Built for the four days a year
that decide your quarter
Retail infrastructure has a brutal test: it either holds through peak or it does not, and everyone finds out at the same time. The trick is passing that test without paying for peak capacity in February.
What retail technology leaders bring us
Checkout degrades exactly when it matters
The platform holds all year, then queues under campaign load. Usually the bottleneck is not compute at all — it is a shared component that serialises under concurrency.
Sized for peak, paid for annually
Capacity provisioned for Black Friday runs from January to December. The worst of both outcomes: expensive and still fragile.
Forecasts nobody in the business believes
Inventory and staffing decisions made on a spreadsheet model that has never been backtested, in a category where being wrong costs margin twice.
What we do for retail clients
Elasticity and cost discipline are the same project. We run them together.
Peak readiness engineering
Load testing at multiples of your real baseline, bottleneck tracing through the full request path, and pre-scaling scheduled against campaign calendars rather than reactive autoscaling alone.
- Load testing as a release gate
- Distributed tracing to find real bottlenecks
- Scheduled pre-scale ahead of campaign start
Platform modernisation
Storefront and checkout services on EKS with Karpenter and Graviton node pools, so capacity follows demand within minutes instead of a change window.
- Kubernetes platform with GitOps delivery
- Graviton migration for price-performance
- Progressive delivery and fast rollback
Demand and inventory forecasting
Forecasting on SageMaker with backtesting your merchandising and finance teams can interrogate line by line — because a forecast nobody trusts changes no decisions.
- Hierarchical forecasting by SKU and location
- Transparent backtesting and error attribution
- Feeds directly into scaling and replenishment
Unit economics
Infrastructure cost per order, per session and per channel — so platform investment can be argued in the language the rest of the business uses.
- Cost per order and per channel
- Savings Plans against the new baseline
- Anomaly alerts routed to owning teams
Consumer obligations that shape the design
Retail is less heavily regulated than finance or health, but consumer data and payment obligations still constrain the architecture.
Privacy Act & APPs
Consent, collection notices and marketing preferences enforced in the data platform rather than in each downstream tool.
PCI DSS
Scope reduction through tokenisation and segmentation so the compliance boundary stays as small as possible.
Australian Consumer Law
Availability and pricing accuracy obligations that make peak-period reliability a compliance matter, not only a commercial one.
What retail clients ask first
Four months is comfortable for a readiness program including load testing and remediation. Six weeks out we can still do meaningful work, but the scope narrows to observability, pre-scaling and a tested rollback plan.
We do not run structural change into a peak window. Modernisation happens in the trough, with progressive delivery and a rollback path proven before each release.
For a narrow proof of value, sometimes. For anything you would run the business on, no — the forecasting work and the data foundation run together, which is why we scope them together.
Yes. Most of our retail work sits around a commercial commerce platform rather than replacing it — the elasticity, forecasting and cost work is largely independent of that choice.
Find the bottleneck before your customers do
A peak readiness assessment takes three weeks and ends with a ranked list of what will break, at what load, and what each fix costs.