Case study · Retail & eCommerce
Forty thousand people, six hundred pairs, no chaos
A streetwear brand's drops were being won by bots and lost by customers, with the site falling over in between. We built a queue that is genuinely fair and a checkout that holds.
Engagement · Build & live operation
Commerce under extreme, deliberate, short-duration load.
- Industry
- Streetwear & footwear
- Solution
- High-demand product drops
- Platform
- Shopify Plus
- Engagement model
- Dedicated product team
- Scope
- Queue, bot mitigation, inventory, checkout
- Peak
- 40,000 checkouts in 10 minutes
- Constraint
- Fairness must be demonstrable
- Target
- No oversells, no outage
Outcomes
What a drop is judged on
By the customers, on fairness. By the business, on whether the site stayed up.
-
40k
Checkouts in the first ten minutes
Queueing absorbs the arrival spike and releases customers into checkout at a rate the platform sustains, so demand is metered rather than dropped.
Measured at the largest drop after launch
-
0
Oversells across every drop
Inventory is reserved when a customer enters checkout rather than when they pay, so two people can never be sold the last pair.
By design: reservation at checkout entry
-
−91%
Automated purchase attempts reaching checkout
Bot mitigation operates at the queue rather than at checkout, so automated traffic is filtered before it consumes any inventory or capacity.
Measured across identified automated traffic per drop
Context
Drops that were being lost by the people who wanted them
A drop that resells instantly is not a successful drop. It is a failed one with good revenue.
The business
A streetwear brand releasing limited-run products on announced dates to a large and highly engaged audience.
The starting point
Products released at a set time with no queue. The site degraded under the arrival spike, and resale listings appeared within minutes.
The trigger
The community was openly frustrated, and the brand's own data showed a large share of purchases going to a small number of automated buyers.
What they wanted
A queue that is fair and demonstrably so, bot mitigation that works, inventory integrity, and a site that does not fall over.
Constraints
Load arrives within seconds, not minutes · fairness must be explicable to a sceptical community · false positives on bot detection block real customers · checkout must stay on Shopify.
System
What it runs at today
A drop as it runs today.
-
40k
Checkouts
In ten minutes
-
0
Oversells
Across every drop
-
−91%
Automated attempts
Filtered at the queue
-
Published
Queue rules
Before each drop
The engineering problem
Four problems with high-demand releases
Three are technical. The fourth decides whether the community stays.
-
An arrival spike no storefront survives
Tens of thousands of simultaneous arrivals will degrade any site, and a degraded site produces failed checkouts and duplicate orders.
What we did
A queue in front of the storefront that admits customers at a sustainable rate, with a position and a wait estimate rather than a broken page.
-
Bots buying faster than people can click
Automated buyers complete checkout in a fraction of the time a person takes, so without mitigation the drop is decided before customers reach it.
What we did
Mitigation at the queue rather than at checkout, combining behavioural signals with challenge escalation, so automation is filtered before it consumes inventory.
-
Overselling the last units
Reserving inventory at payment rather than at checkout entry means several customers can be in checkout for the same last pair.
What we did
Reservation on entry to checkout with a hard time limit, so stock is held for the person actually buying it and released promptly if they do not.
-
A fairness claim nobody believes
A community that suspects the queue is rigged will treat every unsuccessful drop as evidence, and the brand loses more than the sale.
What we did
Queue rules published before each drop, with entry order determined at a fixed point and no mechanism to advance a position — a claim the brand can actually stand behind.
Architecture
How it fits together
Simplified — the shape of the system rather than every service in it.
-
Queue
- Arrival absorption
- Fixed entry order
- Position & wait
The spike absorbed in front of the storefront, with order set at a fixed point.
-
Mitigation
- Behavioural signals
- Challenge escalation
- Rate limits
Filtering at the queue, before inventory or capacity is consumed.
-
Inventory
- Reservation on entry
- Hard timers
- Release
Stock held for the person in checkout, released quickly if unused.
-
Checkout
- Shopify checkout
- Metered admission
Standard checkout, protected by admitting customers at a sustainable rate.
Fairness is a product decision before it is a technical one. Fixing entry order at a published moment, with no way to buy a better position, is what makes the claim credible.
Solutions
What we built
A drop system designed to be trusted as well as to hold.
-
Queue
Arrival absorption with position and wait.
-
Bot mitigation
Behavioural filtering with challenge escalation.
-
Reservation
Stock held on checkout entry with hard timers.
-
Metered admission
Release rate matched to checkout capacity.
-
Live operations
Real-time queue, stock and checkout visibility.
-
Drop reporting
Post-drop figures published to the community.
Key capabilities
What it does day to day
Six capabilities during a drop.
| Capability | Runs | Refresh | What it does |
|---|---|---|---|
| Queue | Automatic | At the drop | Arrival absorbed, position and wait shown |
| Entry order | Automatic | Fixed point | Determined once, with no way to advance |
| Bot mitigation | Automatic | Continuous | Behavioural filtering with challenge escalation |
| Reservation | Automatic | On checkout entry | Stock held with a hard timer |
| Metered admission | Automatic | Continuous | Customers released at a sustainable checkout rate |
| Live operations | Team | During the drop | Real-time visibility of queue, stock and checkout health |
Integrations
How the moving parts plug in
Absorb, filter, meter, sell.
Arrival
- Queue enteredPosition shown
- Order fixedAt a published moment
- Wait estimated
Filtering
- Behavioural signals
- Challenge escalationWhere suspicious
- Rate limitsPer identity
Checkout
- Stock reservedOn entry
- Hard timerThen released
- Shopify checkoutUnchanged
The hard reservation timer is unpopular with customers who lose it and essential to everyone else — without it, abandoned checkouts hold the last units for hours.
Security & data
What makes the drop defensible
The community scrutinises every drop, and they are right to.
-
Published rules
Queue behaviour published before each drop, so the process is inspectable.
-
No position advancement
There is no mechanism — paid, technical or otherwise — to improve a queue position.
-
Drop reporting
Post-drop reporting on volumes, filtering and fulfilment shared with the community.
-
Checkout unchanged
Payment stays inside Shopify checkout; the queue never handles payment data.
The brief
Fairness was the requirement; scale was the easy part
Handling the load is a solved problem — a queue in front of the storefront deals with it. The harder requirement was that the brand's community, which had good reason to be sceptical, should believe the drop was fair.
That shaped the design more than the traffic did: order fixed at a published moment, no mechanism to advance, and post-drop reporting published afterwards.
- Queue absorbing the arrival spike
- Entry order fixed at a published moment
- Bot mitigation at the queue, before inventory
- Reservation on checkout entry with a hard timer
What the build had to survive
- 01Tens of thousands of arrivals within seconds
- 02Automated buyers faster than any person
- 03A community that scrutinises every drop
- 04False positives that would block real customers
Process
We load-tested against the real arrival shape
A drop is not high average traffic. It is everything in the first ninety seconds.
-
Arrival modelling
Historical drop traffic analysed to model the real arrival curve rather than a generic load profile.
-
Queue build
Built and load-tested at several times the largest previous drop before going near a live release.
-
Mitigation tuning
Tuned against recorded automated traffic with explicit attention to false positives, because blocking a real customer is the worse error.
-
Reservation model
Timers chosen from observed real checkout completion times, not guessed.
-
Live rehearsal
A full rehearsal with the operations team before the first real drop on the new system.
Technology
A queue in front of Shopify Plus
Checkout untouched; everything in front of it built for the spike.
Commerce
- Shopify Plus
- Checkout
- Inventory reservation
Queue
- Arrival absorption
- Fixed entry order
- Metered admission
Mitigation
- Behavioural signals
- Challenge escalation
- Rate limiting
Operations
- Live dashboards
- Load testing
- Post-drop reporting
Business impact
What changed for the brand
Three outcomes across a season of drops.
-
Drops complete without incident
40,000 checkouts in ten minutes, no outage, no oversell.
-
Automation largely filtered
Automated attempts reaching checkout down 91%.
-
The community calmed down
Because the fairness claim is one the brand can actually demonstrate.
The result
Drops that end with product on feet, not on resale sites
40,000 checkouts in the first ten minutes with no outage and no oversell, and automated purchase attempts reaching checkout down 91%.
The community's reaction changed more than the numbers did, because the fairness claim is now one the brand can demonstrate rather than assert.
- 40,000 checkouts in ten minutes
- Zero oversells across every drop
- Automated attempts down 91%
- Queue rules published before each drop
What we hold to on high-demand commerce
- 01Absorb the spike in front of the storefront
- 02Filter automation before it consumes inventory
- 03Reserve on checkout entry, with a hard timer
- 04Publish the rules — fairness has to be demonstrable
Free discovery call
Have an idea? Let's turn it into AI-powered software.
Book a free discovery call with our experts. Share your idea and we will help you shape the scope, timeline and budget, under NDA.
- Free consultation
- NDA before we talk
- Transparent estimate