We've introduced a smarter auto-dispatch system that better balances delivery speed and driver productivity. Instead of dispatching tours at a fixed time before the scheduled pickup, the system can now deliberately delay dispatching by a configurable amount after an order comes in. This gives the stacking engine a fair window to find suitable stacking candidates for every order, regardless of when it was placed.

Additionally, the stacking engine now handles scheduled time differences more intelligently, distinguishing between hard-assigned and soft-assigned tours. For hard-assigned tours (where the merchant only starts preparing as soon as a driver is en route to the pickup), the time restriction between stacked orders can be relaxed. For soft-assigned tours (where the merchant is already preparing), a tighter window ensures drivers aren't waiting for orders that are ready at very different times.

Previously, the dispatch system faced a fundamental trade-off: dispatching early meant faster deliveries but fewer stacked orders, while dispatching later improved stacking rates but slowed down delivery times. The new system resolves this tension by making dispatch timing configurable per tenant, service area, and place, allowing each operation to find the right balance for their specific order patterns and volumes.

We'll continue making the dispatch logic smarter by incorporating place-specific and time-of-day-specific configurations. High-volume pickup locations (like popular fast food chains) can benefit from longer delay windows since the chance of finding a stack is much higher, while low-volume locations can dispatch immediately to avoid unnecessary delays. We're also looking at leveraging historical order data to automatically tune these parameters based on real-world patterns.

Webhook subscriptions can now be configured to only receive specific event types directly from the Dashboard. This feature was previously only available via the API - it's now accessible through the Dashboard, making it easier for operations teams to manage their integrations without developer involvement.

Location update notifications for bookings are now processed more efficiently, with smart skipping logic that reduces unnecessary processing without affecting data quality. This results in lower system load during peak hours.

When the routing service experiences temporary disruptions, missing route estimates are now automatically backfilled once the service recovers. This eliminates the need for manual intervention and ensures that all bookings have accurate distance and duration data.

The Booking Details page in the Dashboard received several UI improvements: all stop status badges are now displayed, time formatting is consistent across stop details, and title/subtitle formatting has been aligned for a cleaner look.

As part of our ongoing effort to align API naming with our domain model, we've renamed the filter for finding users on active tours. The previous name will continue to work until August 24, 2026. If your integration uses this filter, please refer to the migration guide for details.