Engineering Note
How we reduced AWS cost by 40% in a SaaS system
Reducing AWS cost is not always an infra problem.
In one SaaS system we were running, costs kept increasing with growth.
Nothing unusual in metrics.
But monthly bill said otherwise.
We didn't start with scaling down resources.
We looked at usage patterns.
What we found:
- Idle instances running 24/7
- Over-provisioned databases
- Background jobs running more frequently than needed
- Large Docker images increasing compute time
Cost was not from traffic.
It was from inefficiency.
What we changed:
- Right-sized instances based on real load
- Scheduled non-critical workloads
- Optimised job frequency and batching
- Reduced Docker image size
- Cleaned unused resources
No major architecture change.
Just better utilisation.
Outcome:
- Around 40% reduction in AWS cost
- More predictable monthly billing
- No impact on performance
Important learning:
Scaling cost is not linear with users.
It is driven by how efficiently systems are used.
In SaaS, cost optimisation is an engineering responsibility.
In one SaaS system we were running, costs kept increasing with growth.
Nothing unusual in metrics.
But monthly bill said otherwise.
We didn't start with scaling down resources.
We looked at usage patterns.
What we found:
- Idle instances running 24/7
- Over-provisioned databases
- Background jobs running more frequently than needed
- Large Docker images increasing compute time
Cost was not from traffic.
It was from inefficiency.
What we changed:
- Right-sized instances based on real load
- Scheduled non-critical workloads
- Optimised job frequency and batching
- Reduced Docker image size
- Cleaned unused resources
No major architecture change.
Just better utilisation.
Outcome:
- Around 40% reduction in AWS cost
- More predictable monthly billing
- No impact on performance
Important learning:
Scaling cost is not linear with users.
It is driven by how efficiently systems are used.
In SaaS, cost optimisation is an engineering responsibility.