Native or Cross-Platform? A Practical Way to Decide in 2026
The honest answer depends on four specific factors. Here is the decision framework we use with clients, and where each...
Read article
Most cloud overspend comes from a handful of predictable causes. Here are the ones we find on nearly every audit, in order of typical savings.
Cloud bills grow quietly. Nobody deliberately overspends - instances get provisioned for a launch and never revisited, and by the time anyone looks, the monthly figure has doubled. Here is what we consistently find.
1. Over-provisioned compute - typically 15 to 25 percent
Instances sized for imagined peak load that never arrived. We routinely find production servers running at four percent average CPU. Look at ninety-day utilisation and right-size against the p95, not the theoretical maximum.
2. Orphaned resources - typically 5 to 15 percent
Unattached storage volumes, old snapshots, idle load balancers, and unused elastic IPs. One client was paying for 3.2 TB of volumes attached to instances terminated two years earlier. Tag everything with an owner and expiry, and audit monthly.
3. No commitment discounts - typically 20 to 30 percent
Steady-state workloads running entirely on on-demand pricing. Reserved instances or savings plans on your genuine baseline are the single largest lever available, and carry no technical risk. Cover the baseline only; keep bursts on demand.
4. Non-production running 24x7 - typically 5 to 10 percent
Development, staging and QA environments running through nights and weekends. A scheduled shutdown outside working hours removes about seventy percent of their cost with no impact on anyone.
5. Storage in the wrong tier - typically 3 to 8 percent
Logs and backups from three years ago sitting on hot storage. Lifecycle policies moving them to infrequent-access and then archive tiers is a one-time configuration change.
6. Data transfer surprises - varies widely
Cross-availability-zone chatter and egress that nobody budgeted for. A CDN in front of static assets and keeping talkative services in the same zone usually fixes it.
Make it stick
One-off optimisation decays within months. Set a monthly budget alert, add cost review to your sprint cadence, tag resources by team, and require an owner and expiry on anything new. The discipline is worth more than the initial cleanup.
Share this article