Alok Sharma
All work

Case study · 2025 · Parcelo

Rebuilding returns for a carrier that never wanted them

A returns experience for 1.4m parcels a year, designed around the drivers and depot staff who actually absorb the work.

Hero image — placeholder

Overview

Parcelo moved parcels one way. Returns were handled by a spreadsheet in the Bergen depot and a phone number that rang in customer support. As volumes grew, the workaround became the second-largest driver of support cost.

I led design on a returns product covering the consumer flow, the depot handling app and the merchant-facing reconciliation view.

Problem

A return looked like one transaction to the merchant and four separate jobs to Parcelo: authorising it, collecting it, sorting it and settling it. No system held all four, so each handoff was reconstructed by a person.

The brief we were given was “build a returns portal”. The problem was that nobody agreed on what state a returned parcel was in at any given moment.

Figure — placeholder
Fig. 1 — Service blueprint of the existing return, four systems and eleven manual handoffs

Research

Two weeks in the Bergen and Oslo depots, eleven shifts observed, plus diary studies with fourteen consumers who had returned an item in the previous month.

The finding that reframed the project: depot staff had invented a physical shorthand — a folded label corner — to mark parcels the system could not classify. Roughly one in nine returns carried one.

Process and explorations

We tested three models of the same idea: return as a new shipment, return as a reversal of the original, and return as a claim against a merchant account. Each implied a different data model and a different owner.

The reversal model tested best with consumers and worst with finance. We shipped the shipment model with a reconciliation layer, which cost us two months and saved a rewrite.

Figure — placeholder
Fig. 2 — Three data models sketched against the same six scenarios

Final design

Consumers get one screen and one decision: where the parcel leaves from. Everything else is inferred from the original shipment.

Depot staff get a scan-first interface with the exception cases promoted to the top level, because exceptions are the job. The folded-label shorthand became a real status.

Figure — placeholder
Fig. 3 — Consumer flow, depot handling screens and the merchant reconciliation view

Prototype

We built the depot app as a coded prototype on real scanner hardware and ran it on live parcels for three shifts before writing production code. Two of the four screens did not survive.

Figure — placeholder
Prototype walkthrough — 3 min 40 s

Impact

Measured against the six months before launch, across all Norwegian depots.

−41%
Returns-related support contacts
2.1 days
Median return settlement, from 9.4
94%
Depot parcels classified without a human decision

Reflection

I would spend the first month on the finance conversation rather than the sixth. The data model was a commercial argument dressed up as a technical one, and design was the only function looking at all sides of it.

The thing I keep: the folded label. Someone had already designed a solution. The work was noticing it.

Next case study Scheduling for four thousand clinicians who all have a good reason