All work

Parcelry

A concept re-architecture for a fast-growing subscription box business. Renewal day stops being an emergency when every order has a queue, every lane has a known capacity, and the floor can see both.

ConceptFictional project. No real client, and nothing here is a measured result.

Sector
Subscription commerce
Service
Business scaling
Year
2026
Concept design: Parcelry

In this concept

  • Capacity planning
  • Queue architecture
  • Monitoring dashboard
  • Load testing

Parcelry is a concept project. There is no real client behind it. We made it to show how we would approach a business that has outgrown the system it started on. Everything below describes design goals, not measured results, and the numbers in the screens are sample data.

The problem

A subscription box business has a strange shape. For most of the month, traffic is calm. Then renewal day arrives and the whole customer base is billed, picked, packed, and labelled at once.

The system that worked at launch was never built for that. Billing, stock, and label printing all run in one chain, so one slow step holds up the rest. The payment provider rate limits the job. A retry charges a customer twice. The warehouse stands idle in the morning and works late at night, and nobody can say how far behind they are until the courier has already left.

Growth makes every one of these worse. More subscribers should be good news, not a bigger emergency.

The approach

We would start by measuring, not rewriting. Trace one renewal from the charge to the courier scan, find what each step can actually hold, and load test it with a copy of a real renewal day.

Three decisions would shape the rest of the work:

  • Every step gets a queue. Renewals stop being one long job and become messages moving through lanes: billing, pick, pack, label, dispatch. A slow lane builds a backlog in its own queue and leaves the others running.
  • Every lane has a known capacity. Each one drains at a rate the people and systems behind it can hold. The spike goes into the queue, not onto the floor.
  • Every retry is safe. Each box carries one key from charge to dispatch, so running a step twice changes nothing. Failed work waits somewhere visible instead of disappearing.

We would move one lane at a time alongside the old system, so there is no big switch-over on a renewal day.

What we would build

  • A queue layer between billing, stock, picking, packing, labels, and dispatch, with retries and a holding area for failed work.
  • Idempotent billing and label jobs, so a repeat can never charge twice or print a second label.
  • Capacity rules for each lane, with workers that scale up inside limits the team sets.
  • A staggered renewal schedule, so the charge run can spread out without changing what a subscriber sees.
  • A monitoring dashboard on the warehouse wall, showing what is queued, what is draining, and whether the courier cut-off is safe.
  • A repeatable load test, so the next growth target gets checked before renewal day and not during it.

Goals, not results

The goals we would design against: renewal day feels like a normal shift, no customer is charged twice, and anyone on the floor can tell at a glance whether today’s boxes will make the courier.

Key screens

The floor

The whole day as a field of boxes, one row per lane and one column per hour. The renewal wave rolls through the lanes. Anything above a lane’s capacity turns orange, and the hours still to come appear as outlines.

The wave

The current renewal as a shipping label: how many boxes, how many have gone, when the queue should clear, and how that sits against the courier cut-off.

Lanes

Each step with its depth, its workers, and a plain status. Holding, draining, or clear. Nobody has to read a graph to know where to go next.