Nine months into launching GOGO Delivery (a crowd-sourced delivery service), the demand grew to a point where the supply couldn’t keep up. Couriers were scrolling through a long list of orders to find orders to take. Response time increased and completion rate dropped. To improve this, our team pivoted our objective to improve courier activation.

My role

I led the design for courier experience. I worked closely with our PM, FE and BE enginers, Data Analyst, and cross-functional team (Operations, Customer Service, Marketing) throughout the design process.

Research

Through user interviews and focus groups, we found that couriers had difficulty in finding relevant orders. A few months earlier, when there was less demand, there were only around twenty pending orders that couriers can pick from at a given time. Now, there’s over a hundred.

We learned that couriers select orders based on distance from pick up location, drop off location and earnings. To empathize with our couriers, our team went out and delivered packages. This enabled further learnings such as order bundling planning to maximize travel efficiency and earnings, and package size and weight considerations.

Metrics

0.4 daily orders per registered courier

42 minutes average response time

Design sprint

I led a design sprint with cross-functional teams to identify the key problems to tackle for courier activation and decide which ideas to test.

Sketches from the design sprint to tackle couriers' problem of finding relevant orders
Sketches from the design sprint to tackle couriers' problem of finding relevant orders
Early prototype to explore suggested bundling for couriers

We explored the most popular idea: a visual map to show couriers available orders based on their location and suggesting bundling to increase efficiency and income. However, due to the technical scope of this idea, we chose not to pursue this.

Order suggestions

Push notifications for new orders based on proximity to active orders
Push notifications for new orders based on proximity to active order

Couriers tend to pick orders that are “on the way.” We tested the idea of sending push notifications for new orders to active couriers based on their orders’ pick up and drop off locations. This enables them to find orders they’re interested in without opening the app.

Order filtering

Courier flow for filtering and sorting available orders
Courier flow for filtering and sorting available orders

We also explored using filters and sorting to help couriers find relevant orders. We tested the prototype with couriers and made several iterations to suit their needs and technical constraints. The final design uses a sticky top bar with dropdown menus to indicate active filters (in a different color) and the sorting method at all times.

Prototype for pick up region, drop off region filters and sorting by time, distance to pick up and earnings

We shipped the order filtering and sorting first. We assumed this had the greatest outcome for couriers due to their strong preference of picking orders by region and earnings.

Additional filters

Additional filters for time, package size, and weight
Additional filters for time, package size, and weight

In addition, we shipped filters for time constraints, package size, and weight. This enabled couriers to further filter orders that are more relevant to them - for example, package weights that they are physically capable of carrying when bundling orders.

Filters usage

After one month of release, the package size and weight filters had the most usage at 95%. Whereas the time filter had 25%, and pick up region filter only at 10%. Overall response time increased. Next steps: research and iterate on improving the time and pick up region filters.

Filter usage one month since launch
Filter usage one month since launch
Adoption vs orders taken 30 days before and after launch
Adoption vs orders taken 30 days before and after launch

Filter adopters increased their orders by 1.8x post-launch

Outcome

1.8x

Daily orders by adopters (0.4 → 0.7)

-83%

Average response time (42m → 8m)

Retro

  • Assumed pick up regions filter and sorting is more impactful than size, weight, time
    • Testing prototype indoor vs actual usage while making deliveries?
  • Why low usage on pick up region filters?
    • How often do couriers pick up orders from different regions?
    • Are the defined regions too large?
    • Interaction cost?
  • When increased completion rate (price adjustment, coronavirus)
    • Less than 5 pending orders to pick from
    • How impactful are the filters?