The real difference between driver-claimed orders and dispatching in logistics is how orders are allocated, not whether one model necessarily offers lower prices. This distinction is especially important at night, when order density is low and fewer drivers are available. Comparing headline acceptance rates alone can easily conflate pricing, order visibility, and the ability to complete deliveries.
Distinguishing Order Claiming, Dispatching, and Marketplace Models
Under a driver-claimed order model, an order is placed in an open order pool or shown to multiple drivers at the same time. Drivers decide whether to accept it, and the first driver to claim it successfully usually receives the job. Order claiming is not the same as competitive bidding. If the platform has already calculated the price, drivers are competing for the right to fulfill the order—not underbidding one another.
With dispatching, the platform assigns orders to candidate drivers sequentially, in parallel, or automatically, based on factors such as distance, vehicle type, capacity, route, historical reliability, and real-time supply and demand. Drivers may still be asked to confirm a dispatched order, or the system may be designed to accept it automatically on their behalf.
These allocation methods are separate from the concepts of one-sided and two-sided markets. The OECD describes online platforms as digital services that facilitate interactions between two or more distinct but interdependent groups of users. However, an operator with its own fleet can still allow employees to claim orders, while a matching platform can use centralized dispatching. The allocation method alone does not determine the business model.
Pricing: Order Claiming Is Not Inherently Cheaper
Both models can use fixed quotes, mileage-based pricing, nighttime surcharges, or dynamic pricing. The key questions are whether compensation attracts enough driver supply and whether the unit economics remain viable after any surcharge.
An open order pool makes driver preferences immediately visible. Orders may remain unclaimed when compensation is too low, the pickup location is remote, the return trip is unattractive, or nighttime conditions involve greater risk. Dispatching can improve acceptance probability through candidate ranking, individualized incentives, or supply-and-demand pricing, but operators must also address allocation fairness and algorithmic transparency.
Lalamove Taiwan, for example, states that drivers may select orders freely and review earnings and delivery details before accepting. Its terms also specify a 1.2-times rate from 10:00 p.m. to 6:59 a.m. This shows that nighttime surcharges can coexist with driver-claimed orders; higher nighttime prices are not a defining feature of dispatching. The pricing and terms were accessed on July 27, 2026, and should be checked again before publication.
Cancellation Rates Must Be Separated by Stage
A single overall cancellation rate is rarely enough to determine whether an allocation model performs well. At minimum, operators should distinguish among the following:
- Pre-match customer cancellation rate: Orders canceled by customers before matching ÷ valid order submissions.
- Unclaimed order rate: Orders not accepted within the time limit ÷ orders entering the open pool.
- Dispatch rejection rate: Rejected or timed-out invitations ÷ valid dispatch invitations.
- Post-acceptance driver cancellation rate: Orders canceled by drivers after acceptance ÷ accepted orders.
- Post-match customer cancellation rate: Orders canceled by customers after matching ÷ matched orders.
- Rematching rate: Orders requiring another driver search ÷ matched orders.
Order claiming allows drivers to apply local knowledge about traffic, return-trip opportunities, and loading conditions. An order that is claimed may therefore align more closely with the driver’s preferences. The trade-off is that unattractive orders may receive no acceptance at all.
Dispatching gives the platform greater control over network-wide allocation, but every rejected or timed-out assignment may require another dispatch attempt. Ride-hailing research also indicates that longer pickup distances increase the risk of customer cancellation while the driver is on the way. Because ride-hailing and parcel logistics differ in vehicle requirements, loading work, and return-trip conditions, this finding should be treated only as directional evidence.
Use Completed Deliveries as the Common Conversion Metric
A claimed order may be displayed to many drivers simultaneously, while a dispatched order may be offered sequentially to only a few candidates. For that reason, the click-through rate of an open order pool cannot be compared directly with the acceptance rate of an individual dispatch invitation.
A more reliable common funnel is:
Quote completed → customer submits order → matched within the time limit → driver arrives → pickup completed → delivery completed
The primary metric should be:
End-to-end fulfillment conversion rate = completed deliveries ÷ valid order submissions
This should be evaluated alongside the five-minute matching rate, P50 and P95 matching times, average number of dispatch attempts, rematching rate, on-time rate, and stage-specific cancellation rates. These measures reveal whether fast acceptance actually results in completed deliveries.
No Taiwanese logistics platform currently publishes data that directly compares the two allocation models during the same period and at the same price. It is therefore not possible to conclude that either model is inherently superior in cancellation or conversion performance.
AI Dispatching Must Be Tied Back to Revenue and Cost
Faster AI matching does not automatically produce better operating results. Management can break the economics down as follows:
- Completed transaction value = valid orders × fulfillment completion rate × average order value
- Contribution margin per order = platform operating revenue − driver incentives − payment costs − variable costs such as customer service, cancellation compensation, and claims
Operators with their own fleets should pay particular attention to load utilization, empty-mile costs, driver working hours, and gross profit per order. Two-sided platforms should examine completed transaction value, commission rate, incentives, completion rate, and contribution margin per order together.
AI can generate measurable results. A study of Meituan’s intelligent dispatching system reported a 20.96% reduction in average delivery time and a 23.77% reduction in travel distance per order. A Lyft study of reinforcement-learning-based matching estimated more than US$30 million in incremental annual revenue. However, food-delivery and ride-hailing results should not be treated as direct forecasts for parcel logistics.
A Hybrid Model Is Worth Testing for Nighttime Logistics
Low-density nighttime markets do not require an immediate choice between a purely open order pool and fully centralized dispatching. One option is to test a hybrid model in which an algorithm selects a small candidate pool and notifies several drivers in parallel, with manual dispatchers or a guaranteed-capacity fleet taking over after a timeout. This preserves some driver choice while reducing the delays caused by sequential dispatching.
In the first testing phase, customer prices and driver compensation should remain fixed while comparing open order claiming, single-driver dispatching, parallel dispatching, and a hybrid model. A second phase can test nighttime surcharges, guaranteed earnings, or cancellation fees. Because drivers move across service zones and orders may compete with one another, switchback experiments should be conducted by area and time period.
The final decision should not be based on acceptance rate alone. Driver-claimed orders make greater use of drivers’ independent judgment, while dispatching emphasizes system-wide optimization. For nighttime logistics, the best configuration is the one that produces the strongest balance among completion rate, cancellation accountability, customer wait time, drivers’ effective hourly earnings, fairness, and contribution margin per order.
Sources
- SMEs in the Online Platform Economy (OECD, accessed July 27, 2026)
- Courier Platform Terms of Service and Guidelines (Lalamove Taiwan, accessed July 27, 2026)
- How Driver Partners Accept Orders (Lalamove Taiwan, accessed July 27, 2026)
- Satisficing Approach to On-Demand Ride Matching (INFORMS Journal on Computing, accessed July 27, 2026)
- Meituan’s Real-Time Intelligent Dispatching Algorithms (INFORMS Journal on Applied Analytics, accessed July 27, 2026)
- A Better Match for Drivers and Riders: Reinforcement Learning at Lyft (INFORMS Journal on Applied Analytics, accessed July 27, 2026)