoid sending messages during night hours.
If localization isn't possible, choose a balanced UTC hour (for example, 14:00 UTC) to minimize off-hour sends across key markets.
5. Messaging channel behavior (scenarios)
In Scenarios, OST behavior depends on your package:
OST can be calculated per messaging channel (Email / SMS / MMS / RCS / Push / All channels).
The messaging channel selector requires the Loomi AI Journey Orchestration package.
Without this package, OST uses engagement data across all channels (generic OST), available in the Loomi AI Audience Optimization package.
Common pitfalls to avoid
Using sparse or incorrect event mappings: Ensure your campaign events are properly tracked as clicked (primary) and opened (fallback).
Not running enough campaigns before using OST: The model requires sufficient historical campaign data to produce accurate predictions.
Setting a default hour that falls outside typical engagement times: Midnight UTC is rarely optimal for global audiences.
Ignoring send strategy selection: Choosing the wrong send strategy can result in messages being sent too late or too early.
Assuming OST alone solves timing issues: OST improves timing, but overall campaign success still depends on good segmentation, targeting, and content relevance.
Summary
By aligning your configuration with how OST works and applying these best practices, you can improve model accuracy and ensure campaigns reach your customers at the most effective times.
For more details, refer to the Optimal Send Time Prediction and Optimal Send Time in scenarios and campaigns documentation.