← All work
Parcelry
Failed deliveries
Half of all failed drops came from three suburbs. Changing their delivery windows cut the failures by a third.
$1.2M
saved a year
Role
Analyst, with logistics and product
Tools
dbt, SQL, Tableau
When
2022
The problem
One delivery in fourteen failed, and every failure meant a second trip and often a refund.
The team assumed drivers were the problem, so the fixes so far had been about driver training.
The result
Failures in the test suburbs fell by about a third, and the later windows became the default there.
The saving in redeliveries and refunds is the figure on this page, and the suburb report still runs every week.
How I got there
I mapped a year of failed drops by suburb, time window and building type in dbt, and found three inner suburbs behind half of all failures.
Most were apartment blocks with no safe drop point before 6 pm. With logistics we tested later delivery windows in those three suburbs for six weeks against matched control areas.