A large directorate in a central government department was delivering an ambitious transformation of how citizens interact with government services. It was the first large-scale user of AWS in the UK, and from 2017 to 2022 the largest government consumer of AWS services in the country — averaging around $20m a year on hosting.
Rapid adoption came with a bill that rose exponentially. Delivery teams were inexperienced at running cloud compute efficiently, and they had no intuitive tooling to show them where the savings were.
As client-side partner for the directorate, we were accountable for driving down cost-to-complete across the portfolio. So we launched an initiative to bring cloud spend under control: a cost-efficiency strategy built with the platform team, then benefit realisation driven service by service.
// WHAT WE DIDA strategy on four pillars
Under our leadership the programme developed and executed a cost optimisation strategy with four parts. Standards came first, because nothing else works without them.
// PILLAR TWOUsing less compute
Four usage reduction strategies, identified and then rolled out across the portfolio.
These are incremental, not alternatives. A right-sized, auto-scaling service that also powers down overnight compounds all three, and in combination the effect is substantial.
// PILLAR THREEPaying less per unit
AWS charges for the amount of compute used and the rate per unit. Usage reduction attacks the first; rate reduction attacks the second. Beyond enterprise pricing instruments — Savings Plans, Enterprise Discount Programmes — the biggest lever was changing the type of compute.
// PILLAR FOURMaking spend visible
Cost only falls when the people making the decisions can see it. We gave teams meaningful reporting on their own spend, exception reports they could act on tactically, and a view of what their costs would be if they adopted the strategies available to them.
Then we put those reports on the agenda. Efficiency and exception reporting became standing items at project and product delivery boards — which is how cost efficiency stopped being a platform team concern and became a delivery one.
// EXECUTIONTurning strategy into savings
With the strategies and tooling in place, the work became delivery. As client-side partner with both delivery and cost accountability, we led benefit realisation: portfolio-wide prioritisation of cost reduction work, visible dashboards and KPIs to create ownership, targeted and actionable exception reporting, and leaders held to account at delivery boards.
Our people also held day-to-day product management accountability for many of the major products and services in the portfolio, and were directly responsible for the technical implementation inside those teams. The strategy and the execution were not separated.
// RESULTS$8m in year one, around $30m by the end of 2022
Year one realised over $8m of savings. By the end of 2022 the figure was around $30m. Significant future cost was avoided on top of that, because new services were now being built cost-efficiently from the start.
Two firsts came out of the programme, and AWS holds the overall approach up as best in class.
The first UK government department to run Spot in production
Spot was the biggest rate lever, but it needs architecture that can lose an instance without drama. Working with the architecture teams, we drove a large shift to containerised Kubernetes services — immutable by design. Having proven graceful handling of Spot reclamation in non-production, we rolled it out to production services: the first government department globally to do so at scale, at around 70% less than standard EC2 pricing.
Shutting down 18,000 containers with one button
We introduced self-service shutdown orchestration, integrated with Slack, so teams could spin down every development and test environment overnight and at weekends. A single action powers down around 18,000 containers and 500 RDS databases, and brings them all back on a schedule in time for the next working morning — saving up to 64% of usage cost.
"Held as a beacon of success and execution by AWS globally."
— AWS GLOBAL ACCOUNT LEAD, ON THE COST REDUCTION INITIATIVE
// THE LESSONCost efficiency is a delivery discipline
The tooling and the pricing instruments are available to everyone. What made the difference here was accountability: standards teams had to meet, reporting they could act on, and cost as a standing item where delivery decisions actually get made.
Architecture matters too. Spot at 70% off is only available to organisations whose services can survive losing a node — so in the end, the cost saving was an engineering outcome.
