Route Optimization for Food Delivery Fleets

15-minute hot-bag SLAs, ghost-kitchen consolidation, and evening peak — solved.

TL;DR

Food delivery is the most time-pressured branch of last-mile logistics. The hot-bag SLA — typically 15 minutes from kitchen to customer — is the dominant constraint, and ghost-kitchen consolidation (one driver picking up from 3–4 brands at the same address) is the dominant operational pattern. SortDrops handles both: time-window-aware routing keeps each delivery inside its SLA, and the dispatcher app collapses same-address pickups into a single virtual stop.

Key data point: The food-delivery industry's de-facto hot-bag SLA is 15 minutes from kitchen-out to customer-doorstep. Across major aggregators (Talabad, Careem, Deliveroo, Noon Food, Snoonu, Rafeeq) this benchmark is published in their merchant SLAs and breach rates above 5% trigger automatic merchant penalties.

Why food-delivery routing is different

The 15-minute hot-bag SLA collapses the planning window. A grocery driver can plan a 3-hour route in advance; a food driver is making decisions every 2–3 minutes about whether to take the next pickup or return to base. Real-time route adjustment matters more than perfect upfront optimization. SortDrops' dispatch mode allows mid-shift re-optimization in under 1 second per stop, so a dispatcher can drag a new pickup into a driver's queue without rebuilding the whole day.

Ghost kitchens consolidate 3–8 virtual brands into one physical kitchen. That means one driver picking up from "Brand A, Brand B, Brand C" all sitting on the same shelf inside the same building. SortDrops collapses same-address pickups into a single dispatch stop with a multi-order pickup checklist, so the driver doesn't make three separate "is my order ready" interactions with the kitchen.

Evening peak (Thursday-Saturday 19:00–22:00 in GCC markets) doubles or triples normal volume. The right operational pattern is to add 30–50% temporary fleet capacity and let the dispatcher re-balance routes every 15 minutes. SortDrops Team supports adding drivers mid-shift without disrupting in-flight routes.

A typical food-delivery dispatch shift

A typical 4-hour evening shift in a Gulf city: dispatcher logs in at 18:00, imports the night-shift driver roster (12 riders), and starts assigning incoming orders as they arrive from the aggregator API or POS. SortDrops shows each driver as a queue card on the dispatcher screen — current order, ETA to current customer, distance from next pickup. As an order comes in, the dispatcher drags it onto the best-fit driver, the optimizer recomputes that driver's queue in real time, and the SLA clock starts.

The hard problem is the 'queue of three' tradeoff: should driver A pick up order 2 before delivering order 1, or after? If pickup 2 is on the way to drop 1, yes. If it's a 4-km detour, no. SortDrops evaluates this on every queue change and surfaces the driver-side ETA so the dispatcher sees the cost in milliseconds, not minutes.

Multi-drop within 30 minutes is the soft cap for hot-bag work — beyond that, food quality degrades unacceptably. SortDrops surfaces a 'queue too long' warning when a driver's projected last-drop ETA exceeds 30 minutes from the first pickup.

Ghost kitchens, restaurant pickup queues, and branding

Ghost-kitchen operators run 3–10 brands from one address. Each brand has its own ticket, sometimes its own packaging shelf. SortDrops collapses the addresses but keeps the per-brand order numbers visible on the driver-link, so the driver can confirm pickup of all three at the kitchen counter without confusion.

Restaurant pickup queues are the hidden cost. A driver who arrives 90 seconds before the food is ready waits idle. A driver who arrives 90 seconds late breaches SLA. SortDrops doesn't predict kitchen prep time today, but our per-stop service-time field lets you encode an average prep time per restaurant, and the optimizer respects it when ordering pickups.

SortDrops features for food-delivery

Operational thresholds we recommend

Frequently asked questions

Does SortDrops integrate with Talabad / Careem / Deliveroo aggregator APIs?

Not directly today. Most food-delivery fleets we work with run a hybrid model: aggregator orders flow into the merchant POS, and from there into SortDrops via Excel/CSV import or a manual paste. Direct aggregator API integration is on our 2026 roadmap.

How does SortDrops handle the 15-minute hot-bag SLA?

The optimizer minimizes total kilometres, which directly minimizes per-stop time-on-route. We also surface a per-stop projected delivery time so the dispatcher can see SLA risk before assigning, and a 'queue too long' warning if a driver's projected last-drop exceeds the 30-minute multi-drop cap.

Can multiple drivers share a ghost-kitchen pickup point?

Yes. SortDrops treats the pickup point as a regular stop — multiple drivers can be routed to the same address at staggered times, each with their own list of orders to collect.

What happens if a customer cancels mid-route?

The dispatcher removes the stop, SortDrops re-optimizes the affected driver's queue in real time, and the driver-link refreshes within 5 seconds. The driver doesn't have to do anything except keep driving — the next stop is already updated.

Optimize your food-delivery fleet free →