Meituan discloses technical and design detail on delivery worker routing algorit...

Meituan discloses technical and design detail on delivery worker routing algorithm

Meituan discloses technical and design detail on delivery worker routing algorithm

Chen Du

posted on November 8, 2021 2:23 pm

As China's first personal data protection law went into effect recently, marking the height of the public's interest in learning about the inner workings of algorithms that dictates our lives, Meituan made some of the details public to address public concern.

As the public’s interest in learning more about delivery routing and algorithms of food delivery and ride hailing services grew, Meituan for the first time disclosed much details on its routing optimizations last week. 

The largest local services internet company of China posted via its WeChat official account some of the rules regarding how it routes riders, and attempted to address public concern regarding riders being assigned with new orders en route, which sometimes causes late delivery. 

Meituan claims that its system attempts to strike a balance among three parties: the customers, the riders, and the restaurants. More specifically, the algorithm works by trying its best to make sure that the restaurants’ prepared orders are picked up promptly, customers receive their food in time, while the riders get rewarded under reasonable labor intensity.

For example, new orders will more likely be assigned to riders who have longer delivery time windows in their existing jobs, as well as those who don’t necessarily need to take a major detour that may result in late delivery. For freshmen riders, the order assignment system is designed to be inclined to favor them with orders that are of shorter distances and lower “difficulty”.

What may come as a surprise to many is the fact that the system does not simply assign new orders to riders on existing jobs based on current distance, but rather added distance, according to a graph demonstration. This means that although rider A’s current route length is longer than rider B’s, A is still going to get assigned with the new order because his/her added distance is shorter.

As for the public’s grievance over late delivery caused by newly assigned orders messing with riders' routing, Meituan said that there can sometimes be anomalies that causes its systems to not perform like it’s optimized for, and that the company is now piloting add-on countermeasures to address them.

One of these countermeasures is an internet-connected device called “Chucanbao”(出餐宝), which is connected to the merchants’ order processing terminals. Restaurant operators can use the device to call in riders only when orders are prepared and ready. Meituan said that about 2,400 restaurants across the country tested the device, and that the data from an early batch of 32 restaurants in Nanchang, Jiangxi Province showed that rider’s average onsite wait time was cut down by half.

Another countermeasure is a direct manual re-assignment option on riders' app, which can be triggered by Meituan's backend assessment algorithm that constantly looks for signs of late delivery risks. 

One scenario is when the system detects a rider has been inactive en route for a while (maybe due to moped having mechanical issues), then the option may be triggered, asking if the rider would like to re-assign queued orders. Preliminary results show that late delivery of those riders utilizing the feature was reduced by about 52%.

Other measures are also being adopted or developed at the moment, said the company.

Meituan, as well as some of its peers in the local services sector that employs armies of riders, have long been accused of exploiting delivery workers as posts venting grievance are commonly found on social networks and delivery platforms. This is the first time that the company, or any other in delivery-related service, has disclosed technical and design details of routing and assignment systems, as companies often treat such systems as corporate secrets. 

China’s first personal data protection law went into effect more recently, marking the height of the general public’s interest in the inner workings of algorithms that dictates many aspects of our lives today. Although what Meituan disclosed this time was far from a complete picture of those algorithms, it marks the potential beginning of a more transparent future for itself, as well as the algorithmically-driven industry.