Work health and safety of food delivery workers in the gig economy: field trial of delivery window messaging

When food delivery workers in the app-based gig economy accept an order from a customer, they are given “time windows” for their trip to the restaurant to pick up the order and for their trip from the restaurant to the customer to deliver the order. In earlier phases of our research, we found that there was a widespread perception among food delivery workers that these time windows are firm expectations rather than the estimates that food delivery platforms intend them to be. This perception is further compounded by the belief among food delivery workers that their in-app ratings, access to preferred shifts, and, ultimately, their tenure with the platform are all affected by their ability to pick up and deliver within specified time windows, which is not the case. Since food delivery platforms do not explicitly communicate the difference between expectations and estimates to their fleet, these misperceptions have persisted.

As a result, food delivery workers may be incentivised to engage in behaviours that pose significant work health and safety (WHS) risks. In the context of road and traffic safety, these behaviours include using mobile phones while riding or driving, pushing themselves to continue working while fatigued, cycling on footpaths and in other pedestrian-only areas to cut down on travel time, and speeding or rushing. These behaviours are extremely widespread - only 5% of the food delivery workers we surveyed in an earlier phase of this project reported “never” rushing to deliver orders - and are driven largely by the desire to achieve and maintain high in-app ratings, satisfy customers, and maximise income (Convery et al., 2020).

The overall aim of this trial is therefore to determine whether reframing in-app messages that communicate pickup and delivery time windows as estimates rather than expectations reduces risk exacerbation and/or increases risk mitigation behaviours on the road. Specifically, we seek to answer the following research questions: (1) Can a behaviourally framed message reduce average maximum travel speed? (2) Can this message change food delivery worker beliefs about the negative consequences of exceeding the delivery estimate? (3) Do any effects on travel speed or beliefs negatively impact business as usual practices (e.g. delivery times, earnings, and customer satisfaction)?